The Pennsylvania State University
The Graduate School
Department of Energy and Mineral Engineering
STUDY OF UTILIZATION FACTOR AND ADVANCE RATE OF HARD ROCK TBMS
A Dissertation in
Energy and Mineral Engineering
by
Ebrahim Farrokh
2012 Ebrahim Farrokh
Submitted in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
May 2013
The dissertation of Ebrahim Farrokh was reviewed and approved* by the following:
Jamal Rostami Assistant Professor of Energy and Mineral Engineering
Mark S. Klima Department Head, Associate Professor of Mineral Processing and Geo-Environmental Engineering
R. Larry Grayson Professor of Energy and Mineral Engineering
Antonio Nieto Associate Professor of Energy and Mineral Engineering
Prasenjit Basu Assistant Professor of Civil and Environmental Engineering
*Signatures are on file in the Graduate School
iii
ABSTRACT
Estimating the penetration rate (PR), utilization (U), and advance rate (AR) is a critical factor in successful selection and application of tunnel boring machines (TBM), but it has remained a challenge to most engineers and contractors. While there have been many studies on accurate prediction of penetration rate with some progress in accounting for various geological parameters, the amount of research performed on TBM utilization and advance rate is still very limited.
The primary objective of this research was to develop a comprehensive database of TBM utilization and advance rate from different hard-rock tunneling projects using a TBM to develop a new model for estimation of machine utilization and advance rate through statistical analysis of available machine field performance information and a new rock mass characterization system.
For this purpose, information for 300 tunnel projects, including rock properties, TBM specification, TBM operational parameters, and achieved performance were compiled in a database to seek significant correlations between these parameters. As the results of statistical analyses show, Unconfined Compressive Strength (UCS), Cerchar Abrasion Index (CAI), and
Rock Quality Designation (RQD) are the most influential parameters in estimation of PR. For utilization factor, PR, UCS, groundwater condition, and tunnel diameter are the most influential parameters. The results of the analyses also indicate that tunnel diameter and UCS are among the primary parameters for prediction of advance rate. Good correlations between the actual and predicted values with R-sq of more than 60% have been obtained in different statistical analyses.
In this thesis, two methods for prediction of AR are offered. One comes from multiplication of the predicted PR and U (indirect methodology) and the other method comes from direct estimation of AR from input parameters. Even though these two ARs are not exactly the same (since the methodologies and the input parameters are different), the results are
iv reasonably close to each other. Although the indirect methodology has more flexibility in changing the related conditions for different down time components, it might produce more errors due to the combination of many parameters. On the other hand, the direct methodology benefits from more tunnel records compared to the indirect methodology and has more reliability. Hence, the direct method is proposed as the primary method for AR prediction and the indirect method is proposed as the supporting methodology to be used for more in-depth estimation using the analysis of different down time components.
The results of analyses for the learning phase period show that, on average, it takes one week per each meter of tunnel diameter. This should be reflected as an adjustment on the results of the analyses for the first sections of the tunnel after completing the AR prediction for the entire length of a tunnel. In order to account for parallel activities and the probability of the parameters values, four different models were developed and calibrated for the real data. The results of simulation modeling show a very good agreement with the actual values of TBM advance rate values.
The outcome of this study is establishing a framework to assist in the prediction of the main components of TBM performance parameters. This will provide an extremely useful tool for developing reliable estimates of the machine advance rate used in estimating project time and costs for a specific site, ground condition, and machine type when all input variables are properly assigned.
v
TABLE OF CONTENTS List of Figures...... ix List of Tables...... xvi Acknowledgements...... xx
Chapter 1 Introduction ...... 1
Introduction and Problem Statement ...... 1 Scientific Objectives and Intellectual Merit ...... 4 Background of Studies on TBM Performance Prediction ...... 5 Scope of Work and Methodology in Performing the Research...... 6 Structure of the Thesis ...... 10
Chapter 2 Literature Review on Hard Rock TBMs ...... 12
Introduction ...... 12 Tunnel Boring Machine (TBM) ...... 14 Open TBM System Description ...... 17 Single Shield TBM ...... 18 Double Shield TBM ...... 20 TBM Performance Parameters ...... 22 Penetration Rate ...... 23 Utilization Factor and Advance Rate ...... 23
Chapter 3 TBM Performance Databases ...... 25
Introduction ...... 25 General TBM Field Performance Database ...... 26 Detailed Database ...... 29 Data Screening ...... 34 Review of TBM Performance Parameters in the Database ...... 37 TBM Operational Parameters ...... 37 Tunnel Location and Application ...... 39 Tunnel Diameter ...... 40 Length of the Tunnel ...... 43 Unconfined Compressive Strength ...... 44 Core Fracture Frequency (CFF) ...... 46 Groundwater Condition ...... 46 Rock Type ...... 47 Major Mining Problems ...... 48 Geological Variability ...... 49 Quartz Content ...... 50 Cutter Diameter ...... 51 Ground Support ...... 51 Tunnel Transport and Muck Haulage System ...... 53 Tunnel Access ...... 54 Year of Completion ...... 55
vi
TBM Condition ...... 56 TBM Type ...... 57 Tunnel Slope ...... 57 Rock Mass Characteristics ...... 57
Chapter 4 Direct Advance Rate Prediction ...... 59
Introduction ...... 59 AR Prediction Methods ...... 61 RME Model...... 63 Evaluation of Previous Advance Rate Prediction Models ...... 67 Proposed New Model ...... 72 Linear Regression Analysis...... 76 Discussion and Conclusions ...... 83
Chapter 5 Penetration Rate Prediction ...... 85
Introduction ...... 85 Objective or Target Parameter ...... 87 Evaluation of Existing Penetration Rate Models ...... 91 Proposed New Model ...... 99 Multivariate Regression Analysis ...... 100 Discussion and Conclusions ...... 102
Chapter 6 Downtime Analysis ...... 105
Introduction ...... 105 Description of the TBM Field Performance Database ...... 106 Comparison the Reported Downtimes with Existing Predictive Models ...... 112 CSM Method ...... 112 NTNU Method ...... 115 Ribacchi and Lembo Fazio’s Proposed Method ...... 117 Proposed Modifications for Estimation of Various Activity Times...... 119 Boring Time ...... 121 Regrip Time ...... 121 Cutter Change Time ...... 121 TBM Repair Time ...... 122 Back-Up Repair Time ...... 123 Maintenance ...... 124 Surveying Downtime...... 125 Utility Installation Downtime ...... 125 Transport Related Downtime ...... 125 Ground Support Installation Downtime ...... 126 Groundwater Condition Related Downtime ...... 127 Other Downtimes ...... 128 Comparative Study for Advance Rate Prediction ...... 128 Discussion of Proposed Method for Estimating Utilization of Hard Rock TBMs ...... 131
Chapter 7 Unit Supporting Time (UST) and Support Installation Time (SIT) ...... 133
vii
Introduction ...... 133 Ground Support Types ...... 136 Tunnel Support Classification ...... 139 Unit Supporting Time (UST) ...... 144 Support Installation Time (SIT) ...... 152 Comparison between O-Norm and Other Rock Mass Classifications ...... 156 Summary and Discussions ...... 160
Chapter 8 Cutter Change Time and Cutter Consumption ...... 162
Introduction ...... 162 Cutter Change Time ...... 165 Number of Changed Cutters and Estimation of Cutter Change Time ...... 170 Cutter Change Time and Penetration Rate ...... 172 Cutter Life ...... 173 Discussion and Conclusions ...... 181
Chapter 9 Learning Phase ...... 183
Introduction ...... 183 Modeling LPP Effect ...... 186 New Methodology ...... 192 Evaluation of Proposed LPP Function Parameters ...... 196 Summary and Discussions ...... 200
Chapter 10 Simulation of Tunneling Activities for Prediction of Machine Utilization ...... 201
Introduction ...... 201 Modeling of Random Variable...... 205 Background of TBM Performance Prediction with Simulation ...... 208 Search Methods ...... 211 Rock Mass Simulation Method ...... 214 Track-bound System Simulation ...... 216 Track-Bound Transport ...... 216 Simulation of the Activities in a Tunnel Project ...... 228 First Simulation Approach ...... 229 Data Analysis ...... 233 Second Simulation Approach ...... 235 Case 1 Data Analysis ...... 243 Case 2 Data Analysis ...... 244 Case 3 Data Analysis ...... 245 Discussions...... 246
Chapter 11 Conclusions and Recommendations ...... 247
Conclusions ...... 247 Recommendations ...... 251
REFERENCES ...... 254
viii
APPENDIX A Database Projects ...... 267 APPENDIX B Typical Minitab Outputs ...... 277
ix
LIST OF FIGURES
Figure 1- 1. Steps for methodology in performing the research ...... 8
Figure 1- 2. General overview of input and output parameters ...... 9
Figure 2- 1. Picture of a typical hard rock TBM and a single disc cutter ...... 12
Figure 2- 2. Machine selection with Rock Mass Rating (RMR) (After Robbins, 1995) ...... 15
Figure 2- 3. Various hard rock TBMs (Maidl et al, 2008) ...... 16
Figure 2- 4. Open TBM system descriptions (WBI, 2007) ...... 17
Figure 2- 5. Open TBM operation cycle (WBI, 2007) ...... 18
Figure 2- 6. Single shield TBM (Herrenknecht, 2010) ...... 19
Figure 2- 7. Single shield TBM operation cycle (WBI, 2007)...... 20
Figure 2- 8. Double shield TBM (Herrenknecht, 2010) ...... 21
Figure 2- 9. Double shield TBM operation cycle (WBI, 2007) ...... 22
Figure 3- 13. TBM utilization/advance rate versus tunnel length ...... 44
Figure 3- 14. ARw/Uw versus UCS...... 45
Figure 3- 15. Cutter wear rate versus UCS ...... 45
Figure 3- 16. RQD versus ARw/Uw ...... 46
Figure 3- 17. Water condition versus Uw/ARw ...... 47
Figure 3- 18. Rock type versus ARw/Uw ...... 48
Figure 3- 19. Geologic variability versus ARw/Uw ...... 50
Figure 3- 20. Quartz content versus ARw/Uw ...... 50
Figure 3- 21. Cutter diameter versus Uw/ARw ...... 51
Figure 3- 22. Support category versus Uw/ARw ...... 53
Figure 3- 23. Support range versus Uw/ARw ...... 53
Figure 3- 24. Tunnel transportation (or muck evacuation system) versus ARw/Uw ...... 54
Figure 3- 25. Shaft depth category (Shd) versus ARw/Uw ...... 55
x
Figure 3- 26. Hole through tunneling year histogram and hole through tunneling versus ARw/Uw ...... 56
Figure 3- 27. TBM type versus Uw/ARw ...... 57
Figure 4- 1. RME index parameters ratings (Bieniawski et al., 2006, 2008) ...... 64
Figure 4- 2. Correlation between RME index and the average rate of advance for double- shield TBMs (redrawn from Bieniawski et al., 2008) ...... 65
Figure 4- 3. Adaptation correction factor (Bieniawski et al., 2008) ...... 66
Figure 4- 4. RME and TBM AR relationship at GIBE outlet drive (data obtained from Grandori, 2007) ...... 66
Figure 4- 5. Results of comparison for different models ...... 68
Figure 4- 6. Inverse correlation between downtime components and penetration rate in NTH model (Bruland, 1988) ...... 69
Figure 4- 7. ARr and ARt for RME model using testing database ...... 70
Figure 4- 8. Correlation between PR and ARw ...... 73
Figure 4- 10. Correlation between PR and UCS for different rock types (Redrawn from Robbins, 1992) ...... 76
Figure 4- 11. Results of linear regression analyses for ARw (Log values) for 2.5
Figure 4- 12. Results of linear regression analysis for ARw (Log values) for the normal conditions ...... 79
Figure 4- 13. Correction factor for base advance rate in different RMR values and TBM types ...... 79
Figure 4- 14. Comparison between six pairs of ARw for similar tunneling conditions (on the basis of ARw of 12 tunnel cases) ...... 81
Figure 4- 15. The comparative results of new model between predicted and actual AR ...... 81
Figure 4- 16. The comparative results of new proposed model and RME model...... 82
Figure 5- 1. Relationship between penetration per revolution (PRev) and cutter normal force (Fn) (modified from Laughton, 1998) ...... 89
Figure 5- 2. Operating limits of a TBM with 17″ disc cutters at different rock strengths after Frenzel et al. (2008) ...... 91
Figure 5- 3. Results of comparison for the selected models ...... 94
xi
Figure 5- 4. Recreated example of the neuro-fuzzy model recreated in an Excel sheet from Alvarez et al. (2000a) ...... 98
Figure 5- 5. An example of the recreated neuro-fuzzy model (Alvarez et al., 2000a) yielding negative value ...... 98
Figure 5- 6. Correlation between FPI and other parameters ...... 100
Figure 5- 7. Results of linear regression analysis for FPI (response is Ln(FPI)) ...... 101
Figure 5- 8. The comparative results of Eq. 5-2 ...... 102
Figure 6- 1. Histograms of different information of the database ...... 107
Figure 6- 2. Histograms of allocated downtimes for different activities for Open TBM (in %) ...... 110
Figure 6- 3. Histograms of allocated time for different activities for single shield TBM (in %), different hatches refer to different projects ...... 110
Figure 6- 4. Histograms of allocated time for different activities for double shield TBM (in %), different columns in each category refer to different projects ...... 111
Figure 6- 5. Comparison between reported and predicted values of different time items of EMI model ...... 114
Figure 6- 6. Comparison between reported and proposed values of different activity time using NTH model ...... 117
Figure 6- 7. Distributions of coefficients values proposed by Ribacchi and Lembo Fazio (2004) in the database ...... 118
Figure 6- 8. Utilization values and boring hour for the different UCS values ...... 120
Figure 6- 9. Cutter downtime, Tc ...... 122
Figure 6- 10. TBM downtime, Ttbm...... 123
Figure 6- 11. Back-up downtime, Tbu ...... 124
Figure 6- 12. Ground support downtime, Tsp ...... 127
Figure 6- 13. Downtime related to water inflow, Tw ...... 128
Figure 6- 14. Indirect methodology for AR prediction ...... 129
Figure 6- 15. Results of comparative study for predicted direct and indirect ARw ...... 130
Figure 7- 1.Normalized tunneling advance rate as a function of support requirement for various TBM types (Schmid, 2004 and Maidl et al., 2008) ...... 134
xii
Figure 7- 2. Misicuni water-transmission tunnel, excavation and support classes and rates of advance (WBI, 2007) ...... 135
Figure 7- 3. Open TBM and different facilities for support installation in working area behind the cutter head (Herrenknecht web, 2010) ...... 138
Figure 7- 4. Support installation behind the cutter head in Open TBM (Tunneltalk web, 2010; Wallis, 2009; Youtube web, 2010) ...... 138
Figure 7- 6. Average supporting time/percent for different case histories ...... 146
Figure 7- 7. Flowchart of calculating UST for Austrian rock mass classes of F1 to F7 using an Excel solver ...... 147
Figure 7- 8. Open TBM support time for different Austrian rock mass classes ...... 148
Figure 7- 9. Best-practice case histories of Open TBM (Laughton, 1998) ...... 149
Figure 7- 10. Comparison of UST values of some tunnel projects with the obtained range of the UST ...... 151
Figure 7- 11. Flowchart of calculating the SIT for different support types using an Excel solver ...... 153
Figure 7- 12. TBM activity time distribution for four Open TBMs ...... 154
Figure 7- 13. Open TBM support installation time for different support types ...... 154
Figure 7- 14. SIT for tunnel project in New York City (diameter 3.65 m completed in 2005) ...... 155
Figure 7- 15. Comparison of NYC tunnel SIT values and the range of the SIT ...... 155
Figure 7- 16. Correlation between UST and some rock mass classifications for Manapouri tunnel ...... 157
Figure 7- 17. Equivalent RMR values for different F classes based on the general trend of the points ...... 159
Figure 8- 1. Typical disc cutters and their arrangement on the cutter head ...... 163
Figure 8- 2. Rock breaking process under a disc cutter (Buchi, 1984) ...... 164
Figure 8- 3. Cutter change time for 36 hard-rock TBM tunnel projects ...... 166
Figure 8- 4. Correlation between cutter consumption and cutter change time ...... 166
Figure 8- 5. Correlation between cutter consumption and cutter change time for different cutter size categories ...... 168
xiii
Figure 8- 6. Scatter of the data points of tc for different cutter sizes and tunnel diameter sizes (solid line is the average slope parameter of the regression analysis) ...... 169
Figure 8- 7. Extreme cutter change time values in the data set ...... 169
Figure 8- 8. Cutter change time for different number of cutters changed at one stop ...... 171
Figure 8- 9. Cutter change time ...... 173
Figure 8- 10. CAI values with average and standard deviation for different rock types (Buchi, 1984) ...... 174
Figure 8- 11. Graph of real cutter life versus predicted cutter lifer (using Eq. 8-3) in m3/cutter ...... 175
Figure 8- 12. Relationship between CAI/Cp with real excavated volume per ...... 175
Figure 8- 13. Graphs of NTH method used in cutter consumption (Redrawn from Bruland, 1998a and 1998b) ...... 177
Figure 8- 14. Graph of actual cutter life versus predicted cutter life (using Eq. 8-3) in m3/cutter ...... 177
Figure 8- 15. Rock type, quartz content, and UCS versus cutter consumption ...... 179
Figure 8- 16. Cutter consumption as a function of UCS and quartz content ...... 180
Figure 9- 1. Schematic sketch of LPP and NPP for two common TBM progress diagrams ... 183
Figure 9- 2. Ring building and cycle times according to the Herrenknecht control system (Waise and Wachter, 2009) ...... 185
Figure 9- 3. Histogram of Start-Up efficiency for gripper TBMs according to Laughton (1998) ...... 186
Figure 9- 5. Learning phase over-estimation because of TBM progress fluctuations of normal weeks (using Waise and Wachter fitting function) ...... 191
Figure 9- 6. Learning phase over-estimation because of concave shape of learning curve function (using Waise and Wachter fitting function) ...... 191
Figure 9- 7. Schematic fitting functions elements ...... 193
Figure 9- 8. Flowchart of finding the parameters of fitting functions ...... 194
Figure 9- 9. Areas under fitting functions ...... 195
Figure 9- 10. Examples of obtained fitting functions ...... 195
Figure 9- 11. Comparison between linear and exponential functions for LPP ...... 197
xiv
Figure 9- 12. Histogram of distribution of learning phase ratio for 44 tunnel projects ...... 197
Figure 9- 13. Analysis results for obtaining X1/D ...... 199
Figure 10- 1. Schematic of TBF, TTR, and DBF with reference to different counters (Abd-Al-Jalil, 1998) ...... 202
Figure 10- 2. Schematic of TBF, TTR, and DBF with reference to different counters for the case in which the exact occurrence time of a delay is not specified ...... 204
Figure 10- 3. Process and outcomes of a typical simulation modeling ...... 205
Figure 10- 4. Different methods of TBM TTCT prediction (revised from Laughton, 1998) .. 210
Figure 10- 5. Typical outcomes of probability method for prediction of TTCT (Laughton, 1998) ...... 211
Figure 10- 6. Method A of TBM performance prediction with data base level 1 (Laughton, 1998) ...... 212
Figure 10- 7. Method B of TBM performance prediction with data base level 2 (Laughton, 1998) ...... 212
Figure 10- 8. Probabilistic estimation of a zone length to be used in simulation of method B (Laughton, 1998) ...... 213
Figure 10- 9. Rock mass simulation steps (Laughton, 1998) ...... 215
Figure 10- 10. Schematic of California switch and locomotive (upper figure from Maidl et al., 2008)...... 218
Figure 10- 11. Schematic view of different stages of tunneling using two trains without any California switch ...... 218
Figure 10- 12. Schematic view of different stages of tunneling using three trains with one California switch ...... 219
Figure 10- 13. Schematic view of exchanging trains on the California switch ...... 220
Figure 10- 14. Schematic view of exchanging trains on the portal switch when there is one loco with two trains ...... 221
Figure 10- 15. Schematic view of different stages of tunneling using four trains with two California switches ...... 222
Figure 10- 16. Schematic view of different stages of tunneling using five trains with three California switches ...... 225
Figure 10- 17. Final locations of different California switches (CS) for a typical tunnel with regripping time of 5 min and train speed of 15 km/hr ...... 227
xv
Figure 10- 18. Location of Karaj water conveyance tunnel, northwest of Tehran, Iran (Hassanpour et al., 2009) ...... 228
Figure 10- 20. Examples of fitted functions on the data for different categories ...... 231
Figure 10- 21. Simulation model in Arena software ...... 232
Figure 10- 22. TBF sampler model ...... 233
Figure 10- 23. Curve fitting on the utilization data for obtaining the end of the start-up phase ...... 234
Figure 10- 24. Typical Open TBM operation cycle ...... 236
Figure 10- 25. Typical single shield (SS) TBM operation cycle ...... 237
Figure 10- 26. Typical double shield (DS) TBM operation cycle ...... 238
Figure 10- 27. Simulation model of tunneling by double shield TBM using two trains without any California switch (case 1) ...... 240
Figure 10- 28. Simulation model of tunneling by double shield TBM using three trains with one California switch (case 2) ...... 241
Figure 10- 29. Simulation model of tunneling by double shield TBM using four trains with two California switches (case 3) ...... 242
Figure 10- 30. Simulation model results of tunneling by double shield TBM using two trains without any California switch (case 1) ...... 243
Figure 10- 31. Simulation model results of tunneling by double shield TBM using three trains with one California switch (case 2) ...... 244
Figure 10- 32. Simulation model results of tunneling by double shield TBM using four trains with two California switches (case 3) ...... 245
xvi
LIST OF TABLES
Table 2- 1. TBM advantages and disadvantages (modified from Maidl et al., 2008)...... 14
Table 2- 2. Machine application range according to Askilsrud (1996) ...... 15
Table 3- 1. Summary of the parameters included in the developed database ...... 27
Table 3- 2. Database levels and schematics of distance increments in each level (modified from Nelson et al., 1999) ...... 28
Table 3- 3. General information of the tunnel projects in testing database ...... 30
Table 3- 4. CFF categorization ...... 46
Table 3- 5. water condition categorization ...... 47
Table 3- 6. Rock type categorization in data base ...... 48
Table 3- 7. Major mining problems categorization ...... 49
Table 3- 8. Geological variability categorization ...... 49
Table 3- 9. Quartz content categorization ...... 50
Table 3- 10. Basic support categorization ...... 52
Table 3- 11. Support range changing categorization ...... 52
Table 3- 12. Tunnel transport system categorization ...... 54
Table 3- 13. Tunnel access categorization ...... 54
Table 4- 1. Advance rate prediction models and their advantages and disadvantages ...... 62
Table 4- 2. RME index parameters (Bieniawski et al., 2006, 2008) ...... 64
Table 4- 3 Criteria for evaluation of coefficients FE1, FE2 and FE3 (Bieniawski et al., 2008, Grandori, 2007) ...... 66
Table 4- 4. Rock type categorization in database (Laughton, 1998) ...... 75
Table 4- 5. Regression coefficient statistics for advance rate prediction for normal condition...... 78
Table 5- 1. Advantages and disadvantages of empirical TBM performance prediction models ...... 86
Table 5- 2. TBM penetration rates ...... 88
xvii
Table 5- 3. TBM operational parameters in different settings ...... 90
Table 5- 4. TBM performance models ...... 92
Table 5- 5. Regression coefficient statistics for FPI prediction ...... 101
Table 6- 1. Summary of the tunnel projects in the database ...... 106
Table 6- 2. Downtime categories identified in different tunnel projects ...... 108
Table 6- 3. Prediction of TBM utilization using CSM method (Ozdemir and Sharp, 1991; US Army, 1997) ...... 113
Table 6- 4. NTH method for prediction of the TBM utilization (Johannessen et al., 1988, 1994, US Army, 1997) ...... 116
Table 6- 5. General maintenance downtime in different conditions ...... 125
Table 6- 6. Muck transport downtime in different conditions ...... 126
Table 6- 7. Comparative study for direct and indirect AR prediction methods for 12 tunnels ...... 130
Table 7- 1.Tunneling classes for TBMs developed in Switzerland (SIA, 1993) ...... 139
Table 7- 2. Tunneling classes for TBMs proposed by Maidl et al. (2008) ...... 140
Table 7- 3. Rock classes and their corresponding support measurement chart developed by Ilbua from Austrian Onorm rock classification system to combine NATM with TBM tunneling (Wallis, 1993 and Scolari, 1995) ...... 143
Table 7- 4. F classes and their support requirements for the Alassio tunnel (3.6 m diameter) in Italy based on observation (GEOTEST, 1993) ...... 144
Table 7- 5. Tunnel case histories used in calculation the UST range ...... 147
Table 7- 6. UST for the Gossensas tunnel ...... 149
Table 7- 7. UST and F classes for the Alpe Devero delivery tunnel (Devero to Agaro lake) ...... 152
Table 7- 8. UST and F classes for the Alpe Devero delivery tunnel (Bodolero to Cairasca) .. 152
Table 7- 9. Comparison of different rock mass classification methods with F-class (Atlas Copco, 2005) ...... 156
Table 7- 10. Equivalent RMR values for different F classes based on Manapouri UST (3.9) ...... 159
Table 8- 1. Cutter diameter and bearing load capacity (Roby et al., 2008) ...... 162
xviii
Table 8- 2. General specifications of tunnel projects used for analysis of cutter change time and change cutter in the group...... 171
Table 8- 4. Quartz content categorization in the database ...... 178
Table 8- 5. Rock type categorization in the database...... 178
Table 9- 1. Learning conditions rating according to Waise and Wachter (2009) ...... 188
Table 9- 2. Corresponding values in standard to good conditions ...... 189
Table 9- 4. LR rating in different learning conditions ...... 198
Table 9- 5. Summary of formulas and calculation results ...... 200
Table 10- 1. Continuous distributions and their parameters (Summarized from Altiok and Melamed, 2007) ...... 206
Table 10- 2. Search criteria for Manhatan South Tunnel like-sites ...... 214
Table 10- 3. Performance parameters prediction for Manhatan South Tunnel (based on 12 cases) ...... 214
Table 10- 4. Average number of kilometers per EMA (Calculated from level 2 database of Nelson et al. (1994)) ...... 216
Table 10- 5. Demands for locomotive and muck train ...... 217
Table 10- 6. Transportation time components of different sections of a tunnel using three trains with one California switch ...... 219
Table 10- 7. Transportation time components of different sections of a tunnel using four trains with two California switches ...... 223
Table 10- 8. Transportation time components of different sections of a tunnel using five trains with three California switches ...... 226
Table 10- 9. Karaj TBM specification (Hassanpour et al., 2009) ...... 229
Table 10- 10. Best fit function for delays for each ring ...... 230
Table 10- 11. Best fit function for daily delays ...... 230
Table 10- 12. Best fit function for TBF and TTR ...... 230
Table 10- 13. The results of simulation model for TTCT and utilization (U) prediction for reaches 2 to 4...... 234
Table 10- 14. Results of simulation model for case 1 ...... 243
xix
Table 10- 15. Results of simulation model for case 2 ...... 244
Table 10- 16. Results of simulation model for case 3 ...... 245
xx
ACKNOWLEDGEMENTS
I would like to express my thanks to all individuals who have encouraged me and helped me during my study to accomplish my goals in my graduate studies at Penn State University.
Special thanks are due to Dr Rostami, my thesis advisor, for guiding and supporting me during my studies and for providing me with invaluable information, recommendations, and suggestions for accomplishing my research goals. Thanks are also due to Dr Grayson, Dr Nieto, Dr Basu, and
Dr Palomino, the committee members of my thesis, for their support and advices. Thanks are due to National Science Foundation (NSF), Robbins company, and Frontier Kemper company for their support of my research study. Many thanks are due to professional researchers who graciously shared their data for using in my analysis. In this regard, I would like to express my special thanks to Mr Askilsrud, Dr Laughton, Dr Kim, Dr Hassanpour, and Dr Ramazanzadeh.
Most importantly, many thanks are due to my family who patiently support me throughout my study at Penn State University. Words alone cannot express my thanks that I always owe them. In this regard I would like to specially thank my father and my mother who always support me and encourage me throughout my life. Special thanks are also due to my wife for her invaluable helps and encouragements.
Lastly, I offer my regards to all EME faculty and staff at Penn State University who supported me in any respect during the completion of my studies.
1
Chapter 1
Introduction
Introduction and Problem Statement
Mechanized excavation methods have become very popular in tunneling and underground construction because of many advantages they offer over conventional drill and blast (D/B) methods. The Tunnel Boring Machine (TBM) is one of the most commonly used excavation equipment in tunneling and in the past decade, it has become more or less the standard excavation method in long circular tunnels (typically over 1 km). There are various types of
TBMs for tunneling in nearly all different geological conditions ranging from soil to hard rocks with machines ranging in size from under 2 to over 15 m in diameter. The boring process in TBM tunneling includes continuous excavation at the face, removal of the muck, and installation of ground support and utilities. All tunneling world records in last few decades have been set by
TBMs, and the process has become very efficient and streamlined.
Estimating the rate of penetration (ROP) and tunnel advance rate (AR) is a critical factor in successful selection and application of TBMs, but it has remained a challenge to most engineers and contractors. Penetration rate (PR), which is the same as ROP, is typically expressed in terms of m/hr and is a function of machine specifications, geological parameters, and operational parameters. Meanwhile, AR is the average of the cumulative footage mined per day, estimated from ROP and machine utilization (U), which is defined as the portion of the time that the machine is excavating to the total time. There have been a lot of studies on accurate prediction of ROP, with some progress in accounting for various geological parameters in recent years
(Rostami, 1993 and 1997, Yagiz, 2002, Ramezanzadeh 2005, and Gong 2005), but the amount of
2 research performed on TBM utilization, which represents the efficiency of the TBM during the excavation of different sections of a tunnel, is still very limited.
TBM utilization is controlled by several factors including rock types and rock mass conditions, TBM operational parameters, slope and turns along the alignment, operational restrictions, site-specific conditions, site management issues, and contractor experience. Failure to accurately predict the TBM utilization factor can cause problems regarding development of a realistic plan for a tunneling operation to meet the project goals and better selection of machinery and methods as well as precautionary measures for efficient operation that could achieve high advance rates.
TBM utilization has a direct impact on TBM advance rates and project costs. The relationship between ROP, AR and U can be simply stated as:
AR = ROP x U x T (1-1)
where T represents the working hours of the day. This simply shows that any increase in utilization can directly impact the advance rate. It also means that even in operations where high
ROP can be achieved, the tunneling rate could still fall short of expectations if the machine utilization is low. Alternatively, in very hard rocks where penetration is difficult, a higher utilization can allow for relatively good advance rates. In practice, TBM utilization normally can vary from single digits to 50%. This indicates that in most operations, the portion of the time spent in excavation is less than the time spent in various forms of delays or so-called down times.
As such, it is necessary to thoroughly study TBM downtime and be able to quantify its components for the purpose of estimating the TBM utilization. The components of downtime can be put in several categories as will be discussed in upcoming chapters.
3 It is generally difficult to distinguish which factors are the most significant parameters influencing the overall machine performance. Close examination of the field data from TBM operations clearly indicates the need for a new modeling approach to adequately address this problem. The main reasons for the necessity of such a study include:
1) TBMs are increasingly being used in challenging geological conditions. This refers to increasing use of the machines in adverse ground conditions, very large or very small diameters, deep tunnels with frequently varying rock types, tunnels under heavy groundwater pressure, and tunneling through fault zones, and sometimes without adequate site investigation.
2) TBM technology has significantly improved in the past couple of decades and machines are more sophisticated and have great sensors that can monitor various aspect of the operation. Thus previous prediction models cannot be directly used in performance analysis unless new field data and information is added to their available databases to develop more realistic models.
3) Many new features of the machines and tunneling operations have not been adequately explored to observe the impact of closer monitoring of the operation and on-time response to improve tunneling efficiency. For example new sensors are available on the machines that can spot failed cutters and warn the operators to avoid cutterhead damage and subsequent downtimes.
4) The amount of research on the subject of TBM utilization and advance rate has been very limited due to the complexity of the problem and the additional non-technical issues such as site management that has impeded the development of more inclusive and comprehensive models.
Thus the primary objective of this research is to develop a comprehensive database of
TBM utilization and advance rate from different tunneling projects, especially in hard rock tunneling, to develop a new model for estimation of TBM utilization and advance rate by the aid
4 of statistical methods and a new rating systems. This involves identification of the parameters influencing TBM performance and downtime and understanding the inter-relationship between these parameters.
Scientific Objectives and Intellectual Merit
The main objective of the study can be briefly stated as development of quantitative models for the prediction of TBM advance rate, utilization factor, and its components by analysis of the full production cycle and various activities during the operation. The focus of the study will be on open and shielded rock TBMs, especially double-shield TBMs that have been increasingly used in recent tunneling projects. The model will allow for evaluation of the impacts of various geological parameters, machine specifications, and operational parameters on machine utilization and therefore provide a basis for proper selection of the machine and backup system and optimization of the operational parameters to achieve the highest advance rates.
The outcome of this study is establishing a framework to assist in the prediction of machine utilization. This will provide an extremely useful tool for developing reliable estimates of the machine advance rate used in estimating project costs for a specific site, ground conditions, and machine type at the planning phase of a project.
In brief the main contributions of this research study can be expressed as follows:
1. To update available TBM utilization formulas. At the present time, TBM utilization models are rarely used since they were developed for tunneling conditions relevant to a couple of decades ago and their results are mostly for well-organized TBM tunneling projects (e.g. utilization>40%). Therefore, one of the contributions in this study is to incorporate weaker rock mass/tunneling conditions in the downtime analyses.
5 2. Obtaining better estimates for tunneling activity time components, utilization, and advance rate. Obtaining better estimates of tunnel completion time (obtained from advance rate) offers more realistic estimates to contractors for bidding purposes, as well as helping owners accurately estimate the project time and cost. The accurate estimates can reduce the unexpected time and cost overrun for a tunneling project passing through certain geological zones by adopting a better level of precautionary measures.
3. Improving the accuracy of estimates through the use of more important parameters or more common rock tests.
4. Understanding different time components, effects in different conditions and finding bottlenecks causing lower utilization.
Background of Studies on TBM Performance Prediction
Since the advent of the first use of TBMs in hard rock in the 1950’s, many researchers have made an effort to model TBM performance for prediction of the penetration rate. Facing the challenge of modeling machine-rock interactions, several different approaches were adopted for this purpose. One approach was to perform field studies to develop empirical models, and the other was to run laboratory tests, including full-scale cutting tests, to estimate machine performance from the measured cutting forces and specific energy. Both approaches have proven to be short of desired accuracy for a variety of reasons. Even if the penetration rates could be accurately estimated, the estimation of utilization rates and daily advance has typically been a wild card. Mostly contractors use their experience to assign a utilization rate to the operation which can ignore many of the facts on the ground, and the result is lengthy claims and related costs.
6 Literature review shows that the amount of work on prediction of TBM utilization and advance rate is very limited and the number of research studies is quite low compared to the volume of research on prediction of the ROP. The existing systematic work on prediction of TBM utilization includes the work by the Earth Mechanics Institute (EMI) of Colorado School of
Mines (CSM) that can be found in Ozdemir and Sharp (1991), the Norwegian Institute of
Technology (NTH) (Jonhannssen, 1988, 1994, Bruland, 1998), the QTBM method (Barton, 2000), the neural network method (Alvarez, 2000), and the Fuzzy Logic-Based Utilization Predictor
Model (Kim, 2004).
Scope of Work and Methodology in Performing the Research
The scope of this research includes various literature reviews of existing TBM utilization prediction methods, compilation of field performance records in a database, statistical analysis of
TBM utilization records within various rock mass and site conditions, and the development of pertinent models for estimation of machine utilization daily advance rates. The study includes: preliminary selection of input-output variables in the model, classification of the knowledge domain, a thorough analysis and interpretation of the results, and comparison of the results with existing TBM field data for validation.
To reach these goals, the following steps of research have been followed:
An extensive literature review on TBM tunneling projects, machine utilization,
prediction models, and modeling techniques in the engineering field was conducted
to identify relevant parameters and to review the current knowledge on the subject
area in detail.
7
Collection and classification of reliable TBM field performance data from various
job sites. The project information include rock mass properties, machine
specifications, operational parameters, and machine performance.
Compiling the data in a database and initial treatment of data by proper
normalization and calculation of performance indices common in this field of study.
Categorizing the data according to machine and ground type for more effective
analysis.
Performing a descriptive statistical analysis which provides an overview of the
database, including the range of variation and distribution of each parameter and the
possible influence (or interaction) of each pair of parameters.
Data analysis and interpretation using statistical methods and regression analysis to
develop proper formulas for estimation of TBM utilization.
Verification of the predictor models with the comparison of actual (or measured)
and predicted values of machine utilization, and developing a guideline for the
proper use of predictor models.
Fine-tuning the results. This includes: 1. Filtering the database for the normal
condition, 2. Finding the formula for the performance parameter for the normal
condition, 3. Introducing a new rating system to fine tune the results on the basis of
detailed information existing in the detail dataset.
In the proposed approach, the main focus is on establishment of a reliable comprehensive database which includes recent TBM tunneling data to develop new models for prediction of
TBM advance rate and utilization of hard rock TBMs. Figure 1-1 shows the schematic view of the flow chart for development and analysis of the TBM field database. Fig. 1-2 shows an expantion on input and output variables used in this research.
8
Literature Review Online Communication with search contractors/owners/researchers
Conferences: Rapid excavation Journals/Magazines: Shift reports Reports: including design tunneling conference (RETC), TUST, Tunnels and and as-built reports International tunneling Tunneling, World association (ITA), ... Tunneling, ....
Identifying important Important points for different TBM Data for different tunnels either for factors/ideas in the field performance models structures/limits stroke by stroke, zone by zone, or of TBM performance entire length
Summarizing information and averaging over geological zones wherever sufficient data is available
Calculating TBM project performance parameters for actual values
Comparing different sources parameters values to distinguish unreliable reported values (high differences) to disregard a tunnel record data
Database development
General database for tunnels with limited Detail database for tunnels with detail data for entire tunnel length* data for different tunnel zones**
* number of tunnel projects≈260 which ** number of tunnel projects=17 which includes includes some tunnel projects with time time components available for most of them components available- Total number of tunnels with available time components is around 90
Modeling phase (70% of all records) Models results verification using a part of database that was not used in the modeling phase
Regression analysis for every response variable using all available information in database the 1. Filtering the database for normal condition 2. Finding the trend/formula for the normal condition 3. Introducing a new rating system to fine tune the results Validation phase (30% of all records) on the basis of detailed information existing in the testing/detail dataset.
Figure 1- 1. Steps for methodology in performing the research
9
Input variables (Xi)
TBM operational: Equipment spec.: TBM type, TBM Intact rock: UCS, Rock mass: Rock type, Thrust, Torque, RPM, condition, Mucking system, Support quartz content Geological variability, water Cutter Size type and range condition, RQD/CFF, EMA type
Tunnel layout: Diameter, *Categorical variables are ranked on the basis of previous literature review and Length, Slope, Access type, performed analysis for verification. Hole through year
Using linear regression method (Minitab Software)/ Neural Output variables Yi=f(Xi) Network/Fuzzy logic methods (Matlab Software) are also used as secondary check/ New rating system
Calculation Penetration Rate Boring Time (PR)
Time TBM components ARa for available time for Utilization Back-Up
ARb for boring days Cutter Advance Rate (AR) ARw for working days Support
ARw for calendar days Regrip
Available time: Time periods in boring days that TBM is available. Transport Working days: Planed days for excavation. Maintenance Ua Ground Ub Utilization (U) Probe Uw Utility Uc Survey Final Results 1. PR Other 2. Time components (especially for understanding bottlenecks of utilization) 3. U either directly or from time components 4. AR=PR*U or directly Obtaining Uw from time components: second approach Units: where: PR (m/hr) U (%) AR (m/day) AR (m/day)= PR (m/hr)*U (%)*24 (hr/day) Uw = Boring Time/Sum(Boring Time, TBM, ..., Other)
Figure 1- 2. General overview of input and output parameters
10 Structure of the Thesis
Chapter 1 presents a brief background on the TBM performance parameters and their importance for successful completion of a tunnel. This chapter also includes the problem statement, research methodology, and contribution of this study to the technical knowledge in this field.
Chapter 2 covers the background and literature review on application of TBMs, classification of different types of hard rock TBMs, and their main performance parameters.
In Chapter 3, the data-gathering process and description of the two field performance databases is offered.
Chapter 4 presents the study on different advance rate (AR) models, evaluation of the existing and proposed models, and the results of a new generated model for AR prediction.
Chapter 5 addresses the study on different penetration rate (PR) models and the results of a new model for penetration rate prediction.
Chapter 6 presents downtime analysis and results of a proposed new model for prediction of different downtime components in various ground conditions and operational settings.
Chapter 7 addresses the analysis of Unit Support Time (UST) and Support Installation
Time (SIT) for Open-type TBMs. This issue is discussed since it is the most important component of downtime that is related to ground conditions.
Chapter 8 explains the results of analysis for cutter change time and cutter consumption for hard rock TBMs. The cutter change component of downtime is related to hardness and abrasivity of the rock and needs to be addressed in various tunneling operations.
Chapter 9 presents evaluation of the Learning Phase Period (LPP) and its effect on the advance rate in the first stage of tunneling with TBM.
11 Chapter 10 explains simulation techniques for evaluation of time to complete tunneling and the advance rate. This is based on combining various components of tunnel activity time and is a new approach to estimation of the TBM advance rate.
Finally, in Chapter 11, summary and conclusions are presented followed by a brief explanation for the limitations of the current study and recommendations for future works.
12
Chapter 2
Literature Review on Hard Rock TBMs
Introduction
The tunneling industry developed rapidly at the second half of the 20th century with the application of the first open gripper TBM developed by James Robbins in 1956 for a sewer tunnel in Toronto. This 3.27-m diameter machine reached advance rates of up to 30 m/day (Maidl et al.,
2008). As a definition, a tunnel boring machine or TBM is a general term that refers to the machine that can excavate a tunnel in a full-face operation and, while the shielded machines are used for weak ground, hard-rock TBMs are used for excavation of more competent rocks. Hard- rock TBMs are equipped with single-disc cutters (see Fig. 2-1).
Figure 2- 1. Picture of a typical hard rock TBM and a single disc cutter
13 Nowadays, TBMs are becoming more and more popular in the excavation of various tunnel types and various geology conditions because of several advantages of mechanized excavation, especially the rate of advance of the tunnel face over the drill-and-blasting method.
TBM performance is a function of machine design, geological conditions, and site management which is a function of the experience of the contractor. The key to successful planning of TBM tunneling is accurate estimation of the project duration and costs. However, in many projects, the uncertainty of machine performance due to complex geological conditions and unforeseen ground conditions is a factor that has to be considered.
Despite the many successful applications of TBMs in various tunnel projects, TBMs are often challenged, especially in unfavorable geological conditions, with a low penetration rate and prolonged delays. Part of the problem is a lack of understanding of the impact of different geological parameters and thus inaccurate prediction of TBM performance parameters in such projects. Other sources of inaccuracies are related to experience of the contractors and their crews. To rectify this problem, some researchers tried to develop more reliable estimates of TBM performance for project bidding and planning purposes. Inaccurate estimates of TBM performance parameters can cause significant project delays and result in potentially large construction claims (Kim, 2004).
There are three types of hard-rock TBMs including Open, Single-Shield, and Double-
Shield machines. A brief summary of various types of machine as well as operational parameters and machine selection issues for these machine types in different rock mass conditions is presented in this section.
14 Tunnel Boring Machine (TBM)
Hard-rock TBMs are used in a variety of rock types and range in size from just under 2 m to over 14.4 m in diameter. They can negotiate turns and can cut grounds as soft as fault breccia to 300-MPa igneous rocks. Table 2-1 summarizes some of the advantages and disadvantages of the TBM excavation method in comparison with drill-and-blast operation.
Table 2- 1. TBM advantages and disadvantages (modified from Maidl et al., 2008)
Advantage Disadvantage Higher advance rate More geological information needed Exact excavation profile High investment Automated and continual work process Longer lead time for machine designing and manufacturing Low personnel expenditure Specific profile (circular) Better working conditions and safety Limits on curve driving Mechanization and automation of the drive Detailed planning required Limits on adaptation to highly variable rock Limits on adaptation to high water inflow Limits on transportation system
In order to overcome ground complexity and variability, different TBM types have been introduced so far. The range of application of TBMs has expanded considerably in the past two decades. Although a more comprehensive tunneling machine classification and selection chart has been developed by ITA/IATES, for simplicity, it is possible to classify the rock TBMs into three classes according to Barla and Pelizza (2000). These include:
1. Open-type or gripper TBM (for hard and sound rock formations).
2. Single-shield TBM (for weak to very weak rock formations in relatively short tunnels).
3. Double-shield TBM (for weak to hard rock formations in relatively long tunnels).
A general sketch of these TBMs is shown in Fig. 2-3. Tunnel stability is the main factor for the selection of TBM type. Open TBMs are basically used in more stable rock mass conditions while shielded TBMs are used in unstable rock mass conditions with the potential of
15 tunnel collapse. Table 2-2 and Fig. 2-2 show general criteria for TBM selection in different ground conditions.
Figure 2- 2. Machine selection with Rock Mass Rating (RMR) (After Robbins, 1995)
Table 2- 2. Machine application range according to Askilsrud (1996)
Range of Rock Strength Machine Type Diameter (m) Hard Soft Mixed Soft Ground Open TBM 2.5-14 Single Shield TBM 2.5-13 Double Shield TBM 3-12
The process of a typical hard-rock TBM tunnel boring and rock material handling can be summarized as follows:
1. Applying a thrust force on the cutter head and disc cutters.
2. Penetration of cutters into the rock to initiate cracks in the rock to create rock chips.
3. Cutter head rotation to apply torque to dislodge the loose rock chips.
4. Scooping up the rock chips with cutter-head peripheral buckets.
5. Transferring rock material to the cutter-head hopper and then to the conveyor belt.
16 6. Transferring rock material to a tunnel muck transportation system.
7. Unloading the transport system at the tunnel portal.
In the following sections, the process of each TBM type is presented with the primary focus on the main sequence of TBM activities.
Figure 2- 3. Various hard rock TBMs (Maidl et al, 2008)
17 Open TBM System Description
Figure 2-4 shows an open TBM designed for hard-rock excavation. An open or “Gripper”
TBM is suitable for application in a rock mass in which a support of the excavated cross-section in the area of the temporary face and of the machine is not required or may be achieved with minor efforts, e. g. rock bolts, steel sets and shotcrete, applied locally at the roof of the tunnel.
There are two types of gripper system including a single gripper system and a double-gripper system. A double-gripper system is used in larger TBMs especially in more stress-sensitive conditions.
The performance of an Open TBM mostly depends on the time required to install rock support. With an Open TBM it is possible to apply rock support measures right behind the cutterhead, in the so-called L1* work area. For example, ring erectors, anchor drilling devices or wire-mesh erectors can be provided for installing the steel supports. The supplementary support for the entire tunnel profile, such as shotcrete and segment, is provided in the back-up area.
Figure 2- 4. Open TBM system descriptions (WBI, 2007)
18 A simplified Open TBM cycle operation is shown in Figure 2-5. It should be noted that in phase (b), the machine slides on an invert shield or sliding shoe (WBI, 2007).
Figure 2- 5. Open TBM operation cycle (WBI, 2007)
Single Shield TBM
The Single Shield TBM (SS TBM) is primarily for use in a low stand-up-time rock mass.
To support the machine and the crew, this machine is protected by a shield (1) which is usually a cylindrical tapered steel structure over the entire machine. The tunnel lining is installed under the protection of the shield tail. In contrast to the Open TBM, the SS TBM and the hydraulic thrust cylinders (2) push against the last segment ring (3) installed. The cutter-head system is the same as for the Open TBM. The excavation diameter is slightly greater than the outer-shield diameter
19 to allow for over-cut around the machine to avoid machine jamming, especially in highly deformed ground.
Figure 2- 6. Single shield TBM (Herrenknecht, 2010)
A simplified SS TBM cycle operation is shown in Figure 2-7. In phase (b), the jack of the corresponding segment is retracted and after segment installation, the jack is extended against the installed segment. After the completion of the last ring assembly, the new stroke begins.
20
Figure 2- 7. Single shield TBM operation cycle (WBI, 2007)
Double Shield TBM
The double Shield (DS) TBM is also known as the "Telescopic Shield TBM" and is a type of shielded TBM that combines the benefits of both the Open TBM and the SS TBM to allow a faster advance rate in complex geological situations. On the one hand it enables the machine to advance forward as fast as an Open TBM in a sound rock mass while the rock support is installed independently under a separate shield named, a tail shield. On the other hand, it is possible to use the concept of the SS TBM in a very unfavorable and weak rock mass where grippers cannot be anchored properly against the tunnel sidewalls. Therefore, it has more flexibility than the Open and the SS TBM in variable geological situations, especially in long
21 tunnels where there are not sufficient detailed geological information before starting tunnel excavation. There are some disadvantages for this type of TBM in some special geological situations such as squeezing (where due to the long length of the shields as compared to the SS
TBM, TBM jamming is more likely to happen) and telescopic joint blocking in a highly fractured rock mass. Also these machines are more expensive. A simplified DS TBM and its cycle operation is shown in Figs 2-8 and 2-9.
Figure 2- 8. Double shield TBM (Herrenknecht, 2010)
22
Figure 2- 9. Double shield TBM operation cycle (WBI, 2007)
TBM Performance Parameters
TBM efficiency and performance parameters are studied by many researches for the purpose of understanding accurate prediction and ultimately, increasing TBM efficiency. The most common TBM performance parameters are Penetration Rate (PR), Advance Rate (AR), and
Utilization (U) factor which will be reviewed in the following sections.
23 Penetration Rate
Penetration rate (PR) or rate of penetration (ROP) is typically expressed in terms of m/hr
(or ft/hr) and is a function of machine specifications , geological conditions, and operational parameters. Many attempts have been made to seek a correlation between penetration rate and rock/TBM characteristics.
Utilization Factor and Advance Rate
The percentage of total time during which mining or rock excavation occurs is the utilization, U. It is usually expressed as an average over a specified time period. TBM utilization is controlled by several factors including rock type and rock-mass conditions, TBM operational parameters, slope and turns along the alignment, operational restrictions, site-specific conditions, and site management issues and contractor experience. Failure to accurately predict the TBM utilization factor can cause problems for development of a realistic plan for a tunneling operation to meet the project goals and better selection of machinery and methods as well as precautionary measures for efficient operation that could achieve high advance rates (TBM progress over a specific period of time such as day). TBM utilization (U) has a direct impact on TBM advance rates and project costs. The relationship between ROP, AR and U can be simply stated as: U =
Boring Time/Total time = (Total time – Downtime)/Total time.
AR = ROP x U x T (2-2)
This simply shows that any increase in utilization can directly impact the advance rate. It also means that even in operations where high ROP can be achieved, the tunneling rate could still fall short of expectations if the machine utilization is low. Alternatively, in very hard rock where
24 penetration is difficult, a higher utilization can allow for relatively good advance rates. In practice, TBM utilization can vary from single digits to 50%. This indicates that in most operations, the portion of the time spent in excavation is less than the time spent in various forms of delays or so-called down times. As such, it is necessary to thoroughly study TBM downtime and to be able to quantify it for the purpose of estimating TBM utilization. Additional discussion of the relevant literature for these parameters will be offered in the related chapters.
Planning a TBM tunnel project regarding required time and cost to complete tunnel crucially depends on TBM performance parameters, mostly utilization and advance rate. While these two parameters are the most important performance parameters, many studies were carried out to develop relationships for penetrations rate. One reason for this is that the penetration rate is mostly influenced by rock and TBM characteristics and can be estimated more easily with a relatively high degree of accuracy. The utilization and the advance rate were not studied in detail especially in recent years. In order to study these two parameters in more detail, it was deemed necessary to compile data from many tunnel projects around the world to include the variation of different parameters and offer more reliable formulas for estimation of AR and U.
25 Chapter 3
TBM Performance Databases
Introduction
Two separate databases were compiled from the review of various technical sources. The first database (general database) was assembled with the objective of developing a new performance model with data from more than 300 projects, records from around the world. The information in this database reflects average values for each parameter over the total length of the tunnels. The general nature of data cannot reflect the detailed information about each geological zone (if it existed) along the tunnels. Furthermore, this database contains data for the past 30 years in order to have as much information as possible. This might overshadow the effects of technological advancements in the past 3 decades (which are hard to account for) in the final analyses and results. In order to remediate these two problems, the second database (named as the detailed database) was developed. This database (which includes 17 recent tunnel projects) was used to represent the new projects while offering a sufficient level of details on variability of performance parameters along the tunnel. The new models are generated on the basis of the analysis of data from both databases in retroactive analysis of specific items versus overall tunneling performance, as deemed relevant. Finally, the results of multiple regression analyses were fine-tuned on the basis of many bi-variate analyses using the general and detailed databases.
Fig. 3-1 shows a general sketch of the databases developed for this study. One important issue to be noticed in this sketch is the missing data for different parameters in different records. This shows the heterogeneous nature of the data which causes a tremendous reduction in the number of records available for regression analyses. The last columns of this database represent downtimes for different activities for a number of tunnels (around 90 tunnels) which will be
26 treated separately for downtime analysis in the subsequent chapters. A list of projects included in each database is provided in Appendix A.
Xi Yi
… …
xi
x2 x3
Tunnel Record No. x1
Ua (%) Ua (%) Ub Uc (%)
Uw (%) Uw
PR (m/hr)
BU (hr/km) BU
Ara (m/day)
ARb (m/day) ARc (m/day)
ARw (m/day)
Other (hr/km) Boring (hr/km) 1 2 General 3 dataset . . . . . j ...... Detail . dataset Totalm tunnel projects with available data <200 <200 <200 <200 <200 <200 <200 <200 <200 <200 <200 <200 <200 <200 ≈90 ≈90 ≈90 ≈90 Total tunnel projects ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260 ≈260
Available data
Figure 3- 1. A general sketch of the developed database (Xi: input parameters, Yi: Target parameters, refer to Fig. 1-2)
General TBM Field Performance Database
The general database on TBM field performance contains different levels of information which defines the tunnel, rock mass conditions, and TBM performance parameters over the full length of a tunnel drive, and some within discrete geological zones or short tunnel reaches. The general database contains data on more than 260 tunnel projects and includes over 300 data sets.
This database contains TBM diameters ranging from 1.63 to 11.74 m. TBM projects compiled in
27 the database were completed between 1966 and 2004. An effort was also made to complete missing data fields within the database by checking other sources and published literature. This new database includes bored tunnel records with a total length of over 1500 km. Table 3-1 lists the main parameters included in this database.
Table 3- 1. Summary of the parameters included in the developed database
Layout Parameters Rock Mass Parameters Equipment Parameters Performance Parameters Project Name Rock type TBM type Project time Location Geological variability General spec. of TBM Support Type Application type Quartz content General spec. of BU Penetration Rate Tunnel diameter UCS TBM condition Advance Rate Zone length RQD/CFF Type of mucking system Utilization factor Slope Water condition Thrust Cutter wear rate Construction access Extreme mining areas Power Activity time distribution Tunneling hole Torque through year RPM Cutter size Cutter Diameter
The original database of Nelson et al. (1994), included data on 640 TBM projects. Data from the UTA database was compiled from diverse sources, including a literature survey, manufacturer records, and detailed project records. Parameters for the database were recorded either as directly reported in documents or as estimated based on references (Laughton, 1998).
The original database contained four levels of information. The first three levels contained progressively more detailed information for a given tunnel project over shorter spatial increments.
Each zone was categorized based on a general geological structure and similar rock material characteristics. This increased level of detail continues down to a mining cell, which is defined as a 10-m length within the tunnel. The fourth database level provides information required to model the TBM mechanical availability and the performance of key mining cycle activities; penetration, mucking and support installation. A schematic representation of the first three database levels is shown in Table 3-2.
28 Table 3- 2. Database levels and schematics of distance increments in each level (modified from Nelson et al., 1999) Database level Distance increments LevelDatabaseDatabase 1 level level DistanceDistanceThe increments whole increments tunnel drive GeologyLevelLevel 1 description 1 for the wholeThe wholeThe tunnelwhole drive tunnel drive tGeologyunnelGeology description description for for the the whole tunnelwhole tunnel
Tunnel Drive Tunnel Drive Tunnel Drive Level 2 Tunnel zone GeologicalLevelLevel 2 2 variation characterizedTunnel Tunnel zone zone byGeological zoneGeological parameters variation variation characterized by zonecharacterized parameters by zone parameters
Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 LevelLevel 3 3 Unit cellUnit within cell awit tunnelhin a zone tunnel zone GeologicalLevelGeological 3 variation variation characterized Unit cell within a tunnel zone byGeological zonecharacterized and variation unit by cell zone characterized parameters and by zoneunit celland parameters unit cell parameters
Unite Cell within Tunnel Zone Unite Cell within Tunnel Zone
Unite Cell within Tunnel Zone
Each data level can be linked to the other levels by means of a Drive Reference Number
(DRN) that is unique for each tunneling project. For this study, only level two of the database was
used. The data in this level contains average parameters for the zones of the tunnel or the different
geological units along the tunnel. It contains 209 records and includes data on tunnel geometry,
TBM characteristics, and geological parameters reported as numerical data, ranks, and categories.
This database accounts for a total length of over 800 km bored length.
This database was one of the first databases compiled for the purpose of predicting TBM
advance rate, the others being the TBM field performance database compiled by Norwegian
29 University of Science and Technology (NTNU) and a partial database by Colorado School of
Mines (CSM). The database parameters defined rock type and related rock mass parameters based on the recommendations of the International Society of Rock Mechanics (ISRM), including the
Basic Geotechnical Description (BGD), and common rock mass classification systems (Laughton,
1998). It should be noted that in categorizing the rock types in this database, only the dominant rock type in each zone was noted.
Detailed Database
To improve the predictive capability of existing models, a testing database including 17 recent hard-rock TBM projects was developed. These projects provided detailed information for
TBM performance in each geological zone. The TBM diameter for these projects ranged from 2.6 m to 11.8 m. As would be expected, the format and the extent of reported TBM operational parameters and geological data varied significantly from project to project. Table 3-3 shows the summary information reported for these projects, notably including a description of the TBM type, diameter and site geology.
30 Table 3- 3. General information of the tunnel projects in testing database
Tunnel TBM Reference No. Project Name Length TBM Type Diame Geology (km)* ter (m) 1 Ghomroud (Iran) 22.8 DS 4.5 Metamorphic rocks ** Hassanpour, 2 Karaj (Iran) 15.7 DS 4.65 Pyroclastic rocks 2009 Hassanpour, 3 Zagros (Iran) 5.3 DS 6.73 Limestone, Marl, Shale 2009 4 Golab (Iran) 0.7 DS 4.5 Diorite, Schist ** Meta gabbro, Meta Sapigni et 5 Maen (Italy) 1.7 Open 4.2 basite, Schist al., 2002 Sapigni et 6 Pieve (Italy) 6.3 DS 4.05 Schist, Granite, Diorite al., 2002 Sapigni et 7 Varzo (Italy) 6.2 DS 4.05 Gneiss al., 2002 Ramezanza 8 Queens (USA) 0.2 Open 7.06 Granite deh, 9 Milyang (S. Korea) 0.6 Open 2.6 Granite, Andesite Kim, 2010 Manapouri Kim, 2010 10 1.7 Open 10.05 Gneiss (Newzealand) 11 S. Manhattan (USA) 5.6 Open 3.84 Gneiss, Schist ** KCRC D 320-First Mixed Ramezanza 12 1.3 8.75 Granite Tube (Hong Kong) Shield deh, 2005 KCRC D 320-Second Mixed Ramezanza 13 1.4 8.75 Granite Tube (Hong Kong) Shield deh, 2005 Barla and Frasnadello-Pilot Dolomite, Limestone 14 1.6 Open 3.9 Pelizza, (Italy) and Argillite 2000 Barla and 15 Antea-Pilot (Italy) 0.7 Open 3.9 Dolomite Pelizza, 2000 Barla and Frasnadello-Main Dolomite, Limestone 16 1.6 SS 11.8 Pelizza, (Italy) and Argillite 2000 Barla and 17 Antea-Main (Italy) 0.7 SS 11.8 Dolomite Pelizza, 2000 Tot 73.6 al DS: Double Shield, SS: Single Shield
* with available data
** data from authors (Farrokh et al., 2006, 2011; Farrokh and Rostami, 2007, 2008, 2009)
Due to the difficulty of dealing with volumes of detailed data in several separate databases for different projects, it was necessary to reduce the number of data sets to a
31 manageable number. Heterogeneity of the data was also an issue which was caused by using different protocols for recording TBM performance data for different tunnel job sites.
To mitigate this problem for future works and obtain uniform data from the ongoing projects, the first step was to establish a proper template for gathering TBM utilization parameters from field recordings (see Fig. 3-2 and 3-3 as examples). This was accomplished during the construction of the Ghomroud water conveyance tunnel which is a 21-km long tunnel recently completed using a double-shield TBM. This procedure was also adopted, with minor modifications, by designers of two other projects including the Zagros and Karaj tunnel projects with a 6.7-m and a 4.4-m diameter TBM, respectively.
32
Figure 3- 2. A typical sheet of recorded data in the Ghomroud tunnel project for geological data as well as TBM operational parameters
33
Net Boring Time Regripping and Rock Support Train Exchanging and Muck Transportation Problems TBM Maintenace and Other Services Air, Water, Electricity Problems TBM Problems Back-up Problems Cutter Head Inspection and Ground Problems Portal and Logestic Problems Laser and Surveying Problems Stoppage and Shift exchange Other Downtimes
1% 1% 0% 2% 3% 4% 20% 2% 5%
5% 19%
38%
Figure 3- 3. A typical monthly time distribution categorization chart of the Ghomroud tunnel project
After collecting the data from various sources, the following steps were taken:
Breaking the tunnel data into certain reaches with similar operational characteristics.
This allows for reducing the size of workable data entries. For example the number
of TBM strokes for the Ghomroud tunnel project was more than 30,000 rows of data
that needed to be reduced to a series of sections and made more manageable. This
allows for averaging and normalization of the data to facilitate objective analysis
and development of predictive models.
Combining all data obtained from various tunnel projects in one database,
Screening the data and eliminating the data containing serious errors in recording, or
incomplete or major outliers,
Adding the geological parameter (rock type) by using codes, where non-quantitative
descriptions was available
34 Other researchers active in this area of work were contacted to examine the possibility of sharing databases on machine performance. While there are some possibilities for data sharing for future work, a list of data bases used in this study are provided in Table 3-3 (page 30).
Data Screening
In reviewing the TBM performance records, there were different approaches that could be used for presenting various parameters. The difference in these approaches are related to the definition of the total time for calculating the Advance Rate (AR) and the Utilization (U). These definitions are listed as follows.
Calendar days: Number of days between the tunnel project start and finish,
Working days: Number of days planned for working, which is generally the total
calendar days minus holidays,
Boring days: Number of days in which the TBM excavates and advances,
Available time periods: This refers to a fraction of the boring days in which the
TBM is available for boring. In other words the total boring days minus the TBM
maintenance and other downtimes related to TBM.
It should be noted that in some papers "Available time" refers to working days.
The TBM performance parameters of AR and U for each above-mentioned categories can be shown as follows.
Calendar days: ARc, Uc
Working days: ARw, Uw
Boring days: ARb, Ub
Available time periods: ARa, Ua
35 The total time used for calculation of the TBM parameters is not always clearly described in various records and publications. In order to control the reliability of the gathered information in the data base and assure consistency of the described time frames, a procedure was adopted to screen and reorganize the data. Fig. 3-4 represents the different steps used in this procedure as a flowchart. This screening procedure will create a consistent and reliable data set needed for the analysis of downtime, AR, and U. It should be noted that in this approach, the assumption was that the reported PR values are the average value for the whole length of the tunnel drive or a given geological zone. Also, the focus of this research was on parameters based on the working days (ARw and Uw).
36
Input: - PR - Calendar dates of project start and finish
ar dates of project start and Calculatingfinish ARc and Uc
- Major stoppages of the project - Holidays
Calculating ARw, Uw, ARb, Ub
Available values Y Output: <= Calendar days- related Consider Available values values as ARc and Uc
N
Available values <= Working Y Output: days- related Consider Available values values as ARw and Uw
N
Available values Y Output: <= Boring days- Consider Available values related values as ARb and Ub
N
Output: Consider Available values as ARa and Ua
Entering calculated parameters in the remaining places of ARw, Uw, ARb, and Ub
Figure 3- 4. Flowchart of Controlling the Reported Performance Parameters.
37
Review of TBM Performance Parameters in the Database
This section briefly describes the machine performance parameters in the database. These parameters are used for all of the statistical analyses performed by the Minitab Software package. A series of bi-variate analyses were conducted to evaluate expected formulas for each response variable (advance rate and the utilization). These analyses could provide detailed information about the role of each predictor on response variable and the best probable combination of parameters that could reflect the possible interaction of such parameters. The results of bi-variate analyses are explained in the following sections
TBM Operational Parameters
TBM operational parameters include Thrust, Torque, Power, and RPM, which are interrelated. The nominal or installed values for the TBM operational parameters are highly related to the tunnel diameter. TBM Penetration Rate (PR) which is the TBM progress in terms of m/hr during the boring process, is reported for more than 90% of the tunnel cases of the data base, while Advance Rate (AR) or Utilization (U) are not reported as often, or it is not clear if they are based on available time, boring days, working days, or calendar days. As shown in Figure 3-5, the correlation between PR and AR/U is significant, which reflects the true nature of the relationship between these two parameters (Eq. 1-1). The graphs also show that AR and U are almost independent of other parameters, or that the relationship is over shadowed by other factors.
38
60 60 S 9.52907 R-Sq 3.2% 50 50
R-Sq(adj) 2.6%
)
) y
40 y 40
a
a
d
d
/
/
m m
( 30
30 (
w
w R R 20
A 20 A S 7.76401 10 R-Sq 49.4% 10 R-Sq(adj) 49.2% 0 0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 14 16 PR (m/hr) RPM
60 60 S 9.39508 S 9.05393 50 R-Sq 0.0% 50 R-Sq 0.0%
R-Sq(adj) 0.0% R-Sq(adj) 0.0%
) ) y
y 40 40
a a
d d
/ / m
m 30 30
( (
w w R
R 20 20
A A
10 10
0 0 0 5000 10000 15000 20000 0 1000 2000 3000 4000 5000 Thrust (kN) Torque (kN-m)
60 60 S 11.3103 S 10.0862 50 R-Sq 0.0% R-Sq 17.4%
R-Sq(adj) 0.0% 50 R-Sq(adj) 16.7% )
y 40 )
a 40
d
/
%
(
m 30
(
w 30
U w
R 20 A 20 10 10 0 0 1000 2000 3000 4000 5000 6000 7000 0 2 4 6 8 10 12 C utter head pow er (kW) PR (m/hr)
60 60 S 10.2331 S 10.2994 R-Sq 2.4% R-Sq 0.3% 50 50
R-Sq(adj) 1.7% R-Sq(adj) 0.0% )
) 40 40
%
%
(
(
w
w 30 30
U U
20 20
10 10
0 2000 4000 6000 8000 10000 12000 14000 0 500 1000 1500 2000 2500 3000 3500 Nominal C utter head torque (kN-m) C utter head pow er (kW)
39
60
50
) 40
%
(
w 30 U
20 S 10.1016 R-Sq 5.4% 10 R-Sq(adj) 4.8%
0 10000 20000 30000 40000 50000 60000 C utter head Thrust (kN)
Figure 3- 5. Bi-variate analysis between TBM performance parameters and ARw/Uw
Tunnel Location and Application
This includes the country name, the city name, and the name of the project (Fig. 3-6).
140 120
100 t
n 80
u o
C 60 40 20 0 r l e y ic o y n t y ir d e r y ly e ne bl a r tr a ia ilo a o a ic e a p th n a w ct e w r P w v o v w w p O u C h e r st il r R r e b u t g l M to e a se e S u S s i e o d R e S S s H ro e r e d M P R te cc y a A H W Tunnel Application
100
80 t
n 60
u o C 40
20
0 r r e ka lia ria a a ia o e se y e g ia ly n a o ay ru n A n d y K S h s ra t ad in b d nc is an ec n d ta pa re ac e io S e an ke U U t la t s n h m a a u e o In I a o n rw P n d rl r O A s u a C lu u r S rm r K J K o o u e e u u A C o c F - e G g N e w z T A C E e G n M R S it c o w an H S Fr Country
Figure 3- 6. Histograms for tunnels locations
40 Tunnel Diameter
Figure 3-7 shows a histogram plot of the diameter of TBMs in the data base. As can be seen, a range of 3-6 m dominates the available cases in the database which reflects the common application range of TBMs. In general, as tunnel diameter increases, the AR and PR decreases due to the reduced RPM for larger TBMs. Figure 3-8 also shows a possible negative relationship between tunnel diameter and advance rate. Figure 3-8 shows a possible positive relationship between tunnel diameter and utilization factor.
This may be explained by an increased time per stroke for a larger TBM. Also larger tunnels mean more space for handling of the muck train and decreased related delays.
Fig. 3-9 to 3-11 indicate that the tunnel diameter has a strong interrelation with TBM characteristics and its performance.
80 70
60 y
c 50
n
e u
q 40
e r F 30 20 10 0 1.5 3.0 4.5 6.0 7.5 9.0 10.5 Excavated Diameter (m)
Figure 3- 7. Histogram of tunnel diameter in the data base
41
70 R² = 1.35% 60
50
40
30 Uw (m/hr)Uw
20
10
0 0 5 10 15 Diameter (m)
Figure 3- 8. Tunnel diameter versus utilization factor/advance rate
9000 R² = 71.13%
8000
7000
6000
m) 5000 -
4000 Tq (kN Tq
3000
2000
1000
0 0 5 10 15 Diameter (m)
Figure 3- 9. Relationship between torque and tunnel diameter
42
18
16
14
12 R² = 49.14%
10
RPM 8
6
4
2
0 0 5 10 15 Diameter (m)
Figure 3- 10. Relationship between RPM and tunnel diameter
70000 R² = 58.54%
60000
50000
40000
Th (kN) Th 30000
20000
10000
0 0 5 10 15 Diameter (m)
Figure 3- 11. Relationship between thrust and tunnel diameter
43 Length of the Tunnel
Figure 3-12 shows the histogram of tunnel length for different projects in the database.
As can be seen, the common range of length is below 10 km. That is mostly because of the transportation problems of long tunnels when there is no intermediate access. Tunnel length can have a crucial effect on TBM utilization. As tunnel length increases, the experience of the crew increases and without considering other parameters, the advance rate might increase (as it can be seen in Figure 3-13). Meanwhile, as tunnel length increases the tunnel transportation delays increase and it can reduce the AR/U. This depends on the management of the tunnel project.
60
50
40
y
c
n e
u 30
q
e
r F 20
10
0 0 3000 6000 9000 12000 15000 18000 21000 Excavated Length (m)
Figure 3- 12. Histogram of tunnel length in the data base
44
90 70 R² = 5.43% R² = 0.05% 80 60
70 50 60 40
50
Uw ARw 40 30
30 20 20 10 10
0 0 0 5000 10000 15000 20000 25000 0 5000 10000 15000 20000 25000 Length (m) Length (m) Figure 3- 13. TBM utilization/advance rate versus tunnel length
Unconfined Compressive Strength
Unconfined compressive strength (UCS) is a commonly-used representative of rock strength in almost all of the TBM tunnel projects. Increasing in UCS causes a decrease in PR, and thus AR. The results show that AR/U are low when UCS is very low or very high (Fig. 3-14).
This is related to the impacts of rock strength on reduced PR in hard rocks and delays in boring for installation of rock support and maintenance of the cutterhead in weaker rock formations.
Also, in stronger, more abrasive rocks the cutter wear increases (Fig. 3-15) and it causes an increase in cutter change time and frequency, resulting in lower utilization.
45
60 60 S 10.1782 50 50 R-Sq 4.3%
R-Sq(adj) 3.2% )
40 y 40
y
a
c
d
n
/
e m u 30
30 (
q
e
r w R
F 20 20 A
10 10
0 0 0 45 90 135 180 225 0 100 200 300 400 UC S (MPa) UC S (MPa)
60 S 10.2403 R-Sq 0.4% 50 R-Sq(adj) 0.0%
) 40
%
(
w 30 U
20
10 0 50 100 150 200 250 UC S (MPa)
Figure 3- 14. ARw/Uw versus UCS
600
R² = 19.80% 500
400
300 #Cutter/km 200
100
0 0 50 100 150 200 250 300 UCS (MPa)
Figure 3- 15. Cutter wear rate versus UCS
46 Core Fracture Frequency (CFF)
CFF is the only rock mass parameter used in the UTA database. Basically, this factor is in close relationship with RQD and refers to the frequency of the fractures in the rock mass. An approximate relationship between CFF and RQD was found as shown in Table 3-4. Figure 3-16 shows that a lower RQD leads to an increase in AR/U values, but it should be mentioned that this occurs beyond a certain limit of jointing frequency, AR/U might decrease as the rock support requirements or water inflow increase.
Table 3- 4. CFF categorization
CFF Code Description Corresponding RQD Range Less than 8 fractures/m S or 1 Low frequency 90-100 8-12 fractures/m M or 2 Medium frequency 60-90 12-16 fractures/m H or 3 High frequency <60
60 60
50
50 )
y 40 )
a 40
d
/
%
(
m 30
(
w 30
w U
R 20 A 20 10 10 0 1 2 3 * 1 2 3 * RQD RQD
Figure 3- 16. RQD versus ARw/Uw
Groundwater Condition
When the rate of groundwater inflow increases, the delays due to handling of water increases and machine utilization decreases. A secondary impact of groundwater inflow is that when water inflow increases, the stability of the rock mass decreases and the need for support
47 installation increases. Also, collecting and transferring wet muck is more time consuming. The main problem for using groundwater condition as a variable in the development of a new model is that the inflow occur locally, but it is cumulative along the tunnel length; thus it is very difficult to assign an average condition to a tunnel length/tunnel zone. Also, this parameter is not reported for most of the tunnels. As a result, a series of codes listed in Table 3-5 was used for analysis of groundwater data (Fig. 3-17).
Table 3- 5. water condition categorization
Ground water condition Code Dry to low water inflow, no effect on tunnel excavation D or 4 Water inflow at tunnel face does not affect the tunnel excavation time M or 3 Water inflow at face affects the tunnel excavation time significantly W or 2 Water inflow at tunnel face stops the tunnel excavation (Extreme Mining Area) EMA or 1
60 60
50
50 )
y 40 )
a 40
d
/
%
(
m 30
(
w 30
w U
R 20 A 20 10 10 0 1 2 3 4 * 1 2 3 4 * Water Water
Figure 3- 17. Water condition versus Uw/ARw
Rock Type
While rock strength and abrasion can be measured by some laboratory index testing, each rock type has its own texture and grain size and can affect the rock mass and cutter wear rate differently. Therefore, defining rock type can be very useful in corresponding categorical
48 parameters for AR/U analysis. Table 3-6 shows the classification of rock types into seven categories, as proposed by Hoek and Brown (1980) for rock engineering purposes. The first four classes are for "Sedimentary Rocks." The fifth, sixth, and seventh classes are for "Metamorphic, igneous, and Volcanic Rocks," respectively, as proposed by Stevenson G.W. (1999). The results of the bi-variate analysis show that at this level of information, there is no significant correlation between rock type and AR/U (Fig. 3-18).
Table 3- 6. Rock type categorization in data base
Rock Type Code Claystone, mudstone, marn, slate, phyllite, argillite C Sandstone, siltstone, conglomerate, quartzite S Limestone, chalk, dolomite, marble L Karstic Limestone K Metamorphic rocks such as gneiss and schist M Coarse igneous such as granite and diorite G Fine volcanic such as basalt, tuff, and andesite V Gneiss GN
80 60 70 50
60 )
y 40 )
a 50
d
/
%
(
m 30
( 40
w w 30 U
R 20 A 20 10 10 0 0 C S L K M G V GN * C S L K M G V GM * Rock Ty pe Rock Ty pe
Figure 3- 18. Rock type versus ARw/Uw
Major Mining Problems
This item relates to long stoppages of the TBM during the mining process for Extreme
Mining Areas (EMAs). In EMAs, normal operation of the TBM is suspended due to adverse geological conditions such as high convergence, tunnel collapse, or a high rate of water inflow.
49 Unfortunately, the number of data points was insufficient to investigate any relationships between these incidents and machine utilization (Fig. 3-7).
Table 3- 7. Major mining problems categorization
Mining problems Code Water drowning W Broken areas which cause tunnel instability B Faults and shear zones F Karstic holes and water inflow with silty material K Not known or other cases N
Geological Variability
In highly variable geological conditions where there are faults, folding, and weathered rocks, the utilization can decrease due to the delays caused by changing the rock support pattern.
The results of the analysis show that at this level of information, there is insignificant correlation between geological variability and AR/U (Table 3-8 and Fig. 3-19).
Table 3- 8. Geological variability categorization
Geological Variability Code Description Uniform* U No or little weathered or broken zones Variable V Highly Variable H Faults, folds, and weathered rocks * mostly can be seen in granitic or sub-horizontal sedimentary rocks
50
80 60 70 50
60 )
y 40 )
a 50
d
/
%
(
m 40 30
(
w w 30 U
R 20 A 20 10 10 0 0 U V H NA U V H NA Geologic V ariability Geologic V ariability
Figure 3- 19. Geologic variability versus ARw/Uw
Quartz Content
When the quartz content increases, the cutter wear increases and this can cause delays for cutter change and machine utilization decreases. Also, accelerated wear on the cutters could decrease the penetration rate. Based on Table 3-9 and Fig. 3-20, it seems that there is no significant correlation between quartz content and AR/U. This indicates the low ranking of this parameter in machine performance analysis.
Table 3- 9. Quartz content categorization
Quartz content in percentage Code Description 0-20 L Low 20-50 S Significant 50-75 H High >75 V Very High
80 60 70 50
60
) y
a 50 40
)
d
/
% m
40 (
(
30
w w
30 U
R 20 A 20 10 10 0 0 L S H V NA L S H V NA Q uartz C ontent Q uartz C ontent
Figure 3- 20. Quartz content versus ARw/Uw
51 Cutter Diameter
As the diameter of the cutter increases, the maximum thrust per cutter and the PR/AR can increase. Figure 3-21 shows that there is no significant correlation between cutter diameter and U.
60 60 S 9.28903 S 11.0005 50 R-Sq 6.5% R-Sq 0.7% R-Sq(adj) 5.3% 50 R-Sq(adj) 0.0%
) 40
y )
a 40
d
/
%
( m
30
(
w 30
w U
R 20 A 20 10 10 0 10 12 14 16 18 20 10 12 14 16 18 20 C utter Diameter (in) C utter Diameter (in)
Figure 3- 21. Cutter diameter versus Uw/ARw
Ground Support
Two parameters can be considered for measuring the impact of ground conditions and required ground support installation on TBM performance (Table 3-10 and 3-11). In shielded
TBMs, the support is mostly segmental lining for the whole length of the tunnel. For open TBMs, based on the ground conditions, the support can be a combination of rock bolt, wire mesh, shotcrete, and steel ribs/ laggings. The frequency of the installed supports depend on the rock mass quality and stability.
52
Table 3- 10. Basic support categorization
Basic Support Code Description Sketch The bolt spacing is more than the Occasional rock bolt TBM excavation course O and shotcrete
Systematic pattern of rock bolt (on a stroke spacing) Pattern of rock bolts P
Systematic pattern of rock bolt Canopy (Rock bolt, Channel, Wire mesh, C Strap)
Systematic pattern of steel ring
Steel ring R
Systematic installation of segmental lining Segment S
Table 3- 11. Support range changing categorization
Support range Code No change of basic support 0 One level change in basic support (e.g. O to P) 1 Two level change in basic support (e.g. O to C) 2 Three level change in basic support (e.g. O to R) 3
The rock support installation operation is mostly performed simultaneously with other
TBM activities and in weak ground conditions, there might be tremendous delays due to the need for installation of additional rock support measures. These delays are highly dependent on the variability of the rock masses along the tunnel and it is very difficult to assign a category of rock support to the whole length of the tunnel/zone. It seems that this is the main reason that a significant correlation between rock support categories and AR/U could not be found (see Fig. 3-
53 22 and Fig. 3-23). Obviously, another factor in lack of correlation between AR and ground support is that in weaker grounds where more support is needed, higher penetration rates (PR) are often achieved, which offsets the lower utilization.
60 80 70 50
60 )
40 y a
) 50
d
/
%
( m
30 40
(
w w U 30
20 R A 20 10 10 0 0 S R C P O NA S R C P O NA Support Support
Figure 3- 22. Support category versus Uw/ARw
60 80 70 50
60 )
40 y a
) 50
d
/
%
( m
30 40
(
w w U 30
20 R A 20 10 10 0 0 0 1 2 3 * 0 1 2 3 * Support Range Support Range
Figure 3- 23. Support range versus Uw/ARw
Tunnel Transport and Muck Haulage System
In general, tunnels with a continuous conveyor belt have higher AR/U (Fig. 3-14). In reality, there is no tunnel cases that are exactly similar in all other issues and only differ in material handling to allow a one-to-one comparison. Thus, it is very hard to find a significant correlation between transport system categories and AR/U. As a result, a coding system (see
Table 3-12) was used to evaluate the impact of tunnel transportation in related analyses
54
Table 3- 12. Tunnel transport system categorization
Tunnel Transport System Code Truck T Train for tunnel inside and shaft S Train for tunnel inside and tunnel outlet A Full conveyor belt for tunnel inside and shaft V Full conveyor belt for tunnel inside and tunnel outlet H
Train Conveyor belt
80 60 70 50
60 )
y 40 )
a 50
d
/
%
(
m 40 30
(
w w 30 U
R 20 A 20 10 10 0 0 T S A V H NA T S A V H NA Muck Ev a Muck Ev a
Figure 3- 24. Tunnel transportation (or muck evacuation system) versus ARw/Uw
Tunnel Access
Tunnel access type includes portal and shaft and the transportation capacity of the operation is dictated by shaft depth. Meanwhile, no significant correlation was found between tunnel access categories outlined in Table 3-13 and AR/U.
Table 3- 13. Tunnel access categorization
Tunnel Access Code Tunnel outlet P Shaft with depth of less than 15 m S Shaft with depth of 15-50 m M Shaft with depth of more than 50 m D
55
60 60
50 50 )
y 40 )
a 40
d
/
%
(
m 30
(
w 30
w U
R 20 A 20 10 10 0 D M S P NA D M S P NA Shaft Depth Shaft Depth
Figure 3- 25. Shaft depth category (Shd) versus ARw/Uw
Year of Completion
As the technology of TBM manufacturing and machine capabilities improve, it is logical to anticipate higher TBM performance to be achieved in various tunneling projects. Figure 3-26 shows the histogram of tunneling projects in the database versus the completion year, and the bi- variate correlation suggests a modest correlation between machine improvements and increased
AR, while machine utilization seems to be insensitive to advances in machine manufacturing.
56
50
40 y
c 30
n
e
u
q
e r
F 20
10
0 1968 1974 1980 1986 1992 1998 2004 Hole Through Year
80 60 S 11.2818 S 10.7203 70 R-Sq 6.3% 50 R-Sq 5.0% R-Sq(adj) 5.3% R-Sq(adj) 4.0% 60
) 40
y a
50 )
d
/
% (
m 30
( 40
w
w U
R 30 20 A 20 10 10 0 0 1965 1970 1975 1980 1985 1990 1995 2000 2005 1965 1970 1975 1980 1985 1990 1995 2000 2005 Hole Through Year Hole Through Year
Figure 3- 26. Hole through tunneling year histogram and hole through tunneling versus ARw/Uw
TBM Condition
For a tunnel project, depending on the availability of the right TBM size, it is usually economical to use refurbished or used machines. There is no straight forward relationship between TBM condition-TBM performance since many other factors affect the end results. The quality of over-hauling of the TBM and the previous TBM modifications are among the most important factors that cannot be qualified very easily in any related analysis.
57 TBM Type
Different TBMs are selected for different ground conditions. While the cycle time of the shielded TBM is more than for the open TBM in a similar condition, the final outcome of the process shown by AR does not suggest a systematic advantage of Shielded machines over open
TBMs or otherwise (Fig. 3-27).
80 60 70 50
60 )
y 40 a
50 )
d
/
%
( m
40 30
(
w w 30 U
R 20 A 20 10 10 0 0 NA DS O pen SS NA DS O pen SS TBM Ty pe TBM Ty pe
Figure 3- 27. TBM type versus Uw/ARw
Tunnel Slope
In general, the steeper the tunnel the lower the AR/U could be anticipated, but the number of the high-slope tunnel cases in the database did not allow for a distinctive analysis of
TBM performance relative to this parameter.
Rock Mass Characteristics
The importance of rock mass characteristics in calculation of TBM performance is sufficiently obvious. While weaker rock masses allow for a higher penetration rate, the increase in ground support requirements typically offset any potential gains in PR. The Rock Mass Rating value (Bieniawski, 1989) is a commonly used parameter to represent rock mass properties. This
58 parameter was only available for a number of projects in the detailed database and was used in subsequent studies to reflect the impacts of rock mass conditions on TBM performance, as will be discussed later. But overall no significant correlation was found between RMR and machine
AR/U in the initial bi-variate analysis.
59
Chapter 4
Direct Advance Rate Prediction
Introduction
TBM advance rate (AR) is one of the key parameters required to calculate the time and cost to complete a tunnel. As a definition, AR is the average rate of TBM progress in a specific period of time which is usually expressed in the unit of m/day. This parameter is obtained directly as a product of Penetration Rate (PR which is sometimes referred to as rate of penetration or
ROP) and utilization (U) as shown in Eq. 4-1.
AR = PR · U · T (4-1)
where PR is the penetration rate in m/hr, U is utilization in %, and T is the working hours per day.
Robbins (1992) noted the geologically related conditions (mining, rock support, and mucking capacity) and tunnel diameter as the most important factors influencing AR. During the past few decades, many studies have been carried out to develop TBM performance prediction models, but the main focus of most of these studies has been on penetration rate (PR) prediction.
The work on the prediction of PR of hard-rock TBMs includes Tarkoy (1973) who examined various geotechnical measurements and offered total hardness for performance prediction. Meanwhile, a great number of efforts was placed on measurement and prediction of the cutting forces acting on the disc cutters to predict machine performance. Roxborough and
Phillips (1975), Graham (1976), Ozdemir et al. (1978), Farmer and Glossop (1980), Cassinelli
(1982), Snowdon et al. (1982), Sato et al. (1991), Sanio (1985), and Rostami et al. (1993, 1997,
60 2008) are among the researchers who performed many tests on various disc cutters and offered force estimation models. On the other hand, Listerud et al. (1983), Nelson et al. (1983), Bamford
(1984), Hughes (1986), Boyd (1986), Innaurato et al. (1991), Sundin and Wanstedt (1994),
Haworth et al. (1995), and Bruland (1998) are among the researchers who focused on machine field performance and offered empirical models for estimating the PR of hard-rock TBMs. The work by Cheema (1999), Yagiz (2002), Ribacchi and Lembo Fazio (2005), Ramezanzadeh
(2005), Gong (2006), Hassanpour et al. (2009a, 2009b, 2011), and Khademi et al. (2010) are among the efforts to reconcile the semi-theoretical and empirical models to incorporate the impact of joints and rock mass in the estimation of penetration rate for more accurate estimates.
The Alvarez (2000a, 2000b) approach is within the very limited approaches of using artificial intelligence measures such as a neural network for estimation of PR.
A limited number of studies have focused on the prediction of utilization (U) and AR.
One reason for the focus on PR prediction may be due to the additional difficulty in modeling the parameters that influence U and AR, especially regarding the analysis of different downtimes accumulated over the duration of a TBM drive. Some downtimes, such as cutter change, are highly correlated with rock properties, while others, such as major TBM system breakdowns cannot be evaluated without knowledge of many parameters such as TBM condition, management, and contractor experience. These parameters are difficult to assess in detail due to the lack of any recent in-depth analyses of TBM utilization and components of system delays.
Given the absence of a reliable predictive method for U, most researchers and practitioners continue to use approximate values for U based on reference to TBM field experience under similar conditions.
One important issue to address in evaluating AR from the multiplication of PR and U is limiting the significance of component errors on the AR result. To avoid this problem, Bieniawski et al. (2006) proposed a rock mass excavability (RME) system to be used for direct estimation of
61
AR. A similar approach was also introduced by Barton (1999, 2011) through the use of QTBM and has been in use by some engineers for the past decade.
This chapter reviews and compares the results of several mainstream AR estimating models for hard-rock TBMs through the evaluation of their predictive abilities. For this purpose the detailed database of performance parameters for 17 recent tunnel projects was used. This database includes information on various geological zones within the tunnel. This chapter also discusses the development of a new model that can be used for direct estimation of advance rate.
This new model is based on the analysis of the general database.
AR Prediction Methods
Table 4-1 categorizes the most important AR prediction methodologies developed so far along with their advantages and disadvantages.
62 Table 4- 1. Advance rate prediction models and their advantages and disadvantages
General Form Example Advantage Disadvantage Indirect CSM, NTH Accounting for time Might yield inaccurate results AR = PR*U Sharp and Ozdemir ( components analysis due to this combination 1991) US Army ( Accounting for geotechnical Limited range of application due 1997), ITA (2000), information /TBM to lack of updating Rostami et al. (1993 specifications and 1997), Bruland (1998)
Semi direct QTBM (Barton, 1999, Many input parameters AR = PR*U 2000, 2011) Relying on good database Complex relationships Requires uncommon tests Probabilistic Models Nelson et al. (1999, Accounting for randomness and Lack of formula and persistent 1994a, 1994b), approximation need for database Laughton (1998)- Accounting for many Most of the data in the original Abd Al-Jalil (1998), parameters through using like- database are too old Laughton et al. cases Cannot be adopted to new (1995) Creating probability density machines and technologies function (pdf) of AR Making "informed decision" regarding choosing similar cases (Abd Al-Jalil, 1998) Computer-Aided Alvarez (2000a and Relying on good database Complex underlying structure Models 2000b) Accounting for complex Over fitting relationships Usually not available in public domain No formulae to use, relies on proprietary software Direct Bieniawski et al. Development of a new rating Similarity to RMR system AR=f(RME) (2006) system for TBM applications. Lack of some basic parameters Incorporating recent tunneling in the original proposed experience with TBM formulae (e.g. tunnel diameter) Limited database
A review of the available literature indicates that many cases inthe database used for the analysis do not include recent tunnel projects or are limited to TBM’s operating in a relatively limited range of geologic settings. However, since the RME model is very recent and some of the available data from the project did contain relevant information, a more in-depth discussion of this model is offered in the following section.
63 RME Model
The Rock Mass Excavability (RME) indicator for predicting boreability of a given rock mass by a TBM is based on a classification proposed by Bieniawski et al. (2006). Thuro and
Plinninger (2003) were among other researchers who investigated RME in drilling and blasting and cutting by TBMs and road headers. The RME system includes five parameters specifically related to rock mass characteristics and the trends of the suggested ratings for most of the parameters are in agreement with Thuro and Plinninger (2003) study results. Initially RME was applied to the data from 22.9 km of four tunnels bored with four TBMs in Spain. A number of statistical correlations have been established between RME and the Average Rate of Advance
(ARA). In recently published works by Bieniawski et al. (2007a, 2008), three main correction factors for the prediction of advance rate were introduced by considering the influences of the
TBM crew, tunnel excavation length, and tunnel diameter (Table 4-2, Fig. 4-1 to 4-4, Eq. 4-2 to
4-4). This model is still under development. Some extensions were recently offered by
Bieniawski et al. (2008) for other types of TBMs as well as for two main categories of uniaxial compressive strength (UCS) (Fig. 4-2). The RME system was also utilized for cutter consumption prediction (Bieniawski et al., 2009). Khademi et al. (2009) offered a fuzzy logic model for application of the RME system. It should be noted that RME calculations are easier using classic ratings since RME authors have offered continuous rating charts (Fig. 4-1). The following equations are offered for calculation of AR using the RME model.
64 Table 4- 2. RME index parameters (Bieniawski et al., 2006, 2008)
Figure 4- 1. RME index parameters ratings (Bieniawski et al., 2006, 2008)
65
UCS<45 MPa Open TBM UCS>45 MPa 60 60 y = 0.324x - 6.866 y = 0.839x - 40.831 50 R² = 0.524 50 R² = 0.580
40 40
30 30
ARt (m/day)ARt (m/day)ARt 20 20
10 10
0 0
0 0
20 40 60 80 60 80 10 30 50 70 90 10 20 30 40 50 70 90
100 100 RME SS TBM RME 60 60 y = 10.059ln(x) - 13.321 y=23 {1-242^[(45-x)/17]} 50 R² = 0.606 50
40 40
30 30
ARt (m/day)ARt (m/day)ARt 20 20
10 10
0 0
0 0
20 40 60 80 60 80 10 30 50 70 90 10 20 30 40 50 70 90
100 100 RME RME DS TBM 60 60 y = 0.66x - 20.4 y = 0.422x - 11.6 50 R² = 0.751 50 R² = 0.433
40 40
30 30
ARt (m/day)ARt (m/day)ARt 20 20
10 10
0 0
0 0
20 40 60 80 60 80 10 30 50 70 90 10 20 30 40 50 70 90
100 100 RME RME Figure 4- 2. Correlation between RME index and the average rate of advance for double-shield TBMs (redrawn from Bieniawski et al., 2008)
AR AR F F F r t E A D (4-2)
F 0.7 F F F E E1 E2 E3 (4-3)
FD 10/ D (4-4) where ARr is real advance rate, ARt is theoretical advance rate, FE is correction factor of crew efficiency (Table 4-3), FA is correction factor of team adaptation to the environment, FD is correction factor of tunnel diameter, and D is tunnel diameter in m (Bieniawski et al., 2008).
66 Table 4- 3 Criteria for evaluation of coefficients FE1, FE2 and FE3 (Bieniawski et al., 2008, Grandori, 2007)
1.4
1.2
1.0
0.8 FA 0.6
0.4
0.2
0.0
2 6 0 4 8
10 12 14 16 Tunnel Length (km)
Figure 4- 3. Adaptation correction factor (Bieniawski et al., 2008)
45 40 ARr (R²<30%) 35 Gilgel Gibe II Tunnel, 30 Ethiopia, Dia.: 7 m, 25 TBM: DS 20
ARr ARr (m/day) 15 10 5
0
0
70 80 90 10 20 30 40 50 60 100 RME
Figure 4- 4. RME and TBM AR relationship at GIBE outlet drive (data obtained from Grandori, 2007)
67
It should be noted that FA, FE, and FD (Bieniawski et al., 2008) were named FL, FC, and kD or CL, CE, and CD in other publications (Bieniawski et al., 2007a, 2007b, 2007c) with some differences in the formulae. Also, it seems the formulae that SELI applies (Bieniawski et al.
2007c, Grandori et al., 2011) to its projects are a little different from what Bieniawski et al.
(2008) presented (especially for tunnel diameter correction factor).
Evaluation of Previous Advance Rate Prediction Models
The purpose of evaluating existing AR prediction models is to examine their capability for TBM advance rate prediction (Table 4-2). This evaluation was over a range of tunnel or rock mass conditions. Also, based on the availability of the required information in the TBM field performance database for each individual model, the number of data points might be different for different models. The graphs of Fig. 4-5 show the comparison between actual and predicted AR for the selected models.
68
120 120
Qtbm model CSM model
90 90
60 60 Predicted ARw (m/day) ARw Predicted 30 (m/day) ARw Predicted 30
0 0 0 30 60 90 120 0 30 60 90 120 Actual ARw (m/day) Actual ARw (m/day) 80 75 RME model NTH model
60 50
40 25
0
20 (m/day) ARrRME Predicted ARw (m/day) ARw Predicted
0 -25 0 20 40 60 80 -25 0 25 50 75 Actual ARw (m/day) Actual ARw (m/day)
Figure 4- 5. Results of comparison for different models
Overall, Figure 4-5 shows that the results of AR estimation using available models tend to overestimate the advance rate. This is in agreement with study results of Goel (2008) for evaluation of Qtbm and RME models for a Himalayan tunnel. The graphs show that the percent differences between predicted and observed values can sometimes be in excess of 100%. The overestimation of utilization (e.g. more than 60% of cases) was noted as a cause for some of the extreme AR overestimates in applying the QTBM model. AR overestimates in the CSM model were mostly attributed to an overestimation of penetration rate (e.g. close to pump limit). The
NTH model tends to have relatively low variations of advance rate. One reason for this fact is
69 related to using inverse correlations for some components of utilization (TBM and Back-Up) and penetration rate in this model (Fig. 4-6). This means that as penetration rate increases, for the same jobsite conditions, utilization decreases. As gains in PR are roughly offset by losses in U, the product of the penetration rate and utilization stays approximately unchanged.
RME model seems to give the best results among the other models tested since the amount of overestimation is lower and the general trend of the points follows the 1:1 line. The
ARt graph of RME model shows better results in the sense of having both overestimation and underestimation, but the points are mostly stacked around one point (here 20 m/day, Fig. 4-7).
ARr results indicate that by applying correction factors, the points have a better spread but still the results are mostly higher than the real values. Another point for RME model predictions is that there are some negative predicted values which are mostly related to very hard rocks (with
UCS>200 MPa). One reason for this phenomenon might be the use of very low rating values for
UCS and the drilling rate index “DRI” (close to zero) and the inter-correlation between the two parameters. Another reason might be related to having relatively high negative intercept values in the RME formulae (Fig. 4-2, e.g. - 40.8 for Open TBM case).
225 NTH model 200
175
150 Ta
125
100 Ttbm
Down(hr/km) Time 75 Tbu 50
25
0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5
PR (m/hr)
Figure 4- 6. Inverse correlation between downtime components and penetration rate in NTH model (Bruland, 1988)
70
75 75 RME model RME model
50 50
25 25
RME ARt(m/day) RME RME ARr (m/day)ARr RME
0 0
-25 -25 -25 0 25 50 75 -25 0 25 50 75 Actual ARw (m/day) Actual ARw (m/day)
Figure 4- 7. ARr and ARt for RME model using testing database
The main advantage of the RME model is that it is based on recent tunneling data and it includes some important factors impacting TBM performance, but the database used for development of this model is limited to a few tunnel cases. Also, the accuracy of the model is rather low (e.g. R2<45% for the case of the double shield TBM), and even applying correction factors have not improved the results of some cases (see an example in Fig. 4-4 as was reported by Grandori (2007)). Some of the discussion points about application of this model can be summarized as follows:
The proposed formulae in the RME model clearly show that there is a positive
relationship between RME and AR.
Although DRI is accepted worldwide, only a few laboratories can run this test and
the test is not as common as UCS. One issue about the RME system is that, although
this system was specifically developed for TBM application as compared to the
original RMR classification, it has the same structure as the RMR in the sense that it
comprises several ratings that are simply added together without any indication of
statistical significance being noted by the authors. Although the general trends of the
71 RME ratings are in agreement with field performance, the ratings are approximate
and could vary somewhat based on the judgment of the individuals applying the
information to obtain the ratings (as is the case with any rock mass classification
system). Obviously the ratings could be re-adjusted and refined in future revisions
of the RME model. Notably the RME developers currently use different formulae
for two ranges of UCS (UCS<45 MPa and UCS>45 MPa).
The FA factor was obtained from information of one tunnel (Guadarrama) as
Bieniawski et al. (2006) mentioned.
For some cases of very hard rock types, the RME model can yield negative values
for AR.
Considering the number and variation of parameters involved in ARr prediction, the
spread of data points observed on some RME model graphs is considered to be
relatively low (Fig. 4-2). In particular, the graph for a single shield TBM operating
in the harder rock class (UCS>45 MPa) does not show a significant spread in results
with the predicted values being almost constant.
Although these models were calibrated to field performance data, their application in a predictive mode will typically be limited to the prediction of TBM performance in a small set of rock types or job-specific conditions. Realistic modeling results would only be expected if the new set of tunnel conditions is similar to the conditions from which the models were developed.
The key point to note relative to the use of all these models is that they will predict machine performance in ideal conditions and that a maximum achievable rate is calculated, leading to an overestimate of the performance achieved in the field. In other words, these models assume that the machine is going to be used at capacity, which in reality does not happen all the time. In addition, the models assume a site layout and management conducive to the maintenance of high utilization rates. These assumed conditions could obviously be compromised by any
72 number of logistical issues, coordination and labor issues, and factors related to contractor/crew experience. In summary, most existing AR models offer the most likely upper bound of TBM performance and logically no estimator can anticipate how badly the ground conditions could be or how bad the machine is operated or the jobsite is managed. Thus any of these variations is likely to reduce machine performance and AR and the estimator cannot predict such issues and the results are the typical overestimations of the predicted rate, as it was shown in this study as well.
Fortunately, since RME has had a higher rate of acceptance worldwide, it is only natural that the future development and improvements will focus to rectify the noted issues with the
RME model.
Proposed New Model
The purpose of this section is to introduce a new model for direct prediction of ARw based on the statistical analysis of a TBM field performance database. The compiled database includes both numerical values (measurements) and data categories and codes described in chapter 3. Dealing with rated or categorized data in statistical analysis can be problematic. One approach is to assign different numbers to different categories to convert them into quantitative values (ranking) as was used by Alvarez et al. (2000a). Another approach is to run the analyses for each category separately. In this study, both approaches were used to develop a new model for estimating the advance rate under a given set of ground conditions. The first approach involves the use of the whole database for the analysis by converting categorical parameters to numerical values. The second approach entails dividing the database into sub-sets based on rock mass condition. Obviously, there are many missing data points (parameters) for the different tunnel records which make the database heterogeneous. This means that only a limited number of
73 records can be used when considering a certain combination of input parameters, thus reducing the population size used in statistical analysis. For this reason, it was difficult to directly utilize two main methods of regression analysis (stepwise and best sub-set) for the whole database. The alternative approach to deal with this problem was to use a combination of parameters or composite parameters. The results of this first set of bivariate analyses indicate low correlation between AR and each individual parameter except for PR which is inherently incorporated in AR
(Fig. 4-8). Many attempts failed to yield significant results for AR with R2 of less than 30% even by considering more than 15 parameters (see chapter 4 for parameters). This was somehow expected as other researchers such as Abd Al-Jalil (1998) and Morgan et al. (1979) noted poor correlations between common parameters and AR.
60
50 )
y 40
a
d
/ m
( 30
w R
A 20 S 7.76401 10 R-Sq 49.4% R-Sq(adj) 49.2% 0 0 2 4 6 8 10 12 PR (m/hr)
Figure 4- 8. Correlation between PR and ARw
The second approach was applied to explore the relationship between AR and the other sub-category parameters in more detail. For this purpose, several analyses were performed by categorizing/ranking the database on the basis of different parameters such as CFF (core fracture frequency) to derive individual formulae for each category. The validation of the results, relative to the application of the derived formulae on the testing database, was used to select the best sub-
74 sets and formulae. The final analysis used database sub-sets for three tunnel-diameter ranges.
Multivariate regression analysis was performed using Minitab Ver.16 with ARw as the objective variable. Based on these analyses, best-fit linear regressions were identified and adopted for different sets of independent variables.
Fig. 4-9 illustrates the correlations between AR and some of the most important variables in the database. Coefficients of determination (R2) of different bivariate analyses show that ARw-
Rock Type has the strongest correlation (R-sq of around 17%), yet it is too weak to be used as a sole predictor.
60 60 S 10.1782 S 9.94491 50 R-Sq 4.3% 50 R-Sq 3.2%
R-Sq(adj) 3.2% R-Sq(adj) 2.4%
) )
y 40
y 40
a
a
d
d
/
/ m 30 m
( 30
(
w
w R
20 R 20
A A
10 10
0 0 0 100 200 300 400 0 5000 10000 15000 20000 25000 UC S (MPa) Length (m)
60 80 S 10.1469 50 R-Sq 2.2% 70
R-Sq(adj) 1.3% 60
)
) y
y 40 a
a 50
d
d
/
/
m m
30 ( 40
(
w
w 30 R
R 20 A A 20 10 10
0 0 2 4 6 8 10 12 C S L K M G V GN * Diameter (m) Rock Ty pe Figure 4- 9. Correlation between ARw and other parameters (rock type codes from chapter 3)
As can be seen in Fig. 4-9, overall, for higher UCS and tunnel diameter, AR decreases.
Also, as tunnel length increases the ARw increases. These are in agreement with the results of the
75 RME model (Bieniawski et al, 2006). Each rock type has its own texture, grain size, and behavior. These rock type properties can affect the advance rate differently, even when the UCS of two different rock types are in the same range. Therefore, it is appropriate that these properties are used as categorical parameters in the analysis. Table 4-4 offers seven rock type categories for rock engineering purposes (Laughton, 1998), modified from Hoek and Brown (1980). It should be mentioned, Gneiss (GN) is inherently metamorphic but it is typically closer to granitic rocks in terms of its behavior, especially where foliation is less pronounced. For this reason, it was categorized as GN in this analysis. Also, metamorphic rocks were divided into two categories of
Meta-sediments (MS) and Meta-igneous (MI) as used by Robbins (1992).
Table 4- 4. Rock type categorization in database (Laughton, 1998)
Rock Type Code Claystone, mudstone, marl, slate, phyllite, argillite C Sandstone, siltstone, conglomerate, quartzite S Limestone, chalk, dolomite, marble L Karstic Limestone K Metamorphic rocks such as gneiss and schist M Coarse igneous such as granite and diorite G Fine volcanic such as basalt, tuff, and andesite V
As can be seen in Fig. 4-9, rock type has a very good relationship with ARw. The graphs show that in general a higher AR is achieved in sedimentary rocks and a lower AR is achieved in igneous rocks. The results are in agreement with the results proposed by Laughton (1998) and
Robbins (1992) for penetration rate (Fig. 4-10).
76
45
40 Sedimentary Cutter diameter: 432 mm Cutter Load: 200 kN 35 Grain Size: 3 mm Joint Spacing: 200-400 mm 30 Metasediment 25
20 Igneous
PRev (mm/rev) PRev 15
10 Meta Igneous 5
0 0 50 100 150 200 250 UCS (MPa)
Figure 4- 10. Correlation between PR and UCS for different rock types (Redrawn from Robbins, 1992)
Linear Regression Analysis
Different multivariate regression analyses were run with ARw as the objective variable; these analyses were performed using Minitab 16. The results of these analyses showed three parameters of uniaxial compressive strength (UCS), tunnel diameter (D), and TBM type (Open, single shield, double shield) as the most influential parameters for AR prediction. Fig. 4-11 shows one typical output of compared results between predicted and actual advance rate values using linear regression analysis. One interesting result noted in Figure 4-11 is that by categorizing the different tunnel diameter ranges, the effect of the tunnel diameter on performance was more readily observed, in contrast to the analysis of the whole database in which the effects of other parameters (such as rock mass conditions) overshadowed the diameter effect. It should be noted that while bivariate analyses showed that rock type has strong correlation with advance rate, it did not appear in the final formulae probably due to lower significance of this parameter in
77 combination with other parameters. This could also be due to the impact of rock type being included in some other parameters, perhaps the measured UCS.
4.0
R2=65%
3.5
e
s
n
o
p
s
e R
3.0
d
e
t
a
l
u
c
l a
C 2.5
2.0
2.0 2.5 3.0 3.5 4.0 Actual Response
3.8
3.6 R2=67%
3.4
e s
n 3.2
o
p s
e 3.0
R
d
e 2.8
t
a
l
u c
l 2.6
a C 2.4
2.2
2.0 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 Actual Response
3.5 R2=62%
e 3.0
s
n
o
p
s
e
R
d 2.5
e
t
a
l
u
c
l
a C 2.0
1.5 1.5 2.0 2.5 3.0 3.5 Actual Response
Figure 4- 11. Results of linear regression analyses for ARw (Log values) for 2.5
78 For the verification purpose, the obtained formulas from regression models were applied to the testing database and the results showed poor agreement for the testing database. In order to fine tune the results of multivariate regression analyses, the whole database was filtered for normal conditions (good rock quality) and the regression analysis was performed again for the new set of data. Then, a new rating system was established to realign the advance rate results in various rock mass rating values and for different TBM types. This rating system is obtained based on having a minimum error (minimum root mean square error (RMSE)) for predicted advance rate. Table 4-5, shows the Minitab output for the filtered database for normal condition.
Table 4- 5. Regression coefficient statistics for advance rate prediction for normal condition
Predictor Coef SE Coef T P VIF Constant 3.66814 0.09071 40.44 0.000 UCS -0.00589 0.000144 -41 0.000 1.022 D -0.08512 0.003197 -26.62 0.000 1.046 Dc 0.028475 0.005345 5.33 0.000 1.037 M 0.09881 0.02733 3.62 0.000 1.043
Equation (4-5) shows the regression function for the advance rate in the filtered database
(named as base advance rate).
ARw Exp(3.67 0.00589UCS 0.0851D 0.0285Dc 0.0988M) R2 93% (4-5)
where UCS is uniaxial compressive strength in MPa, D is tunnel diameter in m, Dc is disc cutter diameter in inch, M is a dummy variable for the mucking system (1 for conveyor belt and 0 for train) (Fig. 4-12).
79
4.0
e s
n 3.5
o
p
s
e
R
d 3.0
e
t
a
l
u
c l
a 2.5 C
2.0 2.0 2.5 3.0 3.5 4.0 A ctual Response
Figure 4- 12. Results of linear regression analysis for ARw (Log values) for the normal conditions
Fig. 4-13 shows the obtained correction factor for different RMR values and TBM types
by minimizing the RMSE.
60 1.2
Double Single Shield Shield TBM * Muckingsystem: Rail-bound TBM 50 * Disc cutter size: 17" 1 * Open TBM * Commonly practiced ARw
40 0.8
D (Tunne diameter) ≈ 4 m ARw (m/day) 30 0.6
D ≈ 6 m RMR RMR correction factor
20 0.4 Open TBM
D ≈ 8 m 10 0.2
D ≈ 10 m
0 0 0 50 100 150 200 250 0 20 40 60 80 100 UCS (MPa) RMR 1.2 Very good experience Figure(No learning 4 phase)- 13. Correction factor for base advance rate in different RMR values and TBM types 1
0.8 AR reduction due to increase in the rate of mucking system failure (e.g. derailments, ...)
0.6 Normal (Commonly practiced)
0.4
Low experience (relatively long learning phase) Tunnel lengthTunnel correction factor 0.2
0 0 2 4 6 8 10 12 14 16 18 20 Tunnel Length (km)
Notes:
1. For the following conditions add up to 20% or more (especially for the case of using tunnel conveyor) to calculated AR. If all consitions exist, the AR increase might be more than 40%.
- Tunnel conveyor belt - High experienced crew - Disc cutter size>17" (Technology Improvement)
2. For the case of low experienced crew, lower the results up to 20%
3. For the case of steep tunnels (around 15 degrees), lower the results up to 20%
4. Multiply the results by RMR (rock mass rating; Bieniawski, 1989) correction factor.
5. Multiply the results by tunnel length correction factor. 80 In order to obtain the advance rate, one should use Eq. 4-5 to obtain the base AR and then find the correction factor for the corresponding RMR and TBM type to adjust the advance rate.
Some noticeable points about Eq. 4-5 and Fig. 4-12 are listed as follows.
1. As the tunnel diameter increases, the amount of AR decreases. This relationship can be attributed to a decrease in penetration rate due to a reduction in cutterhead RPM and an increased demand on muck transportation.
2. As UCS increases, AR decreases. This can be explained by the lower values of penetration rate in stronger rock types.
3. In general the potential AR of open TBM is the highest and the AR of a single shield
TBM is the lowest, which is intuitive knowing the mode of operation of the open versus single shield machines and lower utilization of single shield TBMs in rock. This will be reversed in weaker rock masses due to the frequent interruptions due to ground support installation on the open type machines and thus lower utilization rates.
4. For a rock with a given UCS, it is anticipated that penetration rate would increase at lower RMR values due to the effect of joints; however, in practice operators tend to run the TBM at reduced thrust and RPM. Considering this fact and also the higher demand of support in lower
RMR values, the advance rate is generally lower for all TBM types.
5. The advance rate of TBMs in projects using a conveyor belt as the main tunnel haulage is normally higher than the advance rate of TBMs where trains are used for main tunnel haulage.
Fig. 4-14 shows a comparison between these two cases for six cases of similar TBM tunneling conditions (e.g. similar ground conditions and similar tunnel diameter). The results show that in good tunneling conditions (e.g. no weak ground), the difference between AR for these two cases can be well above 20%.
6. The advance rate increases as experience of the TBM crew increases and a correction factor should be considered for that.
81 Figure 4-15 shows the results of comparison between predicted and actual AR.
80
Comparison between ARw of tunnels using Conveyor belt versus rail in similar tunneling conditions 60 Mucking with Tunnel Conveyor Belt
40 ARw (m/day)
Mucking with Train 20
0 0 1 2 3 4 5 6 Case No.
Figure 4- 14. Comparison between six pairs of ARw for similar tunneling conditions (on the basis of ARw of 12 tunnel cases)
New Proposed Model
30
R² = 0.7918
20 Predicted(m/day) ARw
10
0 0 10 20 30 40 Actual ARw (m/day)
Figure 4- 15. The comparative results of new model between predicted and actual AR As can be seen in Fig. 4-15, the estimated ARw values based on the proposed model are in good agreement with actual ARw presented in the database. Fig. 4-15 shows the graphs
82 comparing the predicted ARw for both the new proposed model and RME model. In these graphs,
RMSE refers to the root mean square error (Alvarez et al., 2000a) (Eq. 4-6) and is used to compare the performance of the new model and RME formulas. Lower RMSE values refer to lower errors and better performance. As can be seen in Fig. 4-16, RMSE value of the new model is lower than the one for RME model.
1 n ˆ 2 RMSE (yi yi ) n i1 (4-6)
where yi is the real value, yˆ i is the predicted value, and n is the number of data points.
New Model RME ARr RMSE=2.7 RMSE=14.2 40 40
R² = 0.7918
20 20
RME RME ARr (m/day) PredictedARrw (m/day) 0 0
-20 -20 -20 0 20 40 60 -20 0 20 40 60 Actual ARw (m/day) Actual ARw (m/day)
Figure 4- 16. The comparative results of new proposed model and RME model
In general, the higher accuracy of the proposed new model indicates the potential for more accurate prediction of AR for hard-rock TBMs. Further study of this approach depends on the development of new databases that incorporate additional TBM project and ground factors.
83 Discussion and Conclusions
Lack of accurate records of actual TBM operational parameters and key rock mass parameters (such as RQD and water conditions) are primary reasons for the existing scatter in the results. Expanding the existing databases with more parameters and adding more precise and consistent details on different geological zones should allow for additional improvement in the results and a higher degree of model accuracy and reliability.
The division of the database for the tunnel diameter into three categories of 2.5 5.5 The new proposed formulae are based on statistical analysis of tunneling records in our databases. Due to the limited amount of details for various parameters included in such data sets, the models cannot rigorously isolate the impact of individual parameters or address the possibility that there is interdependence between various parameters. Some limitations that can be noticed for the proposed formulae are as follows: Applying the new formulae to the testing dataset shows that these formulae do not offer accurate results for TBM operation under extreme conditions associated with phenomena such as high water inflow, gassy ground, and soft ground (e.g. extreme conditions discussed by Farrokh et al. (2006, 2011), Farrokh and Rostami (2007, 2008, 2009), Pelizza (1998), Peila and Pelizza (2009), Grandori et al. (1995), Grandori (1996a, 1996b)). The new formulae lack the full accounting of the impact of the tunnel length parameter noted earlier. As such, these results are not reliable and adjustments are 84 needed to predict machine performance for very short or extremely long tunnels. The results are approximate and reflect the average expected advance rate over the length of a tunnel section where the rock mass conditions are consistent. Given these issues, the proposed models should be used with caution on any new project since they lack potentially important machine and ground parameters. The models do serve to offer preliminary machine performance estimates that should be used in conjunction with other models (especially the RME model) to estimate machine advance rate at an early stage of a TBM project plan. In summary, the user should recognize that the number of input parameters into this model is very limited, while any of the additional parameters can impact the end results. The best use of the proposed model is in planning-level estimates or for cross checking the estimated rates against the historical trends. 85 Chapter 5 Penetration Rate Prediction Introduction A key component in successful planning of TBM tunneling is the accurate prediction of the TBM Penetration Rate (PR). Early applications of TBMs were mainly in relatively massive medium-strength rocks. In such rock masses researchers focused primarily on evaluating the influence of intact rock properties on PR for a given set of TBM parameters. As the use of TBMs and the range of application of the TBM has expanded, TBMs are now frequently used in a wider range of rock mass conditions. Since joints and discontinuities within a rock mass impact TBM performance, a need for an improved penetration rate predictive model for TBMs operating in fractured rock units became evident. Many of the earlier models could not address the impact of discontinuities on TBM penetration rate. Consequently, attempts were made to either modify existing models or develop new models that included rock mass parameters. These models can be categorized into four classes as shown in Table 5-1. One of the main challenges in developing predictive methods for TBM performance is accounting for the interaction between TBM and rock mass. To better model the complexity of this interaction, new tests and indices have been developed. Special testing such as rock boreability, drillability, and indentation are among the specialized tests for an improved prediction of penetration rate for the TBM (Bruland, 1998). Another approach is to simulate the process of rock fragmentation in a laboratory setting in full-scale cutting tests. These tests are described by Rostami (1993, 1997, and 2008), Sato et al. (1991), Sanio (1985), and Ozdemir et al. (1978). Only a few laboratories around the world can perform such tests, therefore where these 86 tests are unavailable TBM performance predictions are based on an adjustment of performance data from sites where rock with similar strength properties were encountered. The estimated rates should be adjusted for rock mass parameters and may introduce significant errors in the estimating process. The amount and likelihood of errors being introduced depend on the accuracy of the underlying model assumptions, and the quality and quantity of TBM-related and ground conditions data. Table 5- 1. Advantages and disadvantages of empirical TBM performance prediction models Example Typical Advantages Typical Disadvantages Simple Models Graham (1976) Easy to apply Might underestimate due to lack of joint parameters Limited range of application Multiple Parameters CSM (Rostami, Accounting for both rock mass Several parameters Models 1993, 1997), NTNU and TBM parameters Complex relationships (Bruland, 1998), Relying on good database Using uncommon tests QTBM (Barton, 1999) Probabilistic Models Laughton (1998) Accounting for randomness and Lack of detailed information approximation from a like-case tunnel Computer-Aided Models Neural network Relying on good database Complex underlying structure models (e.g. Over fitting Alvarez, 2000a, Usually not available in public 2000b) domain Simpler models are often preferred because of their ease of use, but they include only a few basic input parameters (e.g. rock compressive and tensile strength) and can only offer a limited range of application. As such many of the parameters that influence TBM performance in more variable ground conditions, such as rock mass properties (e.g. RQD and rock type), are unaccounted for in the modeling process. Probabilistic models offer a more complex methodology for estimating performance. These models should only be used when it can be demonstrated that the detailed information (e.g. probability distribution functions for various parameters) of a similar tunnel is available to support the prediction of TBM performance on a new project. These models use performance data collected from similar case histories. If there are 87 significant differences in ground conditions or technology choices between the new drive and case histories within the database, substantial errors are likely to be introduced in the estimates when using these models. Another potential problem, which is also common for computer-aided models, such as the fuzzy logic model by Alvares (2000), is that in practice these models are rarely used for TBM performance prediction purposes, even though they offer several advantages over the other methods (e.g. having a higher correlation coefficient and taking complex formula structures wherever needed). This is due to the lack of transparency of the process. Models with multiple parameters use more project-specific data (compared to simple models), and therefore, these models are among the most-favored models used in TBM performance prediction. For evaluation of the existing models the detailed database of TBM field performance was used to represent the new projects. This chapter discusses the development of a new model that can be used for the estimation of TBM penetration rate based on the analysis of data from the general database The results of predictions are compared with the detailed database where more detailed ground and machine parameters are available. Objective or Target Parameter A general review of previous TBM penetration rate models indicates that different target parameters were used to support the modeling of TBM penetration rate. Typically, these models predict performance parameters such as PR and PRev (penetration per revolution of the cutterhead). Other indexed parameters used in some of the prediction models include Specific Penetration (SP) and Field Penetration Index (FPI). These parameters are defined in Table 5-2. 88 Table 5- 2. TBM penetration rates Description Typical Unit Formula PR Penetration Rate m/hr PRev Penetration Rate per revolution mm/rev (1000 PR)/(60 RPM) SP Specific Penetration (mm/rev)/(kN/cutter) PRev/Fn FPI Field Penetration Index (kN/cutter)/(mm/rev) Fn/PRev SER Specific Excavation Rate (m3/rev)/(kN/cutter) A*SP A: Tunnel cross section area, Fn: Normal force per cutter The purpose of using SP (Alber, 2000) is to combine the thrust and cutter head rotational speed with penetration rate so that the penetration rate can be normalized against variations in rock mass strengths (as shown by Alber, 2000). Alber (2000) noted a general correlation between rock mass strength and SP without including any strength factor for the correlation. FPI was also introduced as a normalized measure of penetration rate, specified in terms of applied thrust on cutters for a specific PRev in different geological conditions (Hamilton and Dollinger, 1979; Nelson et al., 1983; Klein et al., 1995). Nelson et al. (1983) offered a relationship between FPI and total hardness (Ht). Klein et al. (1995) presented correlations between FPI and intact rock and rock mass parameters with the primary goal of presenting different classes for different ground conditions in four tunnels. More recently, Hassanpour et al. (2009) and Khademi et al. (2010) used FPI as an objective variable in a multiple regression setting to offer a new way of estimating the rate of penetration. In fact, SP is the inverse of FPI. The advantage of using these combined or normalized parameters is that they can be applied on machines of various sizes and account for cutterhead rotation (RPM), which is typically inversely related to cutterhead diameter. Stevenson (1999) introduced Specific Excavation Rate (SER) as the excavated volume per revolution divided by thrust per cutter to combine SP and the tunnel cross-sectional area. One benefit of using a normalized penetration rate is to allow the inclusion of a TBM-rock mass interaction factor. Laughton (1998) maintained that SP is not an appropriate parameter for TBM performance prediction purposes as it does not reflect the true non-linear nature of the PRev:Fn 89 (penetration-cutter load) trend. Fig. 5-1 shows a typical nonlinear relationship between PRev and Fn (cutter load) and a typical assumption for SP (straight line). As can be seen, the slopes of these two lines might be different drastically. The accuracy of TBM penetration rate estimates based on SP or FPI may contain a significant error if an SP from one tunnel drive is used directly to predict TBM performance on a new tunnel project where a different cutter load is to be applied. This refers to the fact that for the estimation of the actual PR from these indices, the model requires an assumed level of the cutter load. The possibility of applying a certain cutter load on a given TBM requires a close examination of TBM power and torque capacity. In other words, the cutter load rating of a given cutter size cannot be reliably used for PR estimation since in reality the TBM may not have sufficient power to turn the head at the required thrust level. The difference between the results are especially high if the applied cutter load is near the range of the threshold thrust where there is a change of slope in the PRev-Fn curve. PRev, mm Operating Point Line Slope= SP, mm/kN PRev Predictor Fn, kN Thrust Intercept Critical Thrust Value Figure 5- 1. Relationship between penetration per revolution (PRev) and cutter normal force (Fn) (modified from Laughton, 1998) Table 5-3 shows some examples of the significant differences that can exist between nominal cutterhead thrust and RPM capacities and operating values in various geological settings. Even in relatively hard rock (Diorite, Meta-Volcanic) common operating thrust and RPM values 90 can be far away from the design values. Part of the reason for such large discrepancies is related to the fact that there are interactions between the strength of the rock, the thrust level, and TBM cutterhead design parameters. This issue is addressed in the Colorado School of Mines (CSM) model (Rostami et al. 1993; Rostami, 1997). This model was based on the results of full-scale cutting tests, where the relationship between PRev and Fn follows a curve that is closer to actual cutting behavior of a disc cutter. Table 5- 3. TBM operational parameters in different settings RPM Fn (kN) (Gross) Tunnel Name Rock Type Design Max Applied Design Max Applied Ghomroud III&IV Limestone 12 11 ~220 170 Ghomroud III&IV Slate 12 6 ~220 90 Ghomroud III&IV Meta-Volcanic 12 10.3 ~220 130 Ghomroud III&IV Graphite Schist 12 8 ~220 100 Karaj Tuff 11 7 ~220 150 Zagros Limestone 11 6 ~220 100 Golab Diorite 12 8.6 ~220 125 Maen* Serpentinite 11 ~220 155 Milyang Granite 13 10 ~195 160 Queens Granitic Gneiss 8.3 8.3 ~320 250 *Obtained from Sapigni et al. database (2002) The different relationships between the rock strength, cutter thrust, and TBM parameters were very well defined by Frenzel et al. (2008) and are shown in Fig. 5-2. A key reason for observing lower thrusts in the field might be related to the TBM encountering weaker rock mass conditions. In this regard, a big problem for shielded TBMs is that the Fn value used in the predictive calculation is a gross value and does not include any friction losses. The net Fn delivered to the cutterhead may be significantly less than the nominal cutter load calculated from the applied machine thrust. As Laughton (1998) notes, SP and FPI are preferred for the evaluation of TBM performance in the field, where the TBM is operated over the critical thrust (e.g. massive rocks with higher UCS value). In softer, more-jointed rock masses, using these thrust-normalized indices are not necessarily the critical performance factors and there are several other parameters 91 that can come into play, which overshadow the effects of the TBM operational thrust or cutterload. Therefore, normalized factors, FPI and SP, should be used with caution before providing a penetration rate estimate for a new project. Meanwhile for the development of alternative models, PR can be used in statistical studies to allow for an analysis of the real relationship between the various parameters, including TBM diameter. This is the approach that is used in the following sections of this chapter. Figure 5- 2. Operating limits of a TBM with 17″ disc cutters at different rock strengths after Frenzel et al. (2008) Evaluation of Existing Penetration Rate Models The purpose of the study described in this section is to test the capability of some of the commonly used or recently developed TBM performance prediction models as applied over a range of tunnel diameters and rock mass conditions. Table 5-4 summarizes more commonly used TBM performance prediction models. These models have been developed since the early 1970’s. 92 Table 5- 4. TBM performance models Author/Model Year Comment Tarkoy 1973 For limestone, shale, sandstone, quartzite, orthoquartzite, schist, dolomite with total hardness of 2-242 and penetration rate of 0.076-3.716 m/hr. Roxborough and 1975 For UCS of 70-205 MPa, Tensile strength of 5.5-13.8 MPa, Cutter tip width of Phillips 11.4-19 mm, Cutter diameter range 382-432 mm. Graham 1976 PRev=3940 Fn/UCS. Ozdemir et al. 1978 Based on The Robbins Company data in granite, quartzite, schist, and shale. Farmer and 1980 Based on six tunnel projects' data. PRev=624 Fn/TS. Glossop Cassinelli 1982 Using RSR. PR=-0.0059 RSR+1.59 Snowdon et al. 1982 A Formula to demonstrate relationship among normal force, rolling force and penetration per revolution. Lislerud et al. 1983 Based on excavation records in Norway in shale, limestone, gneiss, basalt. Nelson et al. 1983 Based on information of four tunnels in sedimentary rocks Bamford 1984 Based on data of tunneling in claystone on the Thompson project in Australia for bedding spacing range 0.3 to 0.5 m. Sanio 1985 Effect of foliation on penetration rate. Hughes 1986 For sandstone and penetration of up to 10 mm/rev. Boyd 1986 On the basis of cutterhead power, specific energy, and tunnel cross section area. Sato et al. 1991 Followed Sanio's work and used the same approach, but on a rotary cutting machine Innaurato et al. 1991 Updated version of the method presented by Cassinelli, see above. Based on 112 homogeneous sections. No information is provided on the number of bored tunnels. PR=UCS-0.437- 0.047 RSR + 3.15 Rostami and 1993 CSM model. On the basis of LCM tests. Ozdemir Sundin and 1994 For granite, micaschist, gneiss with UCS range 65-200 MPa, point load range Wanstedt 1-9 MPa, CAI=1.9-5.9, Toughness of 2.2-3.3 Haworth et al. 1986 Based on information of excavation in sandstone and marble with Fn=3.16 kN and RPM=14. Rostami 1997 Updated CSM model. On the basis of LCM tests. PRev=f (Fn,Fr) Bruland 1998 NTNU model. PRev = [Mekv / M1] b Barton 1999 -0.2 Qtbm model. PR=5 Q TBM Cheema 1999 Based on information of one project to modify CSM model. Alvarez 2000 Neuro-Fuzzy modeling. a Yagiz 2002 Based on information of one project to modify CSM model. PR=0.859+RFI+BI+0.0969 PRevCSM; RFI=1.44 Log(α)-0.0187 JS; BI=0.0157 Ps. -0.66 Ribacchi and 2005 SP=250 UCScm Lembo Fazio UCScm=UCS exp((RMR-100)/18) Ramezanzadeh 2005 Based on information of 11 projects to modify CSM model. 0.37 et al. PRev=PRevCSM exp(1.8-0.0031 JS-0.0065 α) Gong 2005 Based on information of one project. Hassanpour et 2009 Based on information of two projects. FPI=0.425 RMCI+11.28 al. a RMCI=0.01 UCS RQD2/3 and 2009 b Khademi et al. 2010 Based on information of one project. FPI=4.161+0.091 UCS+0.077 93 RQD+0.117 Jc+1.077 Log(α) Nomenclature: PR: Penetration rate, PRev: Penetration per revolution, SP: Specific penetration rate, FPI: Field penetration index, Fn: Cutter normal force, Fr: Cutter rolling force, UCS: Uniaxial compressive strength of intact rock, BTS: Brazilian Tensile strength, UCScm: Rock mass uniaxial compressive strength, RSR: Rock Structure Rating, RMR: Rock Mass Rating, RQD: Rock Quality Designation, Mekv: Equivalent cutter thrust (kN/cutter), M1: Critical cutter thrust (kN/cutter), that is necessary thrust to achieve 1 mm/rev, b =Penetration coefficient, α: The angle between the tunnel axis and the planes of weakness, Ps = Peak Slope Index (obtained from Punch Penetration test), Fs/Js= Fracture/Joint spacing, RFI: Rock fracture index, BI: Brittleness index, Jc: RMR joint condition partial rating, QTBM: Barton rock mass quality rating for TBM driven tunnels, RMCI: Rock Mass Cuttability Index, LCM: Linear Cutting Machine. Among the models in Table 5-4, 12 models (with given formulae) were selected for evaluation. These models were selected based on the availability of the required information logged in the testing database of TBM field performance. The graphs in Fig. 5-3 compare the actual and predicted PRs for the selected models. A 45-degree dashed line (1:1line) represents the line where predicted and actual rates are the same. Points plotted above the dashed line indicate an over-estimate of PR by predicting models. 94 40 60 32 Graham (1976) Farmer and Glossop Ribacchi and Lembo Fazio (1980) (2005) 30 45 24 20 30 16 Predicted PR (m/hr) PR Predicted (m/hr) PR Predicted 10 PR (m/hr)Predicted 15 8 0 0 0 0 10 20 30 40 0 15 30 45 60 0 8 16 24 32 Actual PR (m/hr) Actual PR (m/hr) Actual PR (m/hr) 8 6 8 Cassinelli et al. Innaurato et al. Yagiz (2002) (1982) (1991) 6 4 6 4 2 4 Predicted PR (m/hr) PR Predicted Predicted PR (m/hr) PR Predicted 2 0 2 Predicted PR (m/hr) PR Predicted -2 0 2 4 6 0 -2 0 0 2 4 6 8 Actual PR (m/hr) 0 2 4 6 8 Actual PR (m/hr) Actual PR (m/hr) 12 8 12 CSM Model NTNU Model Qtbm Model 9 6 9 6 4 6 Predicted PR (m/hr) PR Predicted Predicted PR (m/hr) PR Predicted 3 2 (m/hr) PR Predicted 3 0 0 0 0 3 6 9 12 0 2 4 6 8 0 3 6 9 12 Actual PR (m/hr) Actual PR (m/hr) Actual PR (m/hr) 16 8 12 Hassanpour et al. Ramezanzadeh et al. Khademi et al. (2009) (2005) (2010) 12 6 9 8 4 6 Predicted PR (m/hr) PR Predicted (m/hr) PR Predicted Predicted PR (m/hr) PR Predicted 4 2 3 0 0 0 0 4 8 12 16 0 2 4 6 8 0 3 6 9 12 Actual PR (m/hr) Actual PR (m/hr) Actual PR (m/hr) Figure 5- 3. Results of comparison for the selected models It should be noted that the upper cluster of the CSM model plot is attributed to pump limitations (e.g. maximum propel rate of the hydraulic thrust cylinders during the stroke extension). This means that with the proper information on the applied cutter load, the estimated 95 PR can be adjusted. The three models of Cassinelli et al. (1982), Innaurato et al. (1991), and Yagiz (2002) tend to underestimate TBM performance. One potential problem of these particular models may be related to the absence of any parameter that would account for a variation of tunnel diameter. The absence of a diameter parameter may be a result of the limited number of case histories that were referenced in the development of these models. The penetration rates of larger diameter TBMs are generally lower than those of smaller TBMs. Yagiz (2002) modified the CSM model using field data collected on the performance of a relatively large-diameter TBM tunnel (Queens Tunnel with the diameter of 7.06 m) and does not include a parametric adjustment for TBM diameter. As such, the corresponding graph in Fig. 5-3 shows the variation of this model for the Queens TBM or its equivalent in other ground conditions. The results of the remaining comparisons shown in Fig. 5-3 indicate that these models also generally tend to overestimate the PR. The percent differences between predicted and observed values can be more than 100%. Some likely causes of the tendency of models to overestimate PR are listed as follow: Limited database use in the development of the models; Exclusion of influential parameters, such as tunnel diameter, from the model due to the limited range of TBM diameters in the original databases used for development of the models or the omission of key physical relationships between TBM parameters (i.e. RPM and hence PR are directly related to TBM diameter); Use of inappropriate, inferred, estimated, or inaccurate parameters; Lack of adequate TBM operating information, especially in weaker rock masses; Use of too many parameters. In applying the model interactions between the many parameters may result in unrealistic result. The need for multiple parameters may also oblige the estimator to guess at multiple missing input values; 96 The mere fact that the predictions are based on the machine’s installed capacities (cutter load, power, etc.) whereas in reality machines are operated at lower thrust levels to cope with other field parameters. To provide an example of the latter problem, consider the model introduced by Khademi et al. (2010). The bivariate analysis shows that UCS by itself accounted for 70% of the variation of the FPI. Adding three more parameters led to only a marginal increase in R2 from 0.7 to 0.77. This means that the effects of the additional parameters were largely overshadowed by UCS. It is worth noting that even with very detailed information on selected sections of a tunnel drive, some models still fall short of yielding satisfactory results. From this perspective, although all these models were calibrated to field performance data at one level or another, their predictive power is limited. These models can only be expected to provide realistic predictive results when the new set of conditions is very similar to the conditions on which the models were based. The Nuero-Fuzzy model developed by Alvarez et al. (2000a) is another analytical approach for PR prediction. This model used four major rules as described by Alvarez et al. (2000a). Each rule is a linear combination of five parameters. These rules were obtained based on the Takagi-Sugeno fuzzy method (Takagi and Sugeno, 1985) in combination with the least square method. Fig. 5-4 depicts the recreated Nuero-Fuzzy model rules (r1 through r4) using Excel spreadsheet formulae expressions: r1: If CFF is low and UCS is medium and RPM is medium and Thr/c is medium and Dsize is large, then PR= -0.7288* CFF -0.01444* UCS +0.1076* RPM +0.001287* Thr/c +0.006937* Dsize +0.937 r2: If CFF is medium to high and UCS is medium and RPM is high and Thr/c is medium and Dsize is medium, then PR= -1.95* CFF +0.05495* UCS +0.13* RPM +0.03825* Thr/c +0.04546* Dsize -24.65 97 r3: If CFF is very high and UCS is very low and RPM is low and Thr/c is low and Dsize is small, then PR= -9.639* CFF +0.1399* UCS +3.332* RPM +0.0511* Thr/c - 0.009726* Dsize +1.319 r4: If CFF is medium to high and UCS is high and RPM is high and Thr/c is low and Dsize is large, then PR= -1.459* CFF +0.06171* UCS +1.943* RPM +0.3512* Thr/c - 0.2676* Dsize -8.085 The input parameters for this approach include CFF or the core fracture frequency, UCS in MPa, RPM of the cutterhead, thrust per cutter in kN, and the cutter diameter size in mm. The calculation methodology is explained in detail in Alvarez and BabusÏka (1999). When applying these rules within the testing database, the rules applied in isolation do not generate reasonable PR values. The application of rules number 2 to 4 are particularly suspect as they yield results that are mostly negative. Furthermore, for certain sets of parameters this model yields negative values (see Fig. 5-5). The occurrence of negative results may indicate that although the fuzzy logic models may yield better overall performance accuracy, the interpretation and use of such models might be more problematic. 98 Rule No. 1 1 1 1 1 1 0.5 0.5 0.5 0.5 0.5 w1=0.32, PR=1.9 0 0 0 0 0 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 1 1 1 1 1 2 0.5 0.5 0.5 0.5 0.5 w2=0.54, PR=2.1 0 0 0 0 0 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 1 1 1 1 1 3 0.5 0.5 0.5 0.5 0.5 w3=0.02, PR=41 0 0 0 0 0 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 1 1 1 1 1 0.87 0.7 0.6 4 0.5 0.5 0.5 0.5 0.5 w4=0.87*0.7*0.6* 0.24 0.24=0.03, PR= -10.2 (from 0 0 0 0 0 rule 4) 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 CFF=2 UCS=105 RPM=11.5 Thr/c=190.6 Dsize=355 PR=(0.32*1.9+0.54*2.1+0.02*41- 10.2*0.03)/(0.32+0.54+0.02+0.03)≈ 2.5 Figure 5- 4. Recreated example of the neuro-fuzzy model recreated in an Excel sheet from Alvarez et al. (2000a) Rule No. 1 1 1 1 1 1 0.5 0.5 0.5 0.5 0.5 w1=0.07, PR=1.55 0 0 0 0 0 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 1 1 1 1 1 2 0.5 0.5 0.5 0.5 0.5 w2=0.11, PR=-3.1 0 0 0 0 0 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 1 1 1 1 1 3 0.5 0.5 0.5 0.5 0.5 w3=0.14, PR=-1.6 0 0 0 0 0 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 1 1 1 1 1 4 0.5 0.5 0.5 0.5 0.5 w4=0.006, PR= -79.5 0 0 0 0 0 1 2 3 10 105 200 2 11 20 67 191 315 230 355 480 CFF=3 UCS=60 RPM=5 Thr/c=100 Dsize=432 PR≈ -2.9 Figure 5- 5. An example of the recreated neuro-fuzzy model (Alvarez et al., 2000a) yielding negative value 99 Proposed New Model FPI was used as the objective parameter for the development of the new model for the estimation of TBM penetration rate under a given set of ground conditions, using linear regression analysis (using Minitab 16). As noted before, the different tunnel records within the database were heterogeneous. Only a limited number of records could be used in the regression analysis, thus reducing the population size used in the statistical analyses. As for statistical process for this study, the first step involved bi-variate analyses between penetration rate and other parameters to identify influential parameters. The second step included multiple regression analyses to develop a best-fit combination of key parameters that demonstrate the strongest correlation to FPI. Fig. 5-6 depicts the correlation between some of the most important independent variables and FPI. Coefficients of determination (R2) of the bivariate analyses (Fig. 5-6) show the strongest correlation between FPI and UCS. 100 200 200 S 16.0531 S 30.0047 R-Sq 75.6% R-Sq 14.2% R-Sq(adj) 75.3% R-Sq(adj) 13.6% ) ) 150 v 150 v e e r r / / m m m m / 100 / 100 N N k k ( ( I I P P F 50 F 50 0 0 0 50 100 150 200 250 300 350 20 30 40 50 60 70 80 90 100 110 UC S (MPa) RQD 200 S 25.5366 R-Sq 37.9% R-Sq(adj) 37.4% ) 150 v e r / m m / 100 N k ( I P F 50 0 0 1 2 3 4 5 6 7 CAI Figure 5- 6. Correlation between FPI and other parameters As can be seen in Fig. 5-6, FPI is at a maximum at higher UCS levels. As UCS increases, PR and PRev increase. This is a logical trend and is in agreement with several research studies such as those reported by Laughton (1998), Robbins (1992), Hassanpour et al. (2009), and Khademi et al. (2010). Multivariate Regression Analysis A multivariate regression analysis with FPI as the objective parameter was performed using Minitab 16 (Minitab Inc., 2010). This analysis allowed FPI results to be projected over different machine sizes, with FPI converted to PR using RPM and Fn. Based on this analysis, a best-fit linear regression model was identified. A transformation of the objective parameter was 101 made wherever it was necessary to correct for normality in the regression model (i.e. Ln(FPI) for Eq. 5-1. The results of these analyses are shown in Table 5-5, Fig. 5-7, and Eq. 5-1. 5.5 5.0 e s n 4.5 o p s e 4.0 R d e t 3.5 a l u c l 3.0 a C 2.5 2.0 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 A ctual Response Figure 5- 7. Results of linear regression analysis for FPI (response is Ln(FPI)) 2 FPI Exp(1.97 0.0063 RQD 0.103 CAI 0.00685 UCS) R 85% (5-1) 0.06 RPM Fn PR FPI (5-2) where RQD is rock quality designation, CAI is Cerchar Abrasivity Index, UCS is uniaxial compressive strength in MPa, RPM is revolution per minute, and Fn is disc cutter normal force in kN. Table 5- 5. Regression coefficient statistics for FPI prediction Predictor Coef SE Coef T P VIF Constant 1.9695 0.08382 23.5 0.000 RQD 0.006303 0.001231 5.12 0.000 1.218 CAI 0.10322 0.01407 7.33 0.000 1.4 UCS 0.006852 0.000459 14.93 0.000 1.477 102 As can be seen in Fig. 5-7, the modeled FPI values show good correlation with actual values. Furthermore, the new model yields good improvement in predictive capacity over the other models evaluated (Fig. 5-8). 6 5 ) r h / m 4 ( R P 3 d e t c i d 2 e r P 1 0 0 1 2 3 4 5 6 A ctual PR (m/hr) Figure 5- 8. The comparative results of Eq. 5-2 In general, the proposed new model offers better results than those of the previously discussed models and indicates a potential for better PR prediction. Further study of this approach will require the development of new databases that report additional TBM and ground condition factors. Discussion and Conclusions Comparisons between predicted and actual TBM performance indicated that most of the existing predictive models, especially the simple ones, cannot offer accurate estimates of TBM performance for new projects. A study of these models indicates that using more parameters in a 103 model will not always guarantee improved results due to the lack of inherent limits of the initial models. In some cases, existing models do not include important parameters; as such they cannot adequately distinguish between the ground conditions and job constraints that control TBM performance. To achieve more accurate estimates, the database of TBM field performance was subjected to a statistical analysis. The proposed models offer better accuracy than existing models. Part of the reason for the model improvement may be related to the ability to utilize a larger database of TBM field performance records that includes a wider range of tunnel diameters and ground conditions. Expanding the existing databases with more parameters and adding more details on the different geological zones should allow for more improvement in the results and a higher degree of accuracy and reliability in model prediction. The analyses of the available data indicate that certain sets of available parameters including RQD, UCS, and CAI account for most of the FPI variation. These analyses further indicate that UCS is the single most-important rock parameter controlling FPI. Obviously, frequency and condition of jointing can have a dominant impact on TBM performance, especially in harder rocks, and further study of this phenomenon, notably including a quantitative representation of joint spacing and condition, may be needed to improve model accuracy in harder rock units. There are other areas of concern related to the use of TBM design parameters in performance prediction. The TBMs in operation may only use a fraction of the torque and thrust capacity that is installed in the factory. In the development of new models or the modification of existing models, one should be able to evaluate performance over a range of physical scales, perhaps zone-by-zone analysis, which based on previous experiences can yield the best results. Additionally, one should always be mindful of the tradeoff between model complexity, accessibility of the relevant data and rock property tests for various models, and the desired 104 accuracy and reliability of the model. The proposed new formula can serve the purpose of offering preliminary TBM performance estimates. Used in conjunction with anticipated utilization rates, the models can be used to estimate TBM advance rates for the early stage of a project-planning process. 105 Chapter 6 Downtime Analysis Introduction Among TBM performance parameters, utilization is one of the hardest parameters to model or predict. There are very few models for estimation of the TBM utilization, mostly developed more than two decades ago by CSM and NTNU, and there does not seem to be an in- depth study of TBM utilization in recent years. TBM tunneling is usually performed in a series of cyclic operations which include several activities repeated in sequence. In each excavation cycle, individual activities can cause certain delays, which are usually referred to as the TBM being "down", hence downtime (Nelson et al., 1985). TBM performance and daily advance rate depends on the duration of these down times. As the proportion of downtimes increases, the performance and production of the TBM and hence the footage of tunnel mined in a given time frame decreases. For example, in poor ground, the duration of time spent on ground support installation or ground improvement increases, which results in low utilization, even as low as 10-15%. Understanding the causes of downtimes is the key to successful planning of the TBM tunneling and improving machine performance. In this study, a database of 90 tunnel projects from 20 countries is used to analyze various components of downtimes. This refers to the most frequent causes of downtime and examination of the accuracy of previous TBM utilization models. Some suggested modifications are proposed for better prediction of downtime based on project settings and TBM performance parameters. 106 Description of the TBM Field Performance Database The database for machine performance and downtime analysis includes 89 tunnel projects from 20 different countries, obtained from published papers and reports. This database is a subset of the database on TBM field performance that was discussed earlier. In this group of projects, information on activity time and downtimes were available. The projects were completed between 1975 and 2009 with the various TBM diameters ranging from 2.1-11.52 m and tunnel lengths of 134-17,040 m. The ground conditions in the database vary from poor to good, and in different rock types from sedimentary to volcanic. Table 6-1 and Fig. 6-1 represent descriptive information on the tunnel projects, TBM types, and backup equipment. The majority of the cases were excavated by an Open type TBM as depicted in Fig. 6-1. Table 6- 1. Summary of the tunnel projects in the database Number of TBM Type Location Tunnels Open SS* DS* Australia 1 1 Austria 1 1 Canada 5 4 1 China 3 1 2 Ecuador 2 2 Hong Kong 4 2 2 India 1 1 Iran 4 4 Island 4 4 Italy 11 8 3 Japan 1 1 Korea 1 1 Norway 6 6 SA 8 8 Slovenia 2 2 Sweden 3 3 Switzerland 2 1 1 Turkey 2 2 UK 5 5 USA 23 19 3 1 Total 89 66 7 16 * SS: Single Shield, DS: Double Shield 107 80 70 67 60 50 40 No. of TunnelofDrivesNo. 30 20 16 10 7 0 Open SS DS TBM Type 70 61 60 50 40 30 No. of TunnelofDrivesNo. 20 13 10 10 4 1 0 A S H V T Transportation System 60 55 50 40 30 20 16 No. of TunnelofDrivesNo. 10 5 4 3 2 1 2 1 0 UN Wirth Jarva Lovat Demag Dresser Robbins Voest AlpineVoest Herrenknecht TBM Manufactures Figure 6- 1. Histograms of different information of the database 108 Using the different records of downtimes presented in available literature, the main downtime categories and main tunneling activities were identified as listed in Table 6-2. Downtime associated with each incident is usually reported in percentage of the total shift time. When different categories were reported in the source, a cross mapping of the downtime components was applied where the reported delays were mapped on to the items listed in Table 6- 2 that best matched or described the related activities. Table 6- 2. Downtime categories identified in different tunnel projects No. Category Name Definition 1 TBM , Ttbm TBM breakdowns times 2 BU, Tbu Back-Up breakdowns times 3 Cutter, Tc Cutter check/change time 4 Support, Tsp Support installation time (planned) 5 Regrip, Tr Resetting times of TBM after each excavation stroke 6 Transport, Ttr Times related to muck transportation and unloading 7 Maintenance, Tm Routine maintenance of cutter head, TBM, and Back-Up 8 Ground, Tg Downtimes related to unfavorable ground conditions (additional or supplementary support) 9 Probe, Tp Probing times for ground exploration 10 Utility, Tu Line extension times 11 Survey, Ty Times for changing surveying stations and checking tunnel direction 12 Other, To Unclassified times Note: Some machine types do not require certain activities (i.e. single shield and 8 and 9) Fig. 6-2 to 6-4 show the histograms of downtime categories for three types of rock TBMs in the database. As can been seen, most of the distributions are skewed to one side. Table 6-3 lists the typical parameters affecting the components of downtime and various activity items. 109 16 60 30 14 50 25 12 40 20 10 8 30 15 6 20 10 No. of TunnelDrivesofNo. No. of TunnelDrivesofNo. 4 TunnelDrivesofNo. 10 5 2 0 0 0 10 15 20 25 30 35 45 50 55 60 65 05 40 05 10 15 20 25 30 35 05 15 25 10 20 ------ ------ - - - - - 05 10 15 20 25 30 40 45 50 55 60 00 35 00 05 10 15 20 25 30 00 10 20 05 15 Utilization BU TBM 35 25 60 30 50 20 25 40 15 20 30 15 10 20 No. of TunnelDrivesofNo. TunnelDrivesofNo. No. of TunnelDrivesofNo. 10 5 10 5 0 0 0 05 10 15 - - - 05 10 15 25 30 35 40 50 55 60 05 10 15 20 25 30 35 40 45 50 55 60 65 70 20 45 ------ 00 05 10 00 05 10 20 25 30 35 45 50 55 00 05 10 15 20 25 30 35 40 45 50 55 60 65 15 40 Cutter Support Regrip 30 50 70 45 60 25 40 35 50 20 30 40 15 25 30 20 10 15 No. of TunnelDrivesofNo. TunnelDrivesofNo. No. of TunnelDrivesofNo. 20 10 5 10 5 0 0 0 05 10 15 20 25 30 35 40 05 10 15 20 25 30 35 40 45 50 05 10 15 20 ------ 00 05 10 15 20 25 30 35 00 05 10 15 20 25 30 35 40 45 00 05 10 15 Transport Maintenance Ground 70 70 70 60 60 60 50 50 50 40 40 40 30 30 30 No. of TunnelDrivesofNo. TunnelDrivesofNo. No. of TunnelDrivesofNo. 20 20 20 10 10 10 0 0 0 05 15 25 05 10 15 05 10 15 10 20 ------ 00 10 20 00 05 10 00 05 10 05 15 Probe Utility Survey 110 25 20 15 10 No. of TunnelDrivesofNo. 5 0 05 10 15 20 25 30 35 40 45 50 55 60 65 75 ------ 00 05 10 15 20 25 30 35 40 45 50 55 60 65 Other Figure 6- 2. Histograms of allocated downtimes for different activities for Open TBM (in %) Regrip, Ground, Probe, Utility and Survey were zero 60 50 40 30 Percent 20 10 0 BU TBM Reset Cutter Boring Support Transport Category Name Maintenance Figure 6- 3. Histograms of allocated time for different activities for single shield TBM (in %), different hatches refer to different projects 111 70 60 50 Percent 40 30 20 10 0 BU TBM Cutter Boring Category Name 50 45 40 35 30 Percent 25 20 15 10 5 0 Reset nce Support Category Name Maintena Transport 100 90 80 70 60 Percent 50 40 30 20 10 0 Utility Other Probe Survey Ground Category Name Figure 6- 4. Histograms of allocated time for different activities for double shield TBM (in %), different columns in each category refer to different projects 112 Comparison the Reported Downtimes with Existing Predictive Models The literature review shows that the early study of TBM utilization was conducted at the Earth Mechanics Institute (EMI) of the Colorado School of Mines (CSM) (Ozdemir and Sharp, 1991) and the Norwegian Institute of Technology (NTH or NTNU) (Jonhannssen, 1988, 1994)). In this section the reported downtimes are compared with those predicted by the empirical equations presented by the above-mentioned methods, which relate different downtimes with rock mass and job-site parameters. The purpose of the comparison is to test the predictive capabilities of these models, especially when more recent data are used in the prediction. CSM Method The CSM method was based on analysis of a specific TBM field database compiled by researchers in the mid 1980’s to evaluate TBM utilization and identify major parameters and ways to improve or increase machine advance rate. This approach includes almost all aspects of TBM operations and all activities on a job-site, in addition to ground conditions. For this purpose, delay times associated with machine operations and job-site conditions were predicted in the unit of delay hours per tunnel meter (hr/m) (Table 6-3). 113 Table 6- 3. Prediction of TBM utilization using CSM method (Ozdemir and Sharp, 1991; US Army, 1997) Equations Definition of terms T b Time of boring (hr); T m Time of machine delay; t1 Scheduled maintenance; t2 Unscheduled maintenance; Tr Regrip time; T a all system delays; t s Surveying delays (hr); t w water inflow delays (hr); Tb U (%) 100 t (T T T T T ) f u utility delays (hr); b r m a mu 10 t p Tm t1 t2 Support installation (hr); Tmu t1 0.067Tb mucking delay (hr/m) Mucking method Delay t2 f4 Tb Start up Truck 0.115 15 to 1 Conveyor 0.071 L Tr f3 L 1 to 3 Train 0.056 Ta F(ts ,tw ,tu ,t p ) 3 to 15 Conveyor 0.071 R Radius of curvature of horizontal curves (m); 192 t ( 0.0033) L L Length of tunnel (m); s 2 R 409(m hr) f3(hr / m) 0.03(hr / m) 2 tw f6 L R 1 (hr) tu (0.03 0.0013 ) L f4 0.324 (hr) start up t f L p 9 production phase 0.0056 (hr / m) min imal 3 f 6 0.085 (hr / m) 3 4m / min/ m F(,)(hr / m) high Water inflow rate; Tunnel slope (degrees); 0 hr / m for RMR class I, II, III f9 0.028 hr / m for RMR class IV 0.043 hr / m for RMR class V f 10 1.025 (for labor delay) 114 Using equations listed in Table 6-3 and the reported downtimes for different categories, the predicted values of each downtime item has been calculated and compared with the reported values. The results of this comparison are shown in Fig. 6-5. It should be noted that the charts only illustrate the reported values and respective predicted values and as such, the number of the points in different graphs are different due to the heterogeneity of the available data sets. As can be seen, for the majority of the cases the predicted values are lower than the reported ones and in some cases the difference is several times the predicted values. This means that in most cases, the model underestimates the downtime of the machine, or overestimates the utilization rate. This could most likely be due to the limited data base of the CSM model or absence of any recent tunnel projects and improved machine performance due to technological advances. Furthermore, it seems the database of the CSM model does not include long delays that are common for some projects (see Fig. 6-5). 0.5 Reported 1.4 Reported Tm 0.8 Reported 0.45 Regrip EMI Model Machine EMI Model Mucking EMI Model 1.2 0.7 0.4 0.6 0.35 1 0.3 0.5 0.8 Tr (hr/m) Tr 0.25 0.4 Tm (hr/m) Tm 0.6 (hr/m) Tmu 0.2 0.3 0.15 0.4 0.2 0.1 0.2 0.05 0.1 0 0 0 0 5 10 15 20 25 0 20 40 60 80 100 0 20 40 60 80 Tunnel Drive No. Tunnel Drive No. Tunnel Drive No. 0.03 Reported 0.18 Reported 0.4 Reported Surveying EMI Model Utility EMI Model Support EMI Model 0.16 0.35 0.025 0.14 0.3 0.02 0.12 0.25 0.1 0.015 tp(hr/m) 0.2 0.08 ts(hr/m) tu (hr/m) tu 0.15 0.01 0.06 0.04 0.1 0.005 0.02 0.05 0 0 0 0 5 10 15 0 10 20 30 40 0 20 40 60 80 Tunnel Drive No. Tunnel Drive No. Tunnel Drive No. Figure 6- 5. Comparison between reported and predicted values of different time items of EMI model 115 NTNU Method According to NTH (Jonhannssen, 1988, 1994), in order to predict TBM utilization, some geological, machine, and operational factors should be taken into consideration. These factors will impact the components of TBM operation and various activities and the related times includes mining time, regrip time, cutter change time, TBM/Back-up maintenance time, ground support, and miscellaneous downtimes such as waiting for empty cars, surveying, etc. The formulas for calculating the utilization factor are summarized in Table 6-4. The database of the NTNU model is composed of information from 26 tunnel projects (including some of the high- profile Norwegian tunnel projects completed in the 80’s) that were compiled by Johannessen et al. (1994). As Bruland (1998) noted, this model includes only small amounts of tunnels with extensive rock support requirements. 116 Table 6- 4. NTH method for prediction of the TBM utilization (Johannessen et al., 1988, 1994, US Army, 1997) Equations Definition of terms T U (%) b 100 T b Tb Tt Tk Ttbm Tbu Ta Time of boring (hr/km); T 1000 t Regrip time (hr/km); Tb T Cutter change and inspection (hr/km); I k Maintenance and servicing TBM (hr/km); 1000ttak T tbm Tt 60 Ls Tbu Maintenance and servicing back-up (40 hr/km for single track, 90 hr/km for double track, 55 hr/km for 1000tk T trackless transportation); k L I h T Miscellaneous (time for activities such as a T 150 tbm cleaning, muck car delay, normal rock supporting, surveying, utility in hr/km, 185 hr/km for single track transportation, 95 hr/km for other types); I Machine net advance rate (m/hr); Note: L s Stroke length (m); 1. t k is obtained from cutterheads with front loaded cutters t time per regrip (5.5 or 4.5) changed under favorable working conditions. tak 2. The proposed values for different time items are for "well t time used per changed cutter including time for k organized" tunneling conditions and long failures are not inspection (for cutter diameters≤432 mm is 0.75 hr and included (Bruland, 1998). Therefore, extra times should be for cutter diameters>432 mm is 0.833 hr); considered for unfavorable ground conditions as well as L long delays for major TBM and BU components failures. h Cutter life in hour; In the calculations and graphs generated for comparison of the reported and predicted values the following approaches were used: Reported Ttbm is considered as TBM + Maintenance; Reported Ta is considered as all downtimes items except Regrip, TBM, Maintenance, Cutter, and BU; Cutter life, Lh , is calculated based on the number of changed cutters and total boring time. 117 The predicted and reported downtime values are plotted in Fig. 6-6. Unlike the CSM model, the NTH model has a wider spread from underestimation to overestimation and it gives better results for some cases, especially for cutter change, T . k 200 Reported 600 Reported 1000 Reported Regrip NTH Model Cutter NTH Model NTH Model 500 TBM 160 800 400 120 600 300 Tt (hr/km) Tt 80 (hr/km) Tc TBM TBM (hr/km) 400 200 40 200 100 0 0 0 0 10 20 30 40 0 10 20 30 40 0 20 40 60 80 Tunnel Drive No. Tunnel Drive No. Tunnel Drive No. 250 Reported 1800 Reported BU NTH Model NTH Model Misc. 1500 200 1200 150 900 100 Ta (hr/m) Ta BU (hr/km) BU 600 50 300 0 0 0 20 40 60 0 20 40 60 80 100 Tunnel Drive No. Tunnel Drive No. Figure 6- 6. Comparison between reported and proposed values of different activity time using NTH model Ribacchi and Lembo Fazio’s Proposed Method As Ribacchi and Lembo Fazio (2004) recently noted, in general, the total daily working time, Tp , in which a penetration distance of Lp is obtained, can be subdivided into the following items: - Penetration time Tp - Scheduled maintenance time T0 K0 Td 118 - Unscheduled maintenance time which can be considered proportional to the penetration time (cutter changes, TBM and cutter breakdowns) T1 K1 Tp - Service extension and regripping which are proportional to the penetration length T2 K2 Lp In this approach, there are some coefficients (K0, K1, K2) that are considered as constant values in the mentioned equations. In reality, these coefficients are certainly not constant and vary depending on the job conditions. The graphs in Figure 6-7 are the histograms of the distribution of K0-K2 in the database used in this study, and they show how the three coefficients are scattered for different tunnel projects in different conditions. 18 17 25 16 15 20 14 20 14 18 12 15 14 10 13 9 12 8 10 No. of TunnelofDrivesNo. No. of TunnelofDrivesNo. 6 5 6 5 4 3 5 4 3 2 1 1 1 1 0 0 0 0 0 0 0 0 0 05 10 15 20 25 35 40 50 55 60 65 70 75 30 45 80 05 10 15 20 30 35 45 50 25 40 ------ ------ 00 05 10 15 20 30 35 45 50 55 60 65 70 25 40 75 00 05 10 15 25 30 40 45 20 35 K0 in Percentage K1 in Percentage 18 17 16 14 14 12 10 8 8 No. of TunnelofDrivesNo. 6 5 4 4 3 3 3 3 2 2 2 1 1 1 1 0 0 0 0 0 0 05 10 20 25 30 35 40 50 55 60 65 70 75 80 85 90 95 15 45 ------ 100 - 00 05 15 20 25 30 35 45 50 55 60 65 70 75 80 85 90 10 40 95 K2 *10000 (hr/m) Figure 6- 7. Distributions of coefficients values proposed by Ribacchi and Lembo Fazio (2004) in the database 119 There are several reasons for the scatter of these coefficients in different tunnel projects. Some of these reasons are listed as follows: The definition for the mentioned activities and related time is not unique and consistent between different projects. For example, in some projects, the maintenance includes cutter check/change while in the others, these items are categorized separately. There are some categories in the model that are omitted or ignored, such as transportation delay time. The coefficients are not constant values. Proposed Modifications for Estimation of Various Activity Times While the available data does not allow for offering a comprehensive set of empirical equations for each individual downtime category (or activity time), some improvements can be achieved by considering certain modifications as follows. The proposed modifications to time components are as listed in Table 6-2. In order to obtain reasonable results for each item, the abnormal time distributions for percentage values (cases in which the percentage values of individual activity time were greater than 30% were considered as abnormal) were excluded from the analysis. The excluded cases are either related to the adverse ground conditions or the incomplete recorded data with a high value for the "Other" time category. To obtain better correlations between activity time and ultimately machine utilization and project/operational parameters, a series of statistical analyses were performed. In these analyses 120 the relationship between machine utilization and various parameters were examined. An example of the attempt to establish a relationship between machine performance and geological parameters is the relationship between the UCS and boring time and TBM utilization. Fig. 6-8 shows the utilization values and boring time in hr/km and their relationships with UCS values in the data base. 70 900 R² = 0.3996 R² = 0.1259 800 60 700 50 600 40 500 Utilization(%) 30 400 Boring(hr/km) Hour 300 20 200 10 100 0 0 0 50 100 150 200 250 0 50 100 150 200 250 UCS (MPa) UCS (MPa) Figure 6- 8. Utilization values and boring hour for the different UCS values As can be seen, the R2 values are relatively low. This is partly due to the fact that the UCS values are the average values over a tunnel drive which is not so realistic. Also, there are many other factors affecting machine performance parameters. Additional activity time analysis included the study of relationships between activity time items mentioned in Table 6-2 and other tunnel or rock parameters as follows. 121 Boring Time The common practice in obtaining the boring time is to estimate the penetration rate and then to convert it to boring time (Eq. 6-1 and 6-2) for completion of the project. To estimate the boring time in hr/km one can use the following equations: PR 0.06 RPM Fn / FPI (6-1) 1000 Tb PR (6-2) where RPM is revolution per minute, Fn is thrust per cutter in kN, FPI is field penetration index in kN/mm/rev. Regrip Time On the basis of the information in the database, regrip time is commonly between 20 to 80 hr/km for both open and double shield TBMs. The regrip time can be obtained from the Eq 6- 3. 1000t 409000 T r r 60 L R2 s (6-3) where Ls is stroke length (m), tr is regripping time (min) per stroke which is between 2 to 6 min, and R is radius of curvature of horizontal curves (m). Cutter Change Time Cutter change/inspection time is highly related to penetration rate, rock strength and abrasiveness, and geological setting. Fig. 6-9 shows the results of data analysis for cutter change 122 time for rocks with different quartz contents. The graph includes the impacts of rock strength as part of the penetration rate estimates. 500 Quartz Content ≈ 30-50% 400 300 Tc (hr/km) Tc 200 Quartz Content ≈ 10-20% 100 Quartz Content < 5% 0 0 1 2 3 4 5 6 PR (m/hr) Figure 6- 9. Cutter downtime, Tc TBM Repair Time Fig. 6-10 contains the graphs that show the two most important parameters affecting the TBM downtime, including UCS and penetration rate (PR). In this graph, the impact of tunnel diameter is shown since lower penetration in the rock with a given strength is representative of various tunnel diameters and TBM cutterhead RPM. 123 350 300 UCS>150 MPa 250 200 UCS=100-150 Ttbm(hr/km) 150 MPa 100 UCS<100 MPa 50 0 0 1 2 3 4 5 6 PR (m/hr) Figure 6- 10. TBM downtime, Ttbm Back-Up Repair Time Fig. 6-11 shows the results of analysis for BU-related delays for two different tunnel haulage or mucking systems. 124 120 100 80 Mucking with Train 60 Tbu(hr/km) 40 20 Mucking with Tunnel Conveyor Belt 0 0 1 2 3 4 5 6 PR (m/hr) Figure 6- 11. Back-up downtime, Tbu Maintenance One important issue about maintenance in practice is that it cannot be completely separated from other parallel activities such as utility extension, surveying, probe drilling, etc. Common maintenance-related delays or downtime ranges from 50 to 300 hr/km. The following Table 6-5 gives some guidelines for maintenance time in different conditions. 125 Table 6- 5. General maintenance downtime in different conditions Conditio Tm Comment n (hr/km) Good 50-100 Massive soft to medium rock Normal 100-200 Massive hard rock TBM prone to high clogging and high water inflow in poor cementations, presence of Poor 300 expansive clay, very high rock strength for TBM Surveying Downtime Surveying ranges from 0 to 25 hr/km (close to 0 for most of the cases). In tunnel curves as the CSM model proposed, we can add Ty = 192000/R2, where R is the turning radius of tunnels in m. Utility Installation Downtime Utility extension ranges from 10 to 100 hr/km with an average of 40 hr/km. As proposed by the CSM model, we can add Tu = 1.3 in different tunnel slopes. is the tunnel slope in degree. Transport Related Downtime Table 6-6 shows the approximate muck transport downtime for different conditions. Obviously, in long tunnels, this delay item might increase a lot due to high frequency of equipment breakdowns. This issue is reflected approximately in poor and very poor transportation conditions. 126 Table 6- 6. Muck transport downtime in different conditions T Condition tr Comment (hr/km) Very <50 Tunnel conveyor belt prone to no or very low breakdowns Good Good 50 Tunnel conveyor belt/Train prone to low breakdowns Normal 150 Tunnel conveyor belt/Train prone to normal breakdowns Poor 350 Tunnel conveyor belt/Train prone to high breakdowns (especially in long tunnels) Tunnel conveyor belt/Train prone to very high breakdowns (e.g. simultaneous breakdowns Very Poor >500 for locos, wagons, and switches) Ground Support Installation Downtime In the case of shielded TBMs, the downtime for support is typically fixed for a tunnel project. In the case of open TBM, as RMR value decreases, the demand for ground support installation increases. Fig. 6-12 shows the approximate support installation time for different scenarios. The sharp downturn on the ground support installation downtime in low-RMR values for shielded machines reflect the potential needs for ground improvements in weak rock masses to avoid face collapse and ground squeezing issues. 127 0 Double Shield TBM 200 400 Tsp (hr/km)Tsp 600 Single Shield 800 TBM 1000 1200 0 20 40 60 80 100 RMR Figure 6- 12. Ground support downtime, Tsp Groundwater Condition Related Downtime Water inflow might interrupt the excavation process for different reasons. Some examples are difficulties due to wet muck conveying, pumping, and tunnel face instability. Fig. 6- 13 shows an approximation for downtimes related to water inflow. 128 0 100 200 300 400 Tw (hr/km)Tw 500 600 700 800 900 1000 0 1 2 3 4 Water Condition Code Water condition code: 1: Almost dry 2: Water inflow at tunnel affect the tunnel excavation time (or water inflow/tunnel diameter≈1-3) 3: High water inflow at face (or water inflow/tunnel diameter≈3-4) 4: Water inflow at tunnel face may stop the tunnel excavation (Extreme Mining Area) (or water inflow/Tunnel diameter>10) Note: water inflow in liter/sec and tunnel diameter in m. Figure 6- 13. Downtime related to water inflow, Tw Other Downtimes Consider up to 200 hr/km for the case of a low-experienced crew. For the case of a high- experienced crew, lower the total downtime by 200 hr/km. Comparative Study for Advance Rate Prediction In chapter 4, a direct methodology for advance rate (AR) prediction was introduced. In chapters 5 and 6, two new methodologies for penetration rate (PR) and utilization (U) prediction 129 were presented. From predicted PR and U, it is possible to obtain AR from their multiplication (indirect methodology, Fig. 6-14). Tb: Boring Penetration Rate Time=1000/PR ARw = PR*U*24 (PR) Ttbm: TBM Uw =Tb/ Ʃ Ti Time components (Ti) Tbu: Back-Up 'w' is for working days ( planed days for excavation). Tc: Cutter Units: PR (m/hr) U (%) AR (m/day) Ti (hr/km) Tsp: Support Tr: Regrip Ttr: Transport Tm: Maintenance Tw: Water Tu: Utility Ty: Survey To: Other Figure 6- 14. Indirect methodology for AR prediction From these two methodologies, two ARws can be obtained. Even though these two ARws are not exactly the same (since the methodologies and the input parameters are different), the results are reasonably close to each other. Table 6-7 and Fig. 6-15 show the results of ARw prediction for a couple of tunnel projects as examples using the direct and indirect methodologies. As can be seen the results are very close to each other. Although the indirect methodology has more flexibility in changing the related conditions for different downtime components, it might produce more errors due to the combination of many parameters. On the other hand, the direct prediction of AR benefits from more tunnel records compared to the indirect methodology and has more reliability. 130 Hence, the direct method is proposed as the first option for AR prediction and the indirect method is proposed as the supporting methodology. Table 6- 7. Comparative study for direct and indirect AR prediction methods for 12 tunnels Diameter UCS TBM Tunnel Name Rock Type RMR Fn (kN) RQD PR (m/hr)* Uw (%)* ARw (m/day)** (m) (MPa) Type Ghomroud Sandstone 4.5 53 DS 49 125 60 4.3 (4.4) 29 (26) 30 (31) (28) Zagros Limestone 6.73 50 DS 44 150 60 2.7 (3.0) 30 (35) 20 (22) (25) Pieve Granodiorite 4.05 195 DS 80 220 100 1.5 (1.7) 45 (40) 17 (13) (16) Milyang Granite 2.6 246 Open 84 143 93 0.9 (0.9) 45 (48) 10 (13) (10) Manapouri Granite 10.05 200 Open 61 267 97 1.1 (0.9) 40 (46) 10 (7) (10) Manhattan Gneiss 3.84 62 Open 70 197 80 4.0 (4.3) 32 (27) 32 (32) (28) Frasnadello-Main Argillite 11.8 30 SS 33 100 55 1.7 (1.7) 30 (35) 12 (10) (15) Frasnadello-Pilot Argillite 3.9 60 Open 45 150 60 4.7 (5.1) 22 (19) 25 (21) (24) Rapid transit subway Chalk 6.55 10 Open 60 200 90 ~5 (5.2) - (34) 40-60 (42) (42) River Mt. Conglomerate 4.3 32 Open 60 180 60 9.3 (9.4) 25 (24) 55 (46) (55) Govalle Segment B Chalk 3.2 5 SS 60 180 60 10.6 (11.9) 18 (15) 45 (40) (44) Syar Sedimentary 3.6 50 Open 60 200 60 6.4 (8.5) 30 (21) 47 (44) (42) *(4.4) refers to the predicted value **(31)(28) refer to predicted ARw from direct and indirect methods respectively. 60 Actual ARw 50 Predicted ARw (PR*Uw*24) 40 Predicted Arw (Direct) 30 ARw (m/day) 20 10 0 1 2 3 4 5 6 7 8 9 10 11 12 Tunnel Case No. Figure 6- 15. Results of comparative study for predicted direct and indirect ARw 131 Discussion of Proposed Method for Estimating Utilization of Hard Rock TBMs The study of various causes and components of downtime in the operation of hard-rock TBMs shows the inter-relationship between various parameters and their impact on machine utilization and performance. The existing models seem to fall short of offering a reliable estimate of machine utilization based on the TBM specifications and ground conditions. Part of the problem is the complexity of the jobsite activities and their overlap and parallel or linear relationships as well as the influence of various non-technical or site-management issues on TBM operation that are not directly reflected in various models and their predictions. Thus, the result of different modeling approaches cannot always reflect the detailed variation between the machines, ground conditions, contractor experiences, and site-related requirements. The concluding points of the current chapter can be summarized as follows: There are different approaches for presenting different downtime components and the differences are related mostly to the definition of each individual downtime categoriy as well as total time and also the nature of operational activities that are sometimes simultaneous or performed in parallel (e.g. installation of utility lines while performing maintenance). The previous prediction models of EMI and NTH do not always offer an estimate that is in good agreement with the real TBM records, as shown by data compiled in the database of TBM case histories. A new set of equations is proposed to improve the prediction capabilities of the existing models. The results of the comparison indicate that especially for the lower penetration rates, the NTH model tends to predict higher utilization. To improve the predictive capabilities of the new model, an extensive study of TBM case histories is required. To fulfill this task, a database of TBM field performance 132 has been established and is under expansion, which allows for analysis of various tunneling actives and related time components and resulting delays. 133 Chapter 7 Unit Supporting Time (UST) and Support Installation Time (SIT) Introduction Installation of ground support is one of the essential components of tunneling. Ground support is installed to stabilize the rock mass surrounding the tunnel and provide a safe environment for workers and equipment. If the rock mass is competent and stable, the tunnel does not require support. However, in most cases roof and wall support is necessary because of the limited stand-up time of the rock mass at the heading, or to ensure the safety of the work environment and containment of the possible roof falls. In TBM applications, installation of ground support impacts the overall machine performance. When applying open type TBMs, systematic support installation is necessary in weak ground. As rock mass strength is reduced, more time is required for support installation and the TBM advance rate decreases. Fig. 7-1 illustrates the influence of support installation on the advance rate of a Gripper TBM (Open TBM), compared to the advance rate of a shielded TBM. A gripper TBM can achieve higher advance rates than a shielded TBM only if a small amount of ground support is required. In a highly jointed rock mass, which requires the installation of extensive supports, the risk of experiencing a longer construction period and increased cost for the open type TBMs is higher, thus shielded machines are favored. 134 Figure 7- 1.Normalized tunneling advance rate as a function of support requirement for various TBM types (Schmid, 2004 and Maidl et al., 2008) As can be seen in Fig. 7-1, the tunnel advance rate is almost constant for shielded TBMs over a range of rock conditions (classes). This is because the installation of the segmental lining can be achieved within the shelter of the fully-shielded TBM in every rock class. TBM performance can be improved by increasing the Rate of Penetration (ROP) and/or decreasing the time for ground support installation (if all else remains the same). ROP improvement is limited by the ground material and equipment capacity, such as the maximum permissible cutter loads and the installed torque and thrust. Reducing the time required for ground support installation can also improve machine utilization and thus increase the TBM advance rate. This is especially true for Open type TBMs. Obviously, decreasing other downtime components such as maintenance, utility installation, transportation, surveying, ventilation, etc. is also 135 desirable as part of the jobsite organization, but their contribution to overall downtime are generally small and less affected by ground conditions than ground support installation. Fig. 7-2 shows the excavation support classes and corresponding advance rates achieved in rock masses with different joint spacings in the Misicuni water conveyance tunnel in Bolivia for a gripper-type TBM with an excavation diameter of 3.5 m (WBI, 2007). As can be seen, under massive (un-jointed) conditions, very limited support was required (class I) and advance rates of up to 60 m/d were achieved. In more intensely jointed rock, heavier support was required (class IV). Lower advance rates of 2 to 5 m/d were achieved in this condition. As reported by WBI (2007), massive to slightly jointed rock was only encountered for a very short stretch of the tunnel. The large number of fault zones encountered was unfavorable for TBM operations resulting on a low advance rate. Figure 7- 2. Misicuni water-transmission tunnel, excavation and support classes and rates of advance (WBI, 2007) This example clearly demonstrates the tremendous effect that ground conditions and ground support requirements can have on TBM performance. As noted earlier, this effect has a negative impact on the schedule and cost and can jeopardize project success. This chapter will focus on the review of common ground support systems used in Open TBM projects and discuss 136 the installation time required in different rock mass classes and different support types. The focus on Open TBMs is due to the insensitivity of shielded machines to variation of ground for ground support installation. Ground Support Types A number of support systems are commonly used in conjunction with Open TBM applications. The following list briefly describes these systems. Rock Bolt (Swellex bolt, expansion bolt, resin fiberglass bolt) for stabilizing local blocks, mitigation of rock bursts, or load-bearing capacity of the rock mass when used in combination with shotcrete. Wire mesh typically used in sheets of 1x2 m with the pattern of 10x10 cm (4 inch) and wire thickness of 6-8 mm, although a tighter pattern and heavier wires can be used for heavy loading. Wire mesh can be used as lagging material between steel sets, installed under the rock bolts, or used as reinforcement for shotcrete support. Shotcrete. Shotcrete is normally applied in the backup area, however, under difficult conditions, shotcrete can be applied immediately behind or in front of the cutter head. Applying shotcrete at or near the heading can be problematic in terms of crew health and safety. In addition, substantial downtimes can be accrued in setting up and removing the shotcrete equipment from the heading. Steel arches (in half or full ring by using rigid (I or H beam rib, Wide flange rib) or yieldable supports (Bell, U or TH sections). Bell sections can be installed under the protection of the finger shield. These sections can be expanded to establish rock contact as the TBM advances to limit rock deformation. The spacing between steel rings should ensure that gripper pads can apply gripper bearing forces on to the rock 137 surfaces. In other words, the interval of the steel sets should be selected based on the practical design layout of gripper pad. Liner plate. Liner plates are curved flanged sheets or steel plates that can be bolted together through the flange and are used in weak rock masses where rock bearing capacity is very low. They are also used as lagging material between the steel sets. Lattice girders are mostly used in conjunction with shotcrete on drill-and-blast and road header-mined tunnels. In these tunnels girders can be installed close to the heading in combination with shotcrete. As noted above, placing shotcrete at the heading is problematic within the confined quarters of the TBM heading. If a need is identified to install a lattice girder-shotcrete lining during the excavation phase of construction, the lining may be installed behind the back-up system. Invert segments are used to provide a reliable rail track bed for material transportation and drainage improvement. It is possible to anchor steel arches to the invert segments through the use of special adaptors. Ground modification and stabilization ahead of the face can be accomplished through the use of probe drilling and injection of cement or resin-based grouts. Open or Gripper type machines can be equipped to enable the mechanically assisted installation of a comprehensive suite of rock support measures behind the cutterhead, in the so- called L1* work area (which is within the TBM and immediately behind the cutterhead, see Fig. 7-3). Support installation equipment can include ring erectors (2), anchor drills (3) and wire-mesh erectors (4). Shotcrete and segments are usually installed in the back-up area. 138 Figure 7- 3. Open TBM and different facilities for support installation in working area behind the cutter head (Herrenknecht web, 2010) Figure 7- 4. Support installation behind the cutter head in Open TBM (Tunneltalk web, 2010; Wallis, 2009; Youtube web, 2010) 139 As can be seen in Fig. 7-4, the first supports for Open TBMs can be installed at a distance ranging roughly 4 to 6 m from the face, i.e. immediately behind the cutterhead shield. This is one of the limits for supporting activities in Open TBMs. In unstable grounds with short stand up time, late support installation can aggravate ground instability and prevent proper functioning of some rock support measures such as rock bolts. Also, in unstable blocky rocks, the support installation takes longer due to the rock collapse over the cutterhead shield and the supporting activities would be elongated due to the need to accommodate additional cleaning and excavation activities at almost every stroke. Tunnel Support Classification Several tunnel support classifications have been developed to relate support requirements to TBM support installation times for a given ground class. In these classifications, the amounts and locations of the various ground supports installed are generally specified. As ground conditions deteriorate, the need to install additional support closer to the cutterhead increases. Under really adverse ground conditions, special ground modification measures such as consolidation grouting in front of the cutterhead may become necessary. Tables 7-1 and 7-2 are examples of rock mass classifications developed on the basis of support requirements in Switzerland and Germany. Table 7- 1.Tunneling classes for TBMs developed in Switzerland (SIA, 1993) 140 Table 7- 2. Tunneling classes for TBMs proposed by Maidl et al. (2008) In these tables, working area 1 is defined as the area from the cutterhead to a distance 15 m behind the face (or the length of the TBM). Working area 2 is delineated as extending from the end of working area 1 to a distance of about 60 m behind the tunnel face (effectively the backup system of the machine). 141 Due to the overwhelming success of the New Austrian Tunnelling Method (NATM) there has been a trend towards an evaluation of the rock mass quality based on the criteria of the Austrian Standard ÖNORM B 2203. Using this classification, the ground can be divided into several classes according to the ground type, amount of temporary support needed for stabilization, and specific excavation steps suggested for improving ground stability. The modified ÖNORM support installation for TBM application (F classes, Fig. 7-5) was developed in Austria from the back analysis of TBM case histories, including a number of tunnels mined by Ilbau, an Austrian contractor with recognized expertise in the operation of open TBMs. This contracting standard provides guidance on rock support requirements and serves as a contract mechanism for regulating the payment of ground support work. In better rock conditions (Classes F1, F2 and F3) bolts, mesh and shotcrete can be installed from the working platform without interrupting TBM excavation operations. In Classes F4, F5 and F6, more extensive supports must be installed in closer proximity to the face, resulting in slow-downs or stoppages to the TBM operation (Atlas Copco, 2005). Class F7 constitutes a ground mass with no self-supporting capacity. In such conditions, ground consolidation techniques, or full lining support with ribs and timber lagging, concrete segmental lining, or bolted liner plates may be required at the heading. In zones of extremely difficult ground, consolidation and support measures ahead of the TBM may be required (Atlas Copco, 2005). The lower F classes (F1-F3) correspond to a "penetration-bound" operation where the TBM can operate at the maximum thrust level in stable rock. As support requirements increase, the operation becomes "support-bound" (Laughton, 2005). F-classes and a description of their corresponding support requirements for a typical tunnel of 5.8-m diameter are shown in Table 7- 3. Table 7-4 shows modified support requirements for different F classes for the Alassio tunnel (3.6 m diameter) in Italy. 142 As can be seen in these tables, an F class is determined for a section of a tunnel based on the rock support requirements and not the rock mass conditions. Limits of support installation in Open TBMs play the determinant role in the assignment of an F class. For example, physical obstruction of the main beam in Open TBM can limit immediate application of rock bolting behind the cutterhead shield. 143 Table 7- 3. Rock classes and their corresponding support measurement chart developed by Ilbua from Austrian Onorm rock classification system to combine NATM with TBM tunneling (Wallis, 1993 and Scolari, 1995) Support Measurement for 5.8 m Diameter TBM Influence Rock Typical Cross Rock Mass Place of the on the Class Section Behavior Type Quantity per m Installation Advance Local support Long term F1 stability - Rock bolt (L=2 m) Up to 0.5 Working platform None Local support Local rock - Rock bolt, L=2 m Up to 1 F2 Working platform None fall -Wire mesh, AQ 50 Up to 1 m2 - Shotcrete, 5 cm Up to 0.1 m3 System support Frequent - Rock bolt, L=2 m From 1 to 3 rock falls in From 1 to 1.5 Short F3 -Wire mesh, AQ 50 Working platform machine m2 hindrance area From 0.1 to 0.5 - Shotcrete, 5 cm m3 - Rock bolt, L=2.5 From 3 to 5 m Frequent -Wire mesh, AQ 50 From 5 to 9 m2 Working platform Hindrance rock falls in F4 From 0.5 to 1 behind the cutter after each machine - Shotcrete, 8 cm m3 head stroke area From 40 to 80 - Steel rib, UNP 120 kg - Rock bolt, L=2.5 From 5 to 7 Immediately Frequent m behind the cutter Long rock falls in -Wire mesh, AQ 50 From 9 to 18 m2 head after each hindrance F5 cutter head From 1 to 1.8 stroke. Additional after each area after - Shotcrete, 10 cm 3 m support from stroke each stroke From 80 to 160 - Steel rib, UNP 120 working platform. kg - Rock bolt, L=3 m From 7 to 10 Immediately Large over behind the cutter Long break in From 18 to 27 -Wire mesh, AQ 50 2 head after each hindrance cutter head m F6 partial stroke. after each area after From 1.8 to 3 - Shotcrete, 15 cm 3 Additional partial partial m support from stroke strokes. From 160 to 300 - Steel rib, UNP 120 working platform. kg No self- Special measurements to be decided according to the F7 supporting conditions. capacity 144 Table 7- 4. F classes and their support requirements for the Alassio tunnel (3.6 m diameter) in Italy based on observation (GEOTEST, 1993) Rock Class Installed Support per m - Rock bolt (2-3 bolts) F2 - Wire mesh (1 m2 in crown area) - Shotcrete (2 m2 in crown area) - Rock bolt (3-4 bolts) F3 - Wire mesh (90° in crown area) - Shotcrete (from spring line upward) - Rock bolt (4-5 bolts) - Wire mesh (all circumference excluding the invert) F4 - Shotcrete (complete cover, 5-10 cm) - Steel arch (from spring line upward) - Rock bolt (5 bolts) - Wire mesh (all circumference including some double sections) F5 - Shotcrete (complete cover, 10 cm) - Steel arch or Liner plate (3/4 to full) - Rock bolt (5-8 bolts) - Wire mesh (double layer) F6 - Shotcrete (complete cover, 15-20 cm) - Steel arch (3/4 to full section) * All rock bolts were of the Swellex type of 1.5-2.1 m length Figure 7- 5. Support installation and Austrian rock mass classes (modified from Martin, 1988) Unit Supporting Time (UST) Unit Supporting Time (UST) is the supporting time per unit length of the bored tunnel (Eq. 7-1). 145 UST=ST/L=(ST/TT)/AR (hr/m) (7-1) where ST is supporting time in hour, TT is total time in hr, L is tunnel length in m, and AR is ST advance rate in m/hr. The ratio of /TT is the supporting time percentage. It is important to note that in some tunnel projects, the value of this parameter is more than four times the value of the BT boring time ratio or “ /TT” (Table 7-5). Fig. 7-6 shows the results of an analysis of drive-averaged supporting time for a database of 9 tunnel projects. As can be seen there is no significant correlation between average UST and average UCS or average Rock Mass Rating, RMR (Bieniawski, 1989). Also these graphs show that on average, UST is less than one. The plot of UST versus machine utilization shows that machine utilization decreases as UST increases. Overall the graphs indicate that the average support time for the whole tunnel drive is not well correlated with the identified rock mass parameters. Further analysis and review of more detailed data for discrete, classified tunnel reaches are necessary to better understand the factors that influence UST. 146 1.20 1.20 1.00 1.00 0.80 0.80 0.60 0.60 UST (hr/m) UST (hr/m) UST 0.40 0.40 0.20 0.20 0.00 0.00 0 2 4 6 8 10 12 14 0 50 100 150 200 Tunnel Diameter (m) Average UCS (MPa) 1.20 70 60 1.00 50 0.80 40 0.60 30 UST (hr/m) UST Support(%) 0.40 20 0.20 10 0.00 0 0 20 40 60 80 100 0 20 40 60 80 Average RMR Utilization (%) Figure 7- 6. Average supporting time/percent for different case histories In order to obtain a range for TBM UST’s in different rock classes, support times for a set of small-diameter tunnel case histories (Table 7-5) were used. This subset of tunnels were mined under variable ground conditions where rock support accounted for up to 50% of the total operating time. In the calculation, the F-class system, which is more suitable for Open TBM tunnels, was used. Fig. 7-7 shows the flowchart of different calculation steps used to obtain the UST range for these tunnels. Using the length of the tunnel for each F class in each tunnel case, it was possible to solve the set of time equations (such as Eq. 7-2) by using an Excel solver add-in. It should be noted that since the equations could not be fully constrained, it was not possible to obtain exact numbers from the set of equations. Hence, a range of UST values was obtained for each F class (Fig. 7-8). (7-2) 147 where is length of tunnel in the F class and is Unit Supporting Time in the F class. Table 7- 5. Tunnel case histories used in calculation the UST range Dia. U Support Time ST Project Name (m) (%) ( /TT) (%) Rock Type Rieti-1 3.9 27.16 22.06 Limestone, conglomerate Castellammare 3.6 20.38 46.53 Limestone Dolomite, Limestone, Camporosso 3.9 25.12 56.05 Siltstone Firenze 3.9 14.52 69.48 Slate, Limestone, Sandstone Alassio 3.6 24 52 Clayed and limed mudstone Val D’arzino (Valley of Arzino) 4.5 25.6 44.4 Siltstone and mudstone Alpe Devero Delivery Tunnel (Devero to Agaro Calcareous schist lake) 3.5 22.4 51.9 Guinza 3.6 19.91 25.32 Sandstone, Marl, Siltstone - Supporting time (hr) - Length of different support categories (F1, F2, ..., F3) 1. Individual supporting time of the projects 2. Positive support category norms 3. Norm of lower support category should be lower than or equal to the next category norm met be to Constraints Setting random values for support category norms Goal: Maximizing the number of cases with minimal difference between evaluated support time and encountered support time Support category norm for rock classes of F1 to F7 Figure 7- 7. Flowchart of calculating UST for Austrian rock mass classes of F1 to F7 using an Excel solver 148 45 40 35 Range of Unit Supporting Time 30 25 20 15 10 Unit Supporting Time (hr/m) Supporting(hr/m) UnitTime 5 0 F1 F2 F3 F4 F5 F6 F7 Austrian Rock Mass Class Figure 7- 8. Open TBM support time for different Austrian rock mass classes To check the results of Fig. 7-8, performance data of an open TBM of the Gossensas tunnel project (Laughton, 2005) was analyzed as shown in Table 7-6. In this table UBT, UAT, UDT, and UST are Unit Boring Time, Unit Advance Time, Unit Down Time, and Unit Support Time, respectively, as introduced by Laughton (1998) (except for UST). In calculating the UST, it is assumed that the only difference between downtimes of class F1 or F2 and the other classes is support time, which is considered acceptable with some tolerance for the controlling task. The UDT for the classes F1 and F2 of the Gossensas tunnel is in the range of the "best-practice" case histories of open TBMs (Laughton, 1998), which is noted as being between 0.3 and 0.4 hr/m (Eq. 7-3 to 7-6 and Fig. 7-9). This downtime is mostly due to other activities in TBM tunneling rather than supporting. (hr/m) (7-3) (hr/m) (7-4) 149 (hr/m) (7-5) (hr/m) i=2, 3, ..., 7 (7-6) where PR is penetration rate, AR is advance rate, is the unit downtime of the F1 class (minimum support installation), and is the unit downtime of Fi class. 1.2 1.0 0.8 Data of the Best Practice Cases of Open TBM UDT=0.3-0.4 hr/m 0.6 UBT (hr/m) UBT 0.4 0.2 UAT=UBT 0.0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 UAT (hr/m) Figure 7- 9. Best-practice case histories of Open TBM (Laughton, 1998) Table 7- 6. UST for the Gossensas tunnel Rock Class PR (m/hr) U (%) AR (m/hr) UBT (hr/m) UAT (hr/m) UDT (hr/m) UST (hr/m) F1 3.1 47 1.5 0.32 0.7 0.36 0 F2 3.6 44 1.6 0.28 0.6 0.35 0 F3 3.9 35 1.3 0.26 0.8 0.5 0.1 F4 4.1 20 0.8 0.25 1.3 1.0 0.7 F5 4.2 11 0.5 0.24 2.2 1.9 1.6 F6 4.3 6 0.3 0.23 3.9 3.6 3.3 F7 4.6 2 0.1 0.22 11 10.8 10.4 The UST values for the Gossensas tunnel are plotted over the previously obtained range of UST values in Fig. 7-10. As can be seen, the UST values of the classes F1 to F6 fit very well in the corresponding ranges whilst the UST value of the class F7 is far away from its corresponding range. This might be due to the fact that the F7 class is very sensitive to the ground quality, support system used, and/or the ground behavior. The offered range of the flowchart calculations for the F7 class is part of a set of TBM tunneling in worst case scenarios. The difference could also be due to the fact that the operation in the Gossensas tunnel did not require the use of some 150 of the more extreme measures and thus did not encounter the related delays of more complex ground support systems that may be used elsewhere to control the ground. Fig. 7-10 also shows the approximate UST for two other TBM projects in Italy with a diameter of 3.5 m (Table 7-8). The F classes were not detailed in the source documents and were assessed by reference to the support requirements of each tunnel section. See information in Tables 7-3 to 7-8. As can be seen in Fig. 7-10, similar trends can be observed for these tunnels, and except for F7 class, the UST of other classes match very well with the Fig. 7-8 range of UST. 151 45 40 Range of Unit Supporting Time 35 Gossensas Tunnel 30 25 20 15 10 Unit Supporting Time (hr/m) Supporting(hr/m) UnitTime 5 0 F1 F2 F3 F4 F5 F6 F7 Austrian Rock Mass Class 45 40 35 Range of Unit Supporting Time 30 Alpe Devero Delivery Tunnel (Devero to Agaro lake) 25 20 15 10 Unit Supporting Time (hr/m) Supporting(hr/m) UnitTime 5 0 F1 F2 F3 F4 F5 F6 F7 Austrian Rock Mass Class 45 40 35 Range of Unit Supporting Time 30 Alpe Devero Delivery Tunnel (Bodolero to Cairasca) 25 20 15 10 Unit Supporting Time (hr/m) Supporting(hr/m) UnitTime 5 0 F1 F2 F3 F4 F5 F6 F7 Austrian Rock Mass Class Figure 7- 10. Comparison of UST values of some tunnel projects with the obtained range of the UST 152 Table 7- 7. UST and F classes for the Alpe Devero delivery tunnel (Devero to Agaro lake) Rock Wire Steel Shotcrete UST F- Section Main Rock Bolt Mesh Arch (m3) (hr/m) Class (No.) (m2) (No.) S1 Calcareous Mica Schist 2.7 3.8 0 0 0.07 F3 S2 Marble-Calcareous Mica Schist 10.85 11.3 2.64 0 26.5 F7 S3 Metaconglomerate 2.95 4.6 0.07 0 0 F3 S4 Calcareous Biotite Schist 3.18 5.2 0 0 0.55 F4 S5 Calcareous Schist/Talk and Cholorite 7.25 11.64 1.33 0.96 2.45 F6 S6 Calcareous Schist/Quartz 3.1 5.05 0 0 0.01 F3 S7 Calcareous Mica Gneiss 2.95 5.5 0 0 0.07 F3 Table 7- 8. UST and F classes for the Alpe Devero delivery tunnel (Bodolero to Cairasca) ROP Swellex Wire Mesh Shotcrete Steel Arch UST F- Section Main Rock (m/hr) (No.) (m2) (m3) (No.) (hr/m) Class S1 Calcareous Mica Schist 3.1 0 4.75 0 0 0.01 F3 S2 Metaconglomerate 3.05 0 4.9 0 0 0.09 F3 S3 Calcareous Schist 3.4 1.95 6.75 0.73 0 1.11 F4 S4 CalcareousMica Schist 3.25 1.85 8.45 1.2 1.07 1.67 F5 S5 Marble 2.86 0 4.5 3.5 0 0.1 F3 S6 Calcareous Biotite Schist 6.65 6 21.75 0 3.48 4.69 F6 Support Installation Time (SIT) Another approach for the estimation of UST can be adopted to find an approximate range for the average required time for installation of different support types. In this approach it is not possible to consider the effect of simultaneous activities undertaken during the TBM excavation process. The flowchart in Fig. 7-11 shows different steps of finding the general range of SIT. 153 - Supporting time (hr) - Number of different supports (bolts, mesh, steel set, ...) 1. Individual supporting time of the projects 2. Positive support category norms Constraintsbe tomet Setting random values for support category norms Goal: Maximizing the number of cases with minimal difference between evaluated support time and encountered support time Support Installation Time for different support types Figure 7- 11. Flowchart of calculating the SIT for different support types using an Excel solver For this approach, the monthly data of four open TBMs in Italy were used. The TBM activity time distributions of these tunnels are shown in Fig. 7-12. Fig. 7-13 shows the approximate range of SIT for different support types used in these tunnels. As can be seen, the required installation time for Mesh, Liner Plate, and shotcrete is greater than for other support systems. It should be noted that in these tunnel cases, shotcrete was applied near the tunnel face. This kind of graph can be used to find UST, especially for the tunnel reaches with adverse geological conditions such as sections in F6 and F7 classes for which it is not possible to define a narrow UST range. 154 100 100 Rieti-1 Tunnel 90 Firenze Tunnel 90 Dia.=3.9 m Dia.=3.9 m 80 80 70 70 Bore+Regrip Bore+Regrip 60 Muck 60 Muck 50 Maintenance 50 Maintenance Percent (%) Percent Percent(%) Waiting time Waiting time 40 40 Machine Failure Machine Failure 30 30 Other Other 20 Support 20 Support 10 10 0 0 Jul Jul Jan Jun Oct Oct Sep Feb Sep Apr Apr Dec Mar Mar Nov Aug Aug Nov May May Month Month Average Average 100 100 Camporosso Guinza Tunnel 90 90 Tunnel Dia.=3.63 m 80 Dia.=3.9 m 80 70 70 Bore+Regrip Bore+Regrip Muck 60 Muck 60 Maintenance 50 Maintenance 50 Percent (%) Percent Percent(%) Waiting time Waiting time 40 40 Gas Machine Failure 30 30 Machine Failure Other Other 20 Support 20 Support 10 10 0 0 Jul Jul Jan Jan Jun Jun Oct Oct Sep Feb Sep Feb Apr Apr Apr Dec Dec Mar Mar Mar Aug Nov Aug Nov May May Month Month Average Average Figure 7- 12. TBM activity time distribution for four Open TBMs 45 40 35 Range of Support Installation Time 30 25 20 15 SIT (min/piece m3) m2orSIT or 10 5 0 Bolt Mesh Liner Plate Steel Set Shotcrete Support Type Figure 7- 13. Open TBM support installation time for different support types 155 In order to check the results of Fig. 7-14, support installation data of an open TBM used in a tunneling project in New York City was analyzed as shown in Fig. 7-15. 100% 100% 90% 90% 80% 80% 70% 70% 60% 60% 50% 50% 40% 40% Percent (%) Percent Percent (%) Percent 30% 30% 20% 20% 10% 10% 0% 0% 0-1 1-2 2-3 3-4 4-5 5-6 00-05 05-10 10-15 15-20 20-25 Time of Installation per Bolt (min.) Time of Installation per Steel Channel (min.) Figure 7- 14. SIT for tunnel project in New York City (diameter 3.65 m completed in 2005) The SIT values are plotted over the previously obtained range of SIT’s shown in Fig. 7- 15. As can be seen, the SIT values of the NYC tunnel fit comfortably within the middle of the corresponding SIT ranges. 45 40 35 Range of Support Installation Time NYC Tunnel 30 25 20 15 SIT (min/piece m3) m2orSIT or 10 5 0 Bolt Mesh Liner Plate Steel Set Shotcrete Support Type Figure 7- 15. Comparison of NYC tunnel SIT values and the range of the SIT 156 The approximate nature of SIT range can imply that it depends on other factors such as the design of rock support, installation procedures, and improvements applied on a specific type of support. These factors directly affect the SIT value and therefore the UST value. Also, innovations in rock support types and installation can lower the upper limits of both SIT and UST. This includes the “McNally System” that has recently been introduced by the Robbins Company for application on open types TBMs when encountering broken/ blocky ground. (Home 2010, and Robbins website). Comparison between O-Norm and Other Rock Mass Classifications Table 7-9 shows a comparison between F classes and some other well-established rock mass classifications for a small tunnel diameter (3.9 m). Table 7- 9. Comparison of different rock mass classification methods with F-class (Atlas Copco, 2005) RMR Class Bieniawski 1973 RMR Deere 1969 RQD Barton 1974 Q Class O-Norm (3.9 m dia. TBM) I 83 90 33 F1-F2 II 67 75-90 12.5 F3 III 52 50-90 8.5 F4 IV 29 25-50 1.5 F5 V 15 <25 0.09 F6 As Laughton (2005) noted, if reaches of a tunnel can be divided into sections that are classified using O-Norm or “F class" based on site investigation, this system can be used to develop design and construction scopes that can address both the rock support and performance expectation of a TBM. Unfortunately there is not much information regarding correlation between F-class system and other rock mass classifications with the exception of the guideline shown in Table 7-9. In order to compare the F classes with other rock mass classifications, 157 ground support data for the Manapouri tunnel (Reach No. 1) was used to study UST values in different rock mass and fracture classes (Fig. 7-16). 45 60 40 50 35 30 40 25 30 20 UST (hr/m) UST (hr/m) UST 15 20 10 10 5 0 0 20 30 40 50 60 70 80 90 100 60.0 65.0 70.0 75.0 80.0 85.0 90.0 95.0 100.0 RMR RQD 60 60 50 50 40 40 30 30 UST (hr/m) UST UST (hr/m) UST 20 20 10 10 0 0 0 500 1000 1500 2000 2500 3000 3500 0 0-I I I- II III Joint Spacing (mm) NTH Fracture Class Figure 7- 16. Correlation between UST and some rock mass classifications for Manapouri tunnel Each data point in these graphs represents the average value of UST for a specific rock mass class recorded in the database. This database contains 483 shift-based data records of 1668 m of reach No. 1 of the tunnel. It should be noted that the semi-log graph of Q has the same trend as for the RMR graph, hence this graph is not shown in Fig. 7-16. For RQD, the number of data points and spread are very limited. Spacing and NTH fracture class graphs show better trend and distribution attributes. Among the graphs of Fig. 7-16, the RMR graph is the best graph in terms of having more data points and better spread. Part of the reason for this observation is that the RMR classification is a combination of several parameters, and that the spread in RMR values for these 483 records is more than the other rock mass classification values provided. 158 One important factor in UST determination is tunnel diameter. In the RMR system, support requirements were suggested on the basis of rock mass conditions (Bieniawski, 1989) and are not directly linked to the size of the tunnel. This issue is crucial for large tunnels where the span of the opening dictates low stand-up time, for which the supporting delays might increase tremendously. One solution to this problem is using Q-system guidelines for support selection in which both rock condition and tunnel diameter and considered (Grimstad and Barton, 1993). The following analysis is not intended to address this issue and this needs close examination of comparative support requirements data for the same conditions in a variety of tunnel diameters. If it is assumed that support needs are proportional to tunnel diameter for a given RMR value, the UST would also be proportional to the tunnel diameter. To compare the UST values of different rock classes for the Manapouri tunnel, RMR classes were converted to equivalent UST values for a tunnel of 3.9-m diameter (UST (3.9), Eq. 7-7). (7-7) Considering the proposed equivalence between RMR classes and F classes (Table 7-9), it is possible to compare the UST (3.9) with the corresponding UST range of Fig. 7-17. The results of this comparison for the Manapouri tunnel show that the UST (3.9) values of the tunnel do not correspond well to the UST values of F classes. Part of the reason for the difference could be related to the fact that the real equivalent RMR values of the different F classes are not exactly the same as the ones offered by Atlas Copco (Table 7-9). Table 7-10 represent the equivalent RMR values for different F classes based on the general trend of the points in Fig. 7-17. 159 18 F1-F2-F3 F4 F4-F5 F6-F7 16 14 12 10 8 6 UST (3.9) (hr/m)UST 4 2 0 100 90 80 70 60 50 40 30 20 RMR Figure 7- 17. Equivalent RMR values for different F classes based on the general trend of the points Table 7- 10. Equivalent RMR values for different F classes based on Manapouri UST (3.9) RMR UST UST (3.9) O-Norm 70-100 0.4 0.15 F1-F2-F3 55-70 1.2 0.46 F4 40-55 4 1.59 F4-F5 <40 37 14.35 F6-F7 Fig. 7-17 indicates that in tunnel sections where RMR is greater than 70 (F1-F3), the opening is in a stable condition and support requirements are minimal. For tunnel sections where RMR is less than 70, the TBM performance, especially utilization, will be considerably affected by support requirements. This analysis shows the potential to estimate ground support installation-related downtime in a TBM-driven tunnel based on the use of rock mass classification systems. In other words, once the RMR or Q values for tunnel sections are determined, Unit Support Time (UST) can be determined and combined with a general estimate of machine utilization to calculate 160 utilization and daily advance rates. Obviously, the results of this analysis are based on a limited number of cases. This small data set is inadequate to support a generalization of the conclusions of this study, but it clearly shows the potential for further investigation and development. The formulas and graphs for estimating UST require additional study, but the methodology provides an analytical framework for future studies in this field. One of the most important applications of such an analysis is to increase precision in the planning of a tunnel project using a hard-rock TBM. It also allows for risk mitigation plans relative to ground support selection and installation and provides a meaningful set of quantifiable measures for the analysis of mining delays. These measures can be used to develop ground-compatible machine specifications and support designs. Furthermore, the proposed methodologies can help contractors prevent long delays and operational stoppages during construction by assisting in the timely implementation of ground support changes. The use of the proposed TBM ground support installation analysis will allow for higher construction efficiencies and improved safety. Summary and Discussions Analysis of the TBM downtime related to different ground conditions and different support types were performed in this study. Unit Supporting Time (UST) and Support Installation Time (SIT) were used in the analysis to evaluate the impact of rock mass and ground conditions on downtime and machine utilization. The validity of these concepts was examined by using TBM case history data sets. The conclusions of this study can be summarized as follows: Utilizations of Open TBMs in unstable ground are very low due to the significant impact of ground support installation time and related and delays. Results of UST analysis for case histories using small open-type TBMs show that if the Austrian rock mass classification is used, F1 and F2 rock classes rarely need any 161 ground support. In these ground classes, support installation does not hinder the tunneling process. The amount of required support will increase with Classes F3-F6 and related downtime can be estimated. However, support installation time in stretches of F7 classified rock might need more than 40 hr/m supporting time and is difficult to estimate since it involves extreme and unpredictable measures including the use of ground improvement methods. Analysis of SIT shows that when shotcrete is applied near the tunnel face, it can severely hinder tunneling and can take 40 hr/m on average. This is unless shotcrete is applied in larger-diameter tunnels with special provisions for its applications in the L1 area. Obviously, this explains why, in application of smaller TBMs, installation of shotcrete immediately behind the head is not typically required. The results show that rock bolting is the fastest ground support method in TBM tunneling. Use of rock bolts in F1-F3 Austrian rock classes coincides with their low impact on machine utilization. The comparison of Austrian (Open TBM) and RMR classifications in the cases studied shows that F1-F3 rock classes have an RMR range of 70-100, and generally do not impact TBM operation. A model is offered to allow for comparison of the F1-F7 rock classes and RMR range and for estimation of the related UST. This model which was expressed in a table and a corresponding graph can be used to adjust estimated machine utilization and the related daily advance rate for zones of different rock mass and ground support classes. Use of the proposed system can allow for the fine tuning of utilization and advance rate estimates and corresponding construction plans. It can also help to minimize the impact of ground support-related delays and improve operational safety. 162 Chapter 8 Cutter Change Time and Cutter Consumption Introduction Hard-rock tunnel boring machines are usually equipped with disc cutters to cut the rock at the tunnel face. Early disc cutters were 10 in diameter (Maidl et al., 2008) and were used efficiently for low-strength rocks such as sedimentary formations. Subsequently, larger cutters were introduced to increase the cutter load capacity and to increase TBM penetration and advance rate. The larger cutters could be used in harder rock formations due to higher applicable loads and higher machine thrust and power. To decrease machine downtime, machine manufactures steadily improved cutter life by increasing hardness and toughness of the cutter ring. As a result, disc cutters have experienced significant advances in material and manufacturing technology, especially in recent years. Table 8-1 represents the historical developments of cutters and their bearing load capacity. Table 8- 1. Cutter diameter and bearing load capacity (Roby et al., 2008) Diameter (inch) Bearing Load (kN) Year Introduced 11 85 1961 12 125 1969 13 145 1980 14 165 1976 15.5 200 1973 16.25 200 1987 17 215 1983 19 312 1989 20 312 2006 163 Fig. 8- 1 shows a typical TBM cutterhead equipped with 17-in cutters as well as the arrangement of the cutters on the cutter head in its cross section. The rotation of the cutterhead causes the cutters to roll in concentric tracks on the tunnel face. The common explanation for the cutting of the rock under the cutter is that the high pressure in a crushed zone (pressure bubble) created under the cutter causes tensile stresses within the rock which leads to radial cracks and finally, the formation of rock chips (Ozdemir et al., 1978; Buchi, 1984; Rostami, 1993) (Fig. 8- 2). 210 27-29 Gauge 26 24 25 24 17 17 23 50 22 50 21 50 65 20 Transition 68 19 50° 79 18 80 17 76 16 15 14 13 12 11 10 9 2262 R Flat Face Flat 8 7 R 1842 R 6 5 4 3 2 1 CC6 CC5 CC4 R 463 R CC3 Center Figure 8- 1. Typical disc cutters and their arrangement on the cutter head 164 Figure 8- 2. Rock breaking process under a disc cutter (Buchi, 1984) In practice, cutters are inspected every shift or every day (depending on the rock type at the face) and are changed after a certain tip wear is reached. For changing the cutter, the cutterhead stops and is pulled back and the cutters that have experienced wear in excess of a pre- set amount of tip loss need to be replaced. As a result, the operation incurs a specific amount of downtime to inspect, repair, and replace the disc cutters. Cutter-related downtime is typically not the dominant source of downtime in TBM applications (typically five percent of the total time); however, when the rock is highly abrasive or when a catastrophic cutter failure happens due to a domino effect (for example the "wipeout" phenomenon as described by Roby et al. (2008)), the resulting downtime can be very high due to a high rate of cutter change and repairs. There are three main sources of down times which are related to cutter change and repair as explained by Roby et al. (2009). Cutter inspection; Cutter change: This is related to inspection and replacement of worn-out cutters or damaged bearing/seal assemblies, as well as moving cutters for even ring wear; 165 Cutter head repairs: This is related to either extreme geological conditions (abrasive rock in fault zones or blocky rock) or the failure of a series of cutters which can cause repair of cutterhead, cutters, and/or buckets. This chapter is focused on the evaluation of the average cutter change time based on the study of the data from several tunnel projects. The data was obtained from several projects, related literature and published papers, and contractors’ project documents. Cutter Change Time Cutter change time is defined as the average time required for inspection and changing of a cutter on the cutter head. This item has not been studied in detail in recent years. The NTH model is one of the models that offers some rough estimates for cutter change time for two different cutter size categories as part of the estimation of TBM utilization (refer to Table 6-4). 166 Fig. 8- 3 shows cutter change time of 36 tunnel projects compared with the estimated cutter change time per km of tunnel by the NTH model. 600 Reported NTH Model 500 400 300 Tc (hr/km) Tc 200 100 0 0 10 20 30 40 Tunnel Drive No. Figure 8- 3. Cutter change time for 36 hard-rock TBM tunnel projects As can be seen in Fig. 8- 3, for most of the cases, the NTH model underestimates the cutter change time values. 600 600 y = 1.0926x 600 NTH R² = 0.8012 model 500 with intercept 500 without intercept 500 s=60.08 s=59.87 400 400 400 300 300 300 Tc (hr/km) Tc (hr/km) Tc 200 (hr/km) Tc 200 200 100 100 100 0 0 0 0 200 400 600 0 200 400 600 0 200 400 600 Cutter Consumption (cutter/km) Cutter Consumption (cutter/km) Cutter Consumption (cutter/km) Figure 8- 4. Correlation between cutter consumption and cutter change time Fig. 8- 4 presents the results of analysis of the correlation between cutter consumption and cutter change time for the same tunnel projects. The two graphs on the left-hand side show 167 the linear regression models with and without an intercept. The slope of the line with the without- intercept model is the average of cutter change time (hr/cutter). To choose the most accurate model among the models with and without an intercept, "s" which is defined from the Mean Square Error (Eq. 8-1 and 8-2) calculation was used. It should be noted that R2 is not the best indicator for this analysis and it is unsuitable to use to compare models. For with-intercept model (8-1) For without-intercept model (8-2) where MSE is Mean Square Error and n is the number of data points. As can be seen in Fig. 8- 4, the s value of the best-fit model without-intercept is less than that of the model with-intercept. So the former model presents better results. The slope of this best-fit model is 1.09 hr/cutter or 66 min/cutter. This is the average value for cutter change time, which includes inspection. The right-hand side graph in Fig. 8- 4 shows the comparison between the results of the regression analysis and the general trend of the NTH model for the studied tunnel projects. As can be seen, the NTH trend line is below the obtained regression line since in the NTH model, for cutter diameter ≤17 in and >17 in, the cutter change time of 0.75 hr (45 min.) and 0.833 hr (50 min.) were proposed, respectively. Larger cutters are heavier, and the process of handling is more difficult and time consuming. Therefore, it is common sense to expect higher time consumption to change larger cutters. In order to examine the effect of cutter size on cutter change time, the data set was split into three major sub-categories for 350-394 mm (14-15.5 in), 413-444 mm (16.25-17.5 in), and 483-508 mm (19-20 in) cutters. The results of linear regression analysis for these three subcategories are shown in Fig. 8-5. The lower graphs of Fig. 8-5 represent the proposed models with the best-fit line going through the origin (without intercept), while the upper graphs show the 168 general best-fit (with-intercept) linear models. The slope values of the without-intercept models represent the average cutter change time values for three cutter categories. 500 Cutter Dia.=14-15.5" 600 Cutter Dia.=16.25-17.5" 300 Cutter Dia.=19-20" R² = 0.737 R² = 0.8845 R² = 0.6062 400 s=82 500 s=43 250 s=46 400 200 300 300 150 200 Tc (hr/km) Tc Tc (hr/km) Tc Tc (hr/km) Tc 200 100 100 100 50 0 0 0 0 200 400 0 200 400 600 0 100 200 300 Cutter Consumption (cutter/km) Cutter Consumption (cutter/km) Cutter Consumption (cutter/km) 500 600 300 y = 1.1232x y = 1.1062x y = 0.9791x s= 80 s = 42 250 s=49 400 500 400 200 300 300 150 Tc (hr/km) Tc 200 (hr/km) Tc Tc (hr/km) Tc 200 100 100 100 50 0 0 0 0 200 400 0 200 400 600 0 100 200 300 Cutter Consumption (cutter/km) Cutter Consumption (cutter/km) Cutter Consumption (cutter/km) Figure 8- 5. Correlation between cutter consumption and cutter change time for different cutter size categories The results seem quite surprising, implying that larger cutters are quicker to change than the smaller ones. However, the differences among the slopes of these categories are not so high. The slope of two categories of 350-394 mm (14-15.5 in), 413-444 mm (16.25-17.5 in), and 483- 508 mm (19-20 in) cutters are quite similar. Furthermore, the slope value of the third category (19-20 in cutters) is not as strong as the ones for the first categories, since the R2 value of the third category is much lower; this could be within the margin of error. Perhaps adding more information to the database can improve the results to get a better distinction between these categories. To examine the effect of cutter size and tunnel diameter on cutter change time, the scatter plots of cutter change times for different cutter sizes and tunnel diameters are shown in Fig. 8- 6. 169 The correlation between the cutter change time and cutter size/tunnel diameter is very low. It seems smaller cutter sizes, which are usually used in smaller tunnels, are more scattered around the average cutter change time value, and they have higher tc values, but the left-hand-side graph clearly shows that the number of data points for larger cutters is much less than the number of smaller cutters. If the tc values of more than two (as shown in Fig. 8- 7) are ignored, one can observe that there is no significant difference between the cutter change times of different cutter sizes/tunnel diameters' ranges. 3 3 2.5 2.5 2 2 1.5 1.5 tc (hr/cutter)tc tc (hr/cutter)tc 1 1 0.5 0.5 0 0 12 14 16 18 20 22 2 3 4 5 6 7 8 9 Disc Cutter Diameter (inch) Tunnel Diameter (m) Figure 8- 6. Scatter of the data points of tc for different cutter sizes and tunnel diameter sizes (solid line is the average slope parameter of the regression analysis) 3 3 2.5 2.5 2 2 1.5 1.5 tc (hr/cutter)tc tc (hr/cutter)tc 1 1 0.5 0.5 0 0 12 14 16 18 20 22 2 3 4 5 6 7 8 9 Disc Cutter Diameter (inch) Tunnel Diameter (im) Figure 8- 7. Extreme cutter change time values in the data set 170 Number of Changed Cutters and Estimation of Cutter Change Time This section will discuss the details of cutter change time from shift reports of some recent tunnel projects. It should be noted that generally the cutter change time in reported time components for the whole length of a tunnel includes cutter inspection. However, in the shift based reports, the inspection is usually not considered as a part of cutter change time. Therefore, the time component representing cutter change time based on the shift report is lower than the actual. Cutter inspection is a routine shift activity that is usually performed at the beginning and less frequently at the end of a shift. Fig. 8- 8 presents the average cutter change time for different cases, as a function of the number of cutters changed at one stop in six tunnel projects (Table 8-2). As can be seen, as the number of changed cutters within one stop increases, the cutter change time becomes smaller since several operations can be combined with each other. Furthermore, cutter size does not show a clear trend with cutter change time. For example, 17 in cutters can be changed in a wide range from 48 (Manhattan) to 102 minutes (Ghomroud) per cutter. This difference might be related to the effect of combining TBM maintenance and cutter change in some tunnel projects, such as the Ghomroud tunnel, which could be considered as a reporting or categorization issue. Another important issue that is not addressed here is the experience of the contractor and also the frequency of cutter changes. This refers to the faster cutter change in stronger and more abrasive rocks since the process is streamlined and cutter changes are frequent, as compared to projects where a few cutters are changed every so often and thus due to less frequency, there is no urgency to optimize this activity. 171 Table 8- 2. General specifications of tunnel projects used for analysis of cutter change time and change cutter in the group. Excavated Diameter Cutters Dia. Name of the Project Rock Type (m) (inch) Intrenchment Creek 7.95 15.5 Granitic Gneiss Klippen Hydropower- Schist, Granitic Gneiss, Mica 6.5 20 Headrace Gneiss Manhattan 3.84 17 Mica Schist, Schistose Gneiss Ghomroud 4.5 17 Schist, Slate, Limestone Klippen Hydropower- 6.5 20 Tailrace Tunnel a 3.5 17 a: A tunnel excavated with a TBM type of Jarva MK-12 (Büchi, 1993) 160 Intrenchment Creek-15.5" Cutter a-17" Cutter 140 Manhattan-17" Cutter 120 Ghomroud-17" Cutter Klippen Hydropower- Headrace-20" Cutter 100 Klippen Hydropower- Tailrace-20" Cutter 80 60 40 Cutter Change Time (min.) Change Time Cutter 20 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Number of Cutter Changed at One Stop Figure 8- 8. Cutter change time for different number of cutters changed at one stop Table 8-3 shows the detailed cutter change analysis for a part of water tunnel # 3 in Manhattan. As can be seen in this table, there are two different factors affecting the cutter change time. The effect of changing the cutter concurrent with other activities caused nearly 20% decreases in average cutter change downtime for total changed cutters. On the other hand, the effect of inspection time caused a tremendous increase of nearly 50% from 36 to 62 minutes, if it is done separate from scheduled maintenance activities. These factors sometimes cannot be distinguished clearly, and yet can affect the estimated average cutter change time. 172 Table 8- 3. Cutter change time analysis results for Manhattan south tunnel Number of Cutters Cutter Change Cutter Change Time with Percent Changed Downtime (min.) Inspection (min.) Total 249 100 36 62 1. Cutter change performed as a 195 78 46 - separate activity* 2. Cutter change performed along 54 22 0 - with other activities** *boring process is stopped and cutter change is performed **e.g. cutter change is performed during the general maintenance Average cutter change time depends on inspection time and the percentage of cutters which are changed during other activities. Schubert (2010) noted one minute per 100 m length of a tunnel for the cutter inspection. It is hard to define any average value for the inspection time since it depends on the planned/scheduled activities of a tunnel project, and it varies from one project to another. Based on the data analysis of the Manhattan tunnel, cutter inspection comprises around 40% of the cutter change time. This means that sometimes estimation of cutter inspections time is as important as cutter change time. Most contractors include the inspection time in average cutter change time in their time distributions. Cutter Change Time and Penetration Rate In this section, an attempt is made to estimate the cutter change time without consideration of the rock mass parameters, and based on average Penetration Rate (PR). Fig. 8- 9 shows the cutter change time analysis for different quartz contents with the same data set as previously discussed in Figures 8-4 and 8-5. 173 500 450 Quartz Content ≈ 30-50% 400 350 300 Tc (hr/km) Tc 250 200 Quartz Content ≈ 10-20% R² = 0.6929 150 100 R² = 0.5564 50 Quartz Content < 5% 0 0 1 2 3 4 5 6 PR (m/hr) Figure 8- 9. Cutter change time The charts of Fig. 8- 9 show that as PR decreases the cutter change time increases. This can be related to the fact that lower values of PR are related to the harder rocks or larger machine sizes. Cutter Life While some contend that cutter wear is mainly caused by fine materials produced during the fragmentation process (WBI, 2006), there are other theories and potential explanations for this process. The amount of the cutter wear depends on several geotechnical and machine parameters. 174 One of the common indexes that is used to predict the wear of the cutter and consequently, cutter life (Eq. 8-3) is Cerchar Abrasiveness Index (CAI). The coefficient of wear, Cp, which was offered by WBI (2006), is another index for classification of rock wear (Eq. 8-4). (8-3) (8-4) where CL is cutter life in m3/cutter, CAI is Cerchar Abrasivity Index, d is cutter diameter in inch, PR is TBM penetration rate in m/hr, D is tunnel diameter in m, RPM is revolution per minute, N is number of cutters on the cutterhead, and p is cutter penetration in mm/rev. Fig. 8- 10 shows the mean CAI values and their corresponding standard deviations obtained from more than 200 tests on 10 different rock types (Buchi, 1984). (1) with carbonate or clay cementation (2) with siliceous cementation Figure 8- 10. CAI values with average and standard deviation for different rock types (Buchi, 1984) The number of required cutter changes for a TBM drive is very important for defining the anticipated cutter change time and related downtime as well as the cost of cutters for the whole 175 project. Using TBM penetration rate (PR), TBM revolution per minute (RPM), and tunnel diameter, it is possible to convert the cutter life to cutter/m3. Fig. 8- 11 shows the results of cutter life calculation (in terms of excavated volume per cutter) based on the CAI calculation and the real cutter life. This graph indicates that Eq. 8-3 is unsuitable for prediction of the cutter consumption for these tunnel cases. 600 500 400 300 200 Predictedlifecutter (CAI) 100 0 0 100 200 300 400 500 600 Actual cutter life Figure 8- 11. Graph of real cutter life versus predicted cutter lifer (using Eq. 8-3) in m3/cutter 600 600 500 500 400 400 300 300 Cutter life (m3/cutter) life Cutter Cutter life (m3/cutter)lifeCutter 200 200 100 100 0 0 0 1 2 3 4 5 0 1 2 3 CAI Cp Figure 8- 12. Relationship between CAI/Cp with real excavated volume per 176 One important note for CAI testing is that the results of testing might vary significantly even for the same rock and as Rostami et al. (2005) noted, there is a need to modify the testing procedure to obtain consistent and repeatable test results to reduce the effects of variations of testing procedures, equipments, and operator sensitivity. In order to examine the effect of CAI on cutter life, the reported real cutter life data points are plotted against CAI and Cp (Fig. 8- 12). Even though the data points look scattered, the graphs show that there are some trends, especially for the case of CAI. As these results cannot be generalized for the whole tunnel cases, due to the limited tunnel cases used in the analysis, they do point to the effect of CAI on the cutter life. Another method for cutter life prediction is the NTH method. In this method cutter ring life (Hh) is determined by using equations 5-7 and the graphs of Fig. 8- 13. (5) (6) (7) where Hh= average life of cutter rings H0= basic cutter ring life kD= correction factor for TBM diameter kQ= correction factor for rock quartz content kRPM= correction factor for TBM RPM kN= correction factor for the number of cutters Ntbm= number of cutters on the cutterhead dtbm= TBM diameter N0= normal number of cutters on the TBM 177 2.25 250 2 225 Flat Cutterhead Dc=483 1.75 200 KD 1.5 175 Dc=432 Domed Cutterhead 1.25 150 1 125 Dc=394 0.75 H0 (hr) H0 100 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 Dc=356 Diameter (m) 75 50 2 25 1.8 0 0 10 20 30 40 50 60 70 80 90 100 110 120 130 1.6 CLI (Cutter Life Index) 1.4 70 KQ 1.2 60 Dc=393 1 50 Mica Schist N0 0.8 Mica Gneiss 40 Dc=500 Gneiss Granitic Gneiss 0.6 Granite 30 0.4 0 10 20 30 40 50 60 70 80 90 100 20 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 Quartz Content (%) Diameter (m) Figure 8- 13. Graphs of NTH method used in cutter consumption (Redrawn from Bruland, 1998a and 1998b) Fig. 8- 14 shows the results of estimating cutter consumption based on the NTH method and the real cutter consumption. 1000 800 600 400 Predicted cutter life (NTH)Predicted life cutter 200 0 0 200 400 600 800 1000 Actual cutter life Figure 8- 14. Graph of actual cutter life versus predicted cutter life (using Eq. 8-3) in m3/cutter 178 The graphs in Fig. 8- 15 show the correlation between some rock mass parameters (Tables 8-4 and 8-5) and cutter consumption for 135 tunnel cases. The charts of Fig. 8- 16 show the general trend of cutter consumption for different rock types and quartz (Qtz) contents and UCS values. These charts were obtained from data screening for the 135 tunnels cutter consumption database for different rock types and quartz contents. It should be mentioned that multiple linear regression analysis was also performed on this database, and it yielded a very good coefficient of determination, but in order not to use codes in the final formula, the database screening was applied to generate the most accurate curves and trends to be used as a preliminary tool for prediction of cutter consumption. Table 8- 4. Quartz content categorization in the database Quartz content in percentage Code Description 0-20 1 Low 20-50 2 Significant 50-75 3 High >75 4 Very High Table 8- 5. Rock type categorization in the database Rock Type Code Claystone 1 Limestone 2 Volcanic rocks 3 Sandstone, quartzite 4 Metamorphic rocks 5 Igneous rocks 6 179 3000 2000 R² = 0.4795 1000 Cutter life (m3/cutter) life Cutter 0 0 1 2 3 4 5 6 Rock Type Code 3000 2000 R² = 0.4321 1000 Cutter life (m3/cutter) life Cutter 0 0 1 2 3 4 Quartz Content Code 3000 2000 R² = 0.1474 1000 Cutter life (m3/cutter) life Cutter 0 0 100 200 300 UCS (MPa) Figure 8- 15. Rock type, quartz content, and UCS versus cutter consumption 180 Limestone and Dolomite Sedimentary Rocks except Limestone/Dolomite 3000 3000 ~ 5% Qtz 2500 2500 <5% Qtz 2000 2000 ~ 10% Qtz 1500 1500 1000 1000 lifeCutter(m3/cutter) Cutter lifeCutter(m3/cutter) ~ 20% Qtz ~ 10-20%Qtz 500 500 ~ 50% Qtz > 70% Qtz 0 0 0 50 100 150 200 0 50 100 150 200 UCS (MPa) UCS (MPa) Meta Igneous (such as Mica Gneiss) Volcanics Rocks (such as tuff, andesite, and basalt) 3000 3000 2500 2500 2000 2000 1500 1500 1000 1000 Cutter lifeCutter(m3/cutter) ~ 0-20% Qtz 500 500 Cutter Consumption (m3/cutter) ConsumptionCutter ~ 25% Qtz ~ 45% Qtz ~ 20-50% Qtz 0 0 0 50 100 150 200 0 50 100 150 200 UCS (MPa) UCS (MPa) Granite and Gneiss 3000 2500 2000 1500 1000 Cutter life(m3/cutter)Cutter 500 ~10% Qtz ~ 30-40% Qtz 0 0 50 100 150 200 UCS (MPa) Figure 8- 16. Cutter consumption as a function of UCS and quartz content Being able to estimate cutter consumption, it is possible to predict the cutter change time which was previously discussed. 181 Discussion and Conclusions In order to evaluate and predict the downtime related to cutter change, a database of activity time for various components of TBM operations was established. The analysis of the information in this database, and in particular the time allocated for inspection and changing of the cutters has been the focus of this study. The conclusions of the time studies and related analysis can be summarized as follows: Reported cutter change time usually contains cutter inspection, and the duration of the cutter inspection can sometimes be a substantial amount relative to the actual cutter change time. Also, cutter change simultaneous to other activities is the next important factor for decreasing cutter change-related downtime. Evaluation of the synchronicity of the cutter inspection and other simultaneous activities requires very detailed and consistent reporting of the tunneling activities from various projects, which has not been practiced very often. The type of data available and reporting of the activity times are different from project to project depending on their management and scheduling of operation. Results of the preliminary analysis show that on average cutter change time is 66 minutes and there is a strong relationship between cutter change time and boring time. While it is anticipated that cutter size might have an inverse relationship with cutter change time, the result of the analysis of data from a few projects showed that the differences of cutter change downtime for different cutter sizes are not very high and in some cases larger cutters can reduce cutter downtime. Although, rock abrasiveness is recognized as the most important parameter in cutter wear prediction, our analysis does not show a strong correlation between CAI and 182 cutter consumption. This might be related to the fact that the average CAI values were considered for the whole length of tunnel drives. However, the results show that there is a need for the more in-depth study on the application of CAI in prediction of cutter consumption, especially considering the impact of rock joints and rock mass characteristics. The results of this detailed study can lead to development of more accurate formulas for prediction of cutter consumption from CAI test results. Based on the limited analysis of the available data, it seems like the NTH method can produce better estimates of cutter consumption. The results of the statistical analysis show that the impacts of cutter size or penetration per revolution are less than the effects of rock type, UCS, and quartz content for cutter prediction of consumption. Further analyses of these parameters are still underway to develop more versatile models for estimation of cutter consumption to allow use of different variables and to account for other important machine and cutter parameters for more accurate prediction of cutter consumption and related cutter change time and machine down time. 183 Chapter 9 Learning Phase Introduction Study of TBM performance parameters shows that daily advance rate is typically lower at earlier stages of the operation due to lower utilization. Generally, the daily advance rate of the TBM starts at a low value and increases gradually to reach a normal rate as the operators learn about the machine and its capabilities as well as the machine-ground interaction. During this early stage, the tunnel crew fine tunes the auxiliary operations to achieve a consistent level of production as they streamline the activities and gradually increase machine productivity. Fig. 9-1 shows a diagram of AR versus elapsed time or Cumulative Advance Rate (CAR) versus elapsed time. The LPP diagram is generally curved while the Normal operation/production phase (NPP) portion of the diagram is a straight line (Fig. 9-1) Advance Rate (AR) Cumulative Advance Rate (CAR) LPP NPP LPP NPP Time Time Figure 9- 1. Schematic sketch of LPP and NPP for two common TBM progress diagrams 184 As noted by Wais and Wachter (2009), the major factors influencing the learning phase (or Start-up) can be divided into the four following groups: Man / Personnel o Qualification and motivation o Construction site organization o Communication on the construction site Machine and support system o Machine type o Adaptation measures o Condition of the driving system (TBM) o Conditions of the back-up system Type of ground support o Geology o Geological formation / conditions o Hydro geological conditions o Changes between sections of homogeneous geology General conditions o Degree of difficulty o Intensity of work preparation o Local or Project Specific Conditions Wais and Wachter (2009) presented an example of the effect of learning on ring erection time and cycle time in an EPB shield TBM with the diameter of around 5 m (Fig. 9-2). A simple observation is that, as number of installed rings (horizontal axis) increases, the required time for 185 ring installation (vertical axis) decreases. This means, as tunnel excavation continues, the efficiency of the crew in performing their related tasks improves and overall one can conclude (and it is evident) that utilization increases. Meanwhile the question is how fast an operation can reach a consistent production and how to account for this phase in the calculation of TBM production, advance rate, and overall tunnel completion time, especially for shorter tunnels, where the learning phase could be a substantial part of the entire tunnel excavation time. Figure 9- 2. Ring building and cycle times according to the Herrenknecht control system (Waise and Wachter, 2009) 186 Modeling LPP Effect A rule of thumb in the TBM tunneling industry for evaluation of the LPP is to consider the first-month period as the LPP and the remaining as NPP. Laughton (1998) showed an analysis of 48 TBM data sets using this method by showing the "Start-Up Efficiency," that is the percentage of first-month AR to the Average Advance Rate (AAR) of the remaining months (Fig. 9-3). This method is a rough estimate and cannot reflect the effects of different conditions on LPP. As reported by Wais and Wachter (2001, 2009), among the different functions, the exponential function is one of the common formulas used for estimation of the effect of LPP (Gehring and Wachter methods, Eq. 9-1 and 9-2). In their proposed model, a familiarization factor is calculated, which reflects the speed of adaptation of the procedures for different tunneling activities to achieve the highest possible (normal) production rate. 12 10 8 6 4 No. of Tunnel Drives Tunnel of No. 2 0 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100 Start-Up Efficiency (%) Figure 9- 3. Histogram of Start-Up efficiency for gripper TBMs according to Laughton (1998) (9-1) 187 where = Percentage of the maximum rate of advance (familiarization factor) d = Duration of tunneling in months ( ) (9-2) where t = duration of tunneling in working days [wd] L(t) = Daily advance rate of day ‘t’[m/wd] c = Learning curve parameter obtained from Table 9-1 considering total rating of LRH, which is the summation of proposed ratings for human, machine, surrounding, and rock sub factors in different work conditions of good, standard, and poor (Fig. 9-4). = Parameter of the filter function for penetration rate which is IN/IB. IN = Net penetration rate [m/h] for a specific zone along the tunnel. IB = Reference net penetration rate [m/h] (based on the presented examples by Waise and Wachter (2009) IB is the average penetration rate over the entire tunnel length) a = Learning curve parameter selected from IB ((based on the presented examples by Waise and Wachter (2009) 'a' is the average advance rate over the entire tunnel length) [m/wd] In these methods, the formulas are used for the entire tunnel excavation period from the beginning to the end. The difference between the equations for LPP and NPP is that the familiarization function values in NPP are very close to 1. The component of in Eq. 9-2 (here is called ) is similar to in Eq. 9-1. is the percentage of the NPP advance rate determined at various times during the learning period, assuming that all geological parameters and rock mass stay the same which means . According to Gehring methodology, can represent the end of LPP (Waise, 2002). Table 9-2 shows the back 188 calculated corresponding time for for various types of TBMs under standard to good conditions, by using the average value of c for three categories of learning condition. Single Shield Double Shield Poor Standard Open Good 0 0.05 0.1 0.15 0.2 0.25 Parameter c Figure 9- 4. Learning curve parameter c for different TBM types (Waise and Wachter, 2009) Table 9- 1. Learning conditions rating according to Waise and Wachter (2009) Group Factors Standard Good Poor Points Personal permanent staff 40-50 %, familiar with tunneling, 100 % permanent staff, very flexible working low amount of permanent staff, labors from third enough auxiliary staff avail. flexible working hours, hours, following an earlier site (from former sites) world countries, high fluctuation rigid working hours small fluctuation rules Organization Clear allocation of Function and responsibility, to Organization already in practice (from former Unclear Functions and Responsibility experienced staff sites) Communication good ability to communicate in one common Communication already in practice (from former no or only minor ability to communicate in one language for the key positions sites) language Human Diameter Working space and power of the machine match Lower planned diameter (performance reserve) Diameter does not match with machine and trailer with diameter concept (too big, too small) TBM type and tested and familiarized to the key personal, System already in practice (from former sites) System and its components do not fit together trailer system suitable for soil conditions, suitable trailer, good logistics Condition TBM and trailer in a good refurbished condition, New system , low prone to break down TBM and trailer used, high prone to break down standard prone to break down Support tested and familiarized to the key personal, System already in practice (from former sites) unaccustomed, unpractical support system suitable for TBM type Machine Infrastructure Good accessibility, sufficient area, electric power Good accessibility, sufficient area, electric power poor accessible, poor conditions of area insufficient and water and water. Already developed site from former water and electric power construction Supply competitive suppliers, enough area for storage, already known suppliers from former sites, no time new or unsuitable suppliers, lack of storage area, suitable spare stock pressure insufficient spare stock Starting situation Filling of key positions already known minor complete personal available, no obstacles insufficient staff available, many obstacles by obstacles by temporary measures low weathered temporary measures, no weathered soil temporary measures, insufficient start position, soil and water at the start, secured start position and no water at the starting position secured start completely weathered soil with water during start, (Abutment Frame, Starting trestle, Start Ring) position (Abutment Frame, Starting trestle, Start high time pressure Ring) Surrounding Formation No gas, loose rocks, drilling possible low water No gas, stable, good to very good drillable (not too Gas, unstable soil, Water inflow, many changes in inflow hard), no water inflow soil conditions Rock LRH: * Ratings: 1 for poor; 3 for standard; 5 for good. LRH is the total sum of all of the ratings. LRH ranges for poor, standard, and good conditions are 11-22, 23-43, and 44-55 respectively. ** It should be noted that Waise and Wachter (2009) did not provide any relationships between LRH and ‘c’. 189 Table 9- 2. Corresponding values in standard to good conditions Values of Constant "c" in Corresponding Time of Corresponding Time of TBM Type Standard to Good Conditions (days) (months) Double Shield TBM 0.02 210 7 Open TBM 0.05 60 2 Single Shield TBM 0.025 120 4 According to Waise and Wachter (2009), the corresponding LPP times for Double Shield and Single Shield TBMs are rather high while in common cases they are around two months for these types of machines. Table 9-3 provides the two basic parameters of Eq. 9-2 (a and c) through back analysis on 31 tunnel weekly progress information assuming . It should be noted that a and c were explained previously after Eq. 9-2. Having values of 'a' from real weekly data, it was possible to back calculate parameter 'c' using Eq. 9-2. The condition in Table 9-3 refers to categories of the c parameter shown in Fig. 9-4. The last column represents the corresponding time of the L(t) function (Eq. 9-2) at which 95% of 'a' is obtained. Table 9- 3. Back analysis results of Eq. 9-2 parameters for 31 tunnel cases Excavated Excavated Corresponding Time a c Condition Diameter (m) Length (m) of (weeks) 3.52 6130.3 158.9 0.1813 Good 3 3.35 9520 176.6 0.055 Good 8 3.56 6954 191.95 0.0155 Standard 27 3.56 7350 151.81 0.0176 Standard 24 4.8 5200 110.83 0.0295 Standard 15 4.5 21300 145.56 0.0153 Standard 28 4.5 15686 133.92 0.0139 Standard 30 11.81 1576 104.29 0.0035 Standard >24 6.73 5223 113.56 0.0344 Good 13 3.84 5613.5 142.5 0.028 Standard 15 7 7654 55.79 0.1154 Good 4 7 9559 60.42 0.1525 Good 3 3.6576 2540.2032 109.55 0.0192 Standard 22 3.4 13400 278.67 0.0944 Good 5 3.43 2073 84.34 0.0994 Good 4 3.5 10120 237.23 0.0748 Good 6 4.88 13060 140.72 0.0854 Good 5 8.07 7201.65 148.07 0.0346 Good 12 3.7 15880 160.53 0.0562 Good 8 11.8 3480 68.75 0.0244 Standard 17 11.98 4326 109.12 0.0209 Standard 20 12.35 5620 125.91 0.0161 Standard 26 190 3.9 2960 117.65 0.0237 Standard 18 6.5 10314.5 83.16 0.0227 Standard 19 3.9 2890 80.86 0.0524 Good 0.5 3.9 6558.5 118.33 0.5 Good 0.5 3.62 2820.4 97.82 0.5 Good 0.5 3.63 5930 146.51 0.072 Good 6 3.9 4412 131.21 0.5 Good 0.5 3.9 1919 163.55 0.051 Good 8 6.5 6914 95.2 0.5 Good 0.5 The corresponding time of (which is the end of LPP) for the standard condition in their calculations is between 15 and 30 weeks which seems relatively high. The following list gives some of the reasons that might cause over-estimation of the learning phase period, which can result in under-estimation of the overall utilization and advance rate of the operation. Using the best-fit formula for all working weeks. The learning curve parameters can be affected by the TBM progress fluctuations of the normal-weeks period as shown in the Fig. 9-5 example. The best-Fit function does not form a convex shape as shown in the Fig. 9-6 example. The transition point slope is zero. Since there is one fitting function for learning and normal phases, the transition point is unclear due to having a continuous line and in some cases, learning parameters can push the beginning point of the normal phase- fitted line out of a long period of time. 191 250 200 95% of Normal Advance 150 100 Weekly Advance Rate (m) Rate Advance Weekly 50 0 1 6 11 16 21 26 31 36 41 Week No. Figure 9- 5. Learning phase over-estimation because of TBM progress fluctuations of normal weeks (using Waise and Wachter fitting function) 250 200 95% of Normal Advance 150 100 Weekly Advance Rate (m) Rate Advance Weekly 50 0 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 Week No. Figure 9- 6. Learning phase over-estimation because of concave shape of learning curve function (using Waise and Wachter fitting function) This chapter is an attempt to provide a new methodology for assessing the LPP based on previous case histories of the field performance of several TBMs. 192 New Methodology The practical application of this analysis and the proposed models are in performance prediction of a TBM for a new project. In such applications, one normally develops an estimate of the penetration rate (PR) based on available models and then an estimated Utilization rate (U) is used to calculate an advance rate. Alternatively, other models QTBM and RME can be used to directly estimate AR from a certain set of input parameters. Nonetheless, in both cases the estimated daily advance rate is based on what is called normal operating conditions in a given reach or stretch of the tunnel and does not reflect the learning period. As such, given that the estimated AR reflects NPP, if a model could estimate the LPP and a function could show the progression of the estimated AR or U through this period, a more accurate estimate of both parameters could be achieved. The proposed method involves breaking the graph of advance rate of a specific time increment (i.e. week) into two portions of the Learning Phase Period (LPP) and the Normal Phase Period (NPP). For LPP, a linear function (Eq. 9-3) was selected to represent the gradual increase in advance rate (or alternatively utilization). For NPP, a horizontal line (Eq. 9-4) is selected to represent the average advance rate of the normal phase (Y1) based on the common performance prediction models. It should be noted that for LPP, other functions such as a polynomial function might be more realistic since they can describe the shape of the LPP curve better when it has a convex or concave shape. However, use of a nonlinear equation complicates obtaining the coefficients needed to estimate the LPP. The analyses for polynomial function coefficients turned out to be very susceptible to the variation of the other information (e.g. tunnel diameter) that was collected for several tunnel projects. Hence, the simpler form of a linear function was chosen for LPP analyses. 193 (for 0 (9-4) Fig. 9-7 shows the elements of the LPP function in the new method. X1 and Xn represent the ending time of LPP and the ending time of NPP (or the project), which is a period of operation where the machine can achieve its norms in productivity, respectively. Weekly Advance Rate (AR) Y1 LPP NPP Week No. X1 Xn Figure 9- 7. Schematic fitting functions elements 194 Input: Weekly Cumulative Advance Rate (CARi; i=1 to n) i=1 Y1=(CARn-CARi)/(n-i) Error(i)=Abs{[(Y1*i*0.5)-CARi)]/ CARi} i=i+1 Y1=(CARn-CARi)/(n-i) Error(i)=Abs{[(Y1*i*0.5)-CARi)]/ CARi} No Error(i-1)< Error(i) Yes i=i-1 X1=i Y1=(CARn-CARi)/(n-i) Output: X1, Y1 Figure 9- 8. Flowchart of finding the parameters of fitting functions In order to find the unknown parameters of the LPP and NPP functions, the steps of the flowchart of Fig. 9-8 are followed. The main procedure in this flowchart is to make the area under the LPP fitting function (A1 in Fig. 9-9) equal to or as close as possible to the corresponding CAR of X1. Fig. 9-10 shows two examples of obtained fitted functions for the periods of LPP and NPP. 195 Weekly Advance Rate (AR) Y1 A1 A2 Week No. X1 Xn Figure 9- 9. Areas under fitting functions 200 150 100 Weekly Advance Weekly Advance (ft) 50 0 1 6 11 16 21 26 31 36 41 46 51 Week No. 600 500 400 300 200 Weekly Advance Weekly Advance (ft) 100 0 1 6 11 16 21 26 31 Week No. Figure 9- 10. Examples of obtained fitting functions This procedure can be applied for different periodic times, such as daily, weekly, and monthly. Obviously, in highly variable rock mass conditions, it is better to use the utilization factor instead of the advance rate to compensate for the difference in Rate of Penetration (ROP) within different zones. It should be noted that in this case the utilization factor should represent 196 the LPP, meaning in calculating the utilization factor, the total time should be working time which includes weekend maintenance. One problem with using the utilization factor instead of the advance rate is the unreal high values of utilization factors during LPP due to slow cautious penetration of the TBM that can lead to an inaccurate outcome for LPP parameters. Another point is that best fitting functions are susceptible to errors due the procedures for calculation of the NPP parameter, meaning that having unproductive or low-productive weeks can affect the total area under the functions, and this causes estimation of inaccurate values. Unproductive or low-production weeks especially in "Adverse Ground Conditions” have nothing to do with the learning period and training of the staff or adjusting the TBM operation to the job site. Therefore, if a period of time is certainly not related to LPP or NPP, it should be eliminated from the calculations. Evaluation of Proposed LPP Function Parameters As noted earlier, one important parameter that is necessary for LPP is the duration of LPP (X1). The Waise and Wachter methodology (2009) considered various parameters for evaluation of learning time on TBM advance rate. One ambiguity of this methodology is that the AR for both LPP and NPP is obtained from one formula and there is no explicit definition for the end of LPP. Furthermore, the reduction effects of the learning phase is considered for the whole period of tunneling or a long portion of it (refer to Table 9-2). As discussed before, in the new proposed methodology, two separate fitting functions are introduced to represent the common trends better. Furthermore, a common practice in AR prediction is to add the AR of LPP separately to the AR of NPP (e.g. see Abd Al-Jalil (1998), Laughton (1998)). Fig. 9-11 shows the difference between the new fitting functions and the exponential function of Waise and Wachter (2009). 197 Weekly Advance Rate (AR) Typical trend of Eq. 2 C Y1 A1 A2 A B Week No. X1 Xn Figure 9- 11. Comparison between linear and exponential functions for LPP A series of analyses was conducted on the information from 44 tunneling projects to estimate parameters for the new LPP model (refer to flowchart of Fig. 9-8). Based on the analysis, the ratio of X1/D (called as learning-phase ratio) turned out to be one of the best parameters for LPP evaluation for different tunnel diameter sizes (D). Fig. 9-12 shows the histogram of the distribution of this ratio for these tunnels. 14 Mode=0.5 Median=1 120 44 Tunnel Cases 12 100 10 80 8 Average=1.5 60 Count 6 40 4 Cumulative Percentage 2 20 0 0 0-0.5 0.5-1 1-1.5 1.5-2 2-2.5 2.5-3 3-3.5 3.5-4 4-4.5 Learning Phase Ratio Figure 9- 12. Histogram of distribution of learning phase ratio for 44 tunnel projects As can be seen in Fig. 9-12, the distribution of the learning-phase ratio is highly skewed; therefore, the average value is not an appropriate parameter for this data. The median simply shows the most appropriate central value for the data set examined. This means that if the 198 influence of other factors are ignored, on average we need to consider 1week/m of diameter for LPP. If we want to follow the most commonly practiced LPP, this ratio is 0.5 week/m of diameter. Overall, in lack of information, especially crew experience which is very hard to judge, the ratio of 0.5-1 weeks/m of tunnel diameter is recommended for LPP calculations. Further analyses were performed to include other factors and to have better understanding of the learning-phase ratio. The results of the final analysis are shown in Fig. 9-13. In this graph, LR is LPP rating which can roughly be obtained from Table 9-4. This table is similar to what Waise and Watcher (2009) presented for obtaining the learning-condition rating (see Table 9-1). D is tunnel diameter in m and RMR is rock mass rating. Table 9- 4. LR rating in different learning conditions Good Standard Poor -High experienced crew - Experienced crew -No/low experience - TBM and BU low prone to break - TBM and BU standard prone to - TBM and BU high prone to break down break down down - Good logistics - Regular logistics - Bad logistics LR=100+RMR LR=50+RMR LR=RMR RMR: Rock Mass rating 199 5 y = -0.0255x + 5.1 4 R² = 0.92 3 2 X1 /D (week/m)X1 1 - 0 50 100 150 200 LR Figure 9- 13. Analysis results for obtaining X1/D Once the ratio of X1/D is obtained, X1 and A1, which is the excavated length of tunnel during LPP, can be easily calculated. Having A1 and X1, it is possible to calculate the Xn (total tunneling time) as shown in Eq. 9-5. (9-5) In this calculation, ‘A’, the area under the curve, is in fact the total length of the tunnel and ‘A1’ is the length of tunnel excavated in the learning period. Having X1, Xn, and Y1, it is possible to modify the tunnel advance for the learning phase. Table 9-5 gives the summary of the procedure for estimation of LPP and related parameters. 200 Table 9- 5. Summary of formulas and calculation results Parameter Formula Unit Learning Rate as a function of LR Ground conditions per Table 9-4 D TBM diameter m X1/D -0.0255LR+5.1 Week/m X1 Week A1 X1*Y1 *0.5 m Xn X1+(A-A1)/Y1 week Average AR for A/Xn m/week entire tunnel Summary and Discussions The learning Phase Period (LPP) in the TBM excavation process is a low-production phase which can be distinguished easily from the Normal Phase Period (NPP) in almost all TBM tunneling projects. LPP can affect scheduling the tunnel excavation and assessing the tunnel project cost. This is especially true for shorter tunnels in hard rock. The prediction of LPP depends on several factors, and it is still a difficult and uncertain task. In this chapter, a new methodology for estimation of the LPP was introduced. A linear function is used to obtain the LPP parameters for simplicity and improved accuracy, especially when an estimate of anticipated advance rates during the normal period is available. The proposed formulas can provide a very good preliminary estimate for calculation of the LPP and related characteristic parameters. 201 Chapter 10 Simulation of Tunneling Activities for Prediction of Machine Utilization Introduction The process of boring with a TBM consists of some sub-systems and activities. Sub- systems are components of the related equipment, used as a resource to accomplish a certain task. Activities refer to the tasks that are completed in sequence to allow the tunneling to progress. Any of these subsystems and activities can introduce certain delays in the whole process. The most common sub-systems and activities in TBM tunneling are: Sub-systems: Cutter head, TBM, Back-up, Tunnel haulage system. Activities: Maintenance, Ground Supporting, Boring, Regrip, Surveying, Extending utilities, probing and exploration, and inspections. Obviously, on top of the time spent on the tunneling subsystem and activities and related downtimes, there are other time intervals that are unproductive and are miscellaneous delays, often caused by site-management issues. In simulations, each sub-system has its own sources and probability of failure which causes specific interruption in the boring process for repairing. “Failure” refers to any of the sub- systems being unable to perform its designated tasks, and “Repair” refers to the period of time in which a sub-system is out of service to restore the ability of the given subsystem to perform its related tasks. Based on these two definitions, it is possible to divide the total working time into two main time categories including Time Between Failure (TBF) and Time To Repair (TTR). There are three reference counters for the analysis of the tunneling activities and related modeling 202 and simulation including “Boring Time” (BT), Shift Time (SHT), and Distance along the tunnel, which are used to calculate TBF, TTR, and “Distance Between Failure” (DBF) as shown in Fig. 10-1. TTRi TTRj TTRk TTRl TTRi TBFi 0 SHT TTRi TTRi TBFi 0 BT TTRi TTRi DBFi 0 Distance Figure 10- 1. Schematic of TBF, TTR, and DBF with reference to different counters (Abd-Al- Jalil, 1998) Having mean values of TBF and TTR, it is possible to evaluate the approximate value of the TBM utilization (U) (Eq. 10-1) (Abd-Al-Jalil, 1998). (10-1) For simulation purposes, the distributions of each sub-system TBF and TTR should be constructed based on the previous TBM case histories. These distributions are used, to be sampled randomly during each cycle of the simulation. Different sub-systems might work in sequence or parallel. Depending on the method of downtime (DT) recording. there are usually two methods for calculating the sub-system parameters as follows: 203 The exact start and end times of each subsystem repair are available so that it is possible to simulate parallel activities. The delays of each sub-system and activity are available so that it is necessary to use some simplifications to calculate sub-systems parameters. For example, it is possible to assume that the time of a failure is in the middle of the corresponding BT or SHT (Fig. 10-2). The outcome of a simulation would be finding the critical activities, utilization of each sub-system, Time To Complete Tunnel (TTCT, Cost To Complete Tunnel (CTCT) (if corresponding data is available), and average Advance Rate (AR) for the whole length or part of a tunnel (Fig. 10-3). The analysis for the whole length is used in the pre-construction period while the analysis for sections of a tunnel is used during the construction period to reflect the updated information in the final time and cost predictions. It should be noted that the Learning Phase Period (LPP) delays, Major delays (such as delays of Extreme Mining Areas (EMA), TBM overhaul, …), and Holidays are defined deterministically (unique value) since they are specific to each tunnel case. 204 TTRi BT TTRi TBFi 0 SHT TTRi TTRi TBFi 0 BT TTRi TTRi DBFi 0 Distance Figure 10- 2. Schematic of TBF, TTR, and DBF with reference to different counters for the case in which the exact occurrence time of a delay is not specified 205 Input: Distributions of TBF/TTR of different categories of Boring, TBM, BU, … Simulation LPP Delays Major Delays Holidays Delays Adding manual operations to the simulation results Output: Critical Activity Distribution of U Dis. of TTCT, CTCT, and AR Figure 10- 3. Process and outcomes of a typical simulation modeling Modeling of Random Variable For simulation of the tunneling process, the input values of parameters are either probabilistic or deterministic. Probabilistic parameters can be selected from their corresponding Probability Density Functions (PDF) using some known distribution equation or via empirical equations. Table 10-1 is a summary of the most commonly used continuous distribution functions to be used as PDFs and their related parameters. 206 Table 10- 1. Continuous distributions and their parameters (Summarized from Altiok and Melamed, 2007) Distribution Input PDF formula, Common use Name Parameters Uniform Unif (a, b) In absence of information Step Unif (lj, rj) Empirical distribution Triangular Tria(a, c, b) Unknown distribution with known min, max, and most likely values Exponential Expo( ) Inerarrival times, memoryless property Normal Norm( ) Evenly distributed values Lognormal Logn( ) Gamma Gamm(α,β) Beta Beta(α,β) Weibull Weib(α,β) 207 Distribution Mean Variance Example of PDF Name Uniform Step Triangular Exponential Normal Lognormal 208 Gamma Beta Weibull If none of the known distributions fit the data collected from field application of TBMs, it is possible to use discrete or continuous empirical distributions (Empirical Density Function (EDF), or Empirical Cumulative Density Function (ECDF)) which are simply the set of probabilities and their corresponding values. Background of TBM Performance Prediction with Simulation There are three types of TBM performance prediction methods in the tunneling industry including deterministic, worst-expected-best-case-scenario based, and probabilistic methods. In the deterministic method, which is usually useful for more homogeneous rock masses, the predicted TBM performance parameters are unique, while in probabilistic methods, the parameters used in the modeling are varied among a range according to their corresponding probabilities. The latter method is mostly useful for the conditions in which the rock mass parameters have either high variation or uncertainty. The worst-expected-best-case-scenario 209 based method gives a range of values without any implication regarding confidence values and probability. This method provides the planner the means of expressing doubt of the different parameters including management and completion time (Laughton, 1998). The probabilistic method uses a variety of probable scenarios to obtain Probability Density Functions (PDFs) and risk relative to TTCT (Fig. 10-4). In this case, the area under the curve between two TTCT values represents the probability of the tunnel excavation within that time frame. This information can be used to select proper methods and means of tunnel excavation by taking into account both the severity and likelihood of an occurrence. It should be noted that risk here refers to a deviation from “a desired outcome that is expected or hoped for” as noted by Redja (1992) and Laughton (1998). 210 Tunnel Longitudinal Profile Tunnel Time (day) Determinitic Method 40 35 30 25 20 15 Holethrough 10 5 0 0 100 200 300 400 Distance (m) Time (day) TTCT Worst-Expected- Best Distribution 40 Case Senario-Based Method 35 30 25 20 15 10 Holethrough 5 0 0 100 200 300 400 Distance (m) Time (day) TTCT Probabilistic Method Distribution 40 35 30 25 20 15 Holethrough 10 5 0 0 100 200 300 400 Distance (m) Figure 10- 4. Different methods of TBM TTCT prediction (revised from Laughton, 1998) 211 f (x) F(x) 1.1 0.2 1 Longest Credible 0.9 0.8 0.7 0.6 Best Estimate 0.1 0.5 0.4 0.3 0.2 Shortest Credible 0.1 0 0 28 33 38 28 33 38 TTCT(day) TTCT(day) Figure 10- 5. Typical outcomes of probability method for prediction of TTCT (Laughton, 1998) Search Methods Nelson et. al (1994), Abd-Al Jalil (1998), and Laughton (1998) are among the first researchers who utilized simulation techniques to predict Time To Complete Tunnel (TTCT) and Cost To Complete Tunnel (CTCT). They developed a very comprehensive data base in four levels for which the details of information increases as the number of level increases. The summary of their methods are explained in Fig. 10-6 and 10-7. 212 Figure 10- 6. Method A of TBM performance prediction with data base level 1 (Laughton, 1998) Figure 10- 7. Method B of TBM performance prediction with data base level 2 (Laughton, 1998) 213 The basic idea for these methods is to search in the data base for a sample to find like- case tunnels performance parameters (mostly AR) and to generate a sample of driving rate (AR-1). This sample will be fitted to a PDF to be used in the simulation to develop probabilistic TTCT values. In method B, the length of a tunnel zone might be determined by using sampling from a PDF (such as a triangular distribution as shown in Fig. 10-8) where the uncertainty of the exact location of the boundary of two consecutive zones is relatively high. In a highly unpredictable setting, the sequencing of the units may also be predicted by using transition state probabilities as described by Einstein (1992) and Laughton (1998). Tunnel Longitudinal Profile Zone 1 Zone 2 Tunnel Length Minimum Estimated Zone 1 Length Maximum Estimated Zone 1 Length Probability Zone 1 Length Figure 10- 8. Probabilistic estimation of a zone length to be used in simulation of method B (Laughton, 1998) The above-mentioned simulation methodology was used to predict the TBM performane of Manhattan (south loop). The South Tunnel is approximately 5,613.50 m (18,417-ft) long and its depth ranges from 156.36 m to 175.87 m (513 ft and 577 ft) below the ground surface. The excavated diameter is 3.84 m (12.6 ft) and the finished tunnel diameter is 214 3.05 m (10 ft). The South Tunnel drives through predominantly mica schist and schistose gneiss rock (Kwiatkowski et al., 2007). In order to find a sample set of like-sites from the TBM data base, a set of search criteria (Table 10-2) similar to the ones used by Luaghton (1998) was used, which includes the range of tunnel diameter, rock type, rock strength, and TBM type. Once the set of like-sites was selected, the performance parameters were obtained as listed in Table 10-3. Table 10- 2. Search criteria for Manhatan South Tunnel like-sites Data Base Parameter Search Criteria Diameter (m) 3-5 TBM Type Open Cutter Diameter (in) 16-19 Rock Type Schist and Granite Shaft Entry Depth None Muck Evacuation System Train and Wagon Length (km) 3-10 UCS (MPa) <100 Table 10- 3. Performance parameters prediction for Manhatan South Tunnel (based on 12 cases) PR ARw ARb Uw Ub TTCT (day, working Parameter (m/hr) (m/day) (m/day) (%) (%) /boring) Average 2.96 23 28 30 36 242/203 Std 0.99 7 7 10 8 7/7 COV (%) 34 32 24 32 22 32/24 Min 0.70 15 18 20 27 160/160 Max 5 35 35 46 49 374/312 Real Values 4.23 24.2 27.4 23.8 27 232/205 Rock Mass Simulation Method Laughton used Rock Mass Condition (RMC) as the cornerstone of his model to evaluate the Rock Mass Behavior (RMB). The different steps of this model are depicted in Fig. 10-9. 215 Tunnel Zone Tunnel Cells Block Size Step A Alteration State Step B RMC 3 1 5 3 1 3 4 6 1 3 Step C RMB II I III II I II II IV I II Step D Support C O O P O C C R C C Step E (O, occational bolts; P, pattern bolts; C, canopy; R, ribs) Figure 10- 9. Rock mass simulation steps (Laughton, 1998) The variability of a cell parameter is obtained from probability data derived from like- case histories/real data. In step A of this model, the next cell state is determined by Markov transition probabilities through back analysis of case history map records. The transition counts yield the probability of transitioning to a new state, which is used to generate the model block size. The average distance between Exreme Mining Areas (EMA) obtained from the level 2 data base (Table 10-4) is then interjected in the mining cells to account for these extreme conditions in the transition counts. After defining the step A cell states, the other step cell states are defined based on the probabilities of the previous cell states. 216 Table 10- 4. Average number of kilometers per EMA (Calculated from level 2 database of Nelson et al. (1994)) Track-bound System Simulation Track-Bound Transport In hard rock TBMs, excavated material is usually transmitted from the face to the backup or tunnel haulage system using a conveyor belt. The tunnel transportation system can be a rail- bound system, a trackless system, or a continuous conveyor belt. Among the choices of material haulage, the track-bound system is more common, especially for small-to-medium-sized tunnels. A track-bound system includes rail tracks, a locomotive, and a muck car train. This section will 217 discuss the evaluation of the track-bound transportation system for optimal excavation process not disturbed or delayed by insufficient transport capacity. For this analysis the same notations as the ones used in the NTH model offered by Bruland (1998) are used (Fig. 10-10). The main idea or optimal design of the tunnel transportation system is to maximize TBM excavation and minimize time lost for muck car trains exchanging. In order to achieve this goal, the cycle time of one train (going from the face to the portal or shaft) should be less than or equal to the cycle time of one TBM stroke. Obviously, as tunnel length increases, the cycle time of trains between the tunnel portal and the California switch increases. When this cycle time becomes larger than the cycle time of loading and train exchanging between back-up and TBM, the number of California switches should be increased. In figure 10-11 and 10-12, different arrangements of trains for different states of California switches inside a tunnel are depicted. The main idea in these arrangements is to consider the optimum locations of the California switch for the lowest haulage delays. This is done with the consideration of safety, to have more freedom in the case of an emergency, when continuous excavation should be paused for evacuation or for shift change. Table 10-5 compares the estimated demand for the locomotive and trains for different scenarios with various numbers of California switches inside the tunnel with the ones offered by Bruland (1998). Tables 10-6 to 10-8 and Figures 10-12 to 10-16 and Eq. 10-2 to 10-10 describe different stages of tunneling and its related components in various scenarios. Table 10- 5. Demands for locomotive and muck train Number of California Switches Demand for Loco Demand for Train inside the Tunnel 0 1 (1) 2 (2) 1 2 (2) 4 (3) 2 3 (3) 5 (4) 3 4 (4) 6 (5) (NTH model) 218 Figure 10- 10. Schematic of California switch and locomotive (upper figure from Maidl et al., 2008) Dump Station TBM T0 : Start of a cycle 1 2 T : End of a cycle 1 1 2 1 TStart2 : Loaded of a cycle train out 2 1 TStart3 : Loco of a cycleshunting 2 Start of a cycle 1 T4 : Empty train in 2 1 StartT5 : End of a of cycle a cycle 2 2 TStart6 : Loaded of a cycle train out 1 Figure 10- 11. Schematic view of different stages of tunneling using two trains without any California switch 219 Dump Station TBM T : Start of a cycle 2 1 0 3 Start of a cycle T1 : End of a cycle 2 1 3 Start of a cycle T2 : Train exchange 1 2 3 Start of a cycle 1 T3 : Loaded train out 2 3 Start of a cycle 1 T4 : Empty train in 3 2 Start of a cycle 1 T5 : Train exchange 2 3 Start of a cycle 2 T6 : Loaded train out 3 1 Start of a cycle 2 T7 : Empty train in 1 3 Start of a cycle Figure 10- 12. Schematic view of different stages of tunneling using three trains with one California switch Table 10- 6. Transportation time components of different sections of a tunnel using three trains with one California switch From To Portal-California Switch 1 California Switch 1-TBM T4 T5 tex-cs+2 tbu-cs T5 T6 tout+tex-p tload T6 T7 tin T4 T7 tin+ tout+tex-p tex-cs+2 tbu-cs+tload Note: T0-T3 are just to show the first steps of the beginning of the excavation process and will not repeat. T4- T7 are repeated. 220 Time Item Explanation Remarks Driving from the tunnel portal to the California t t =l /v in switch (including one switch passing in portal) in cs train lcs Distance from tunnel portal to California switch vtrain Train speed inside the tunnel tex-cs Train exchange on California switch Driving from the California switch to the tunnel t t =l /v out portal out cs train tex-p Train exchange in tunnel portal tbu-cs Driving distance between BU and California switch tload Loading the train at the BU tload=tstroke Nstroke+tregrip (Nstroke-1) tstroke Excavation stroke time Nstroke Number of strokes per train tregrip Regripping time tin+tout+tex-p= tex-cs+2 tbu-cs+tload (10-2) lcs = 8.33 vtrain (tex-cs+2 tbu-cs+tload - tex-p) (10-3) In order to have continuous excavation, the cycle time of Portal-California Switch 1 should be less than or equal to the cycle time of California Switch 1-TBM. A cycle time is the total duration to exchange a loaded train with an empty train including driving and loading. When the cycle time of Portal-California Switch 1 exceeds the cycle time of California Switch 1-TBM, another California switch should be added to the system. The optimal location of California switch can be obtained as follows. T BM II: Driving loaded train off I: Driving loaded train on the California switch the California switch I: Driving empty train II: Driving empty train on the California switch off the California switch Figure 10- 13. Schematic view of exchanging trains on the California switch 221 Dump Station 4 1 I: Leaving loaded train in tunnel access and shunting loco towards empty train 4 1 II: Driving empty train towards tunnel access 4 1 III: Shunting empty train 1 4 IV: Shunting loco 1 4 V: Shunting loaded train towards dump station 1 4 VI: Shunting loco towards empty train 1 4 VII: Driving empty train towards tunnel Figure 10- 14. Schematic view of exchanging trains on the portal switch when there is one loco with two trains 222 TBM 1 2 2 2 3 3 3 2 1 3 2 4 1 2 3 4 4 1 1 2 3 3 4 4 1 1 Dump Station 4 5 6 0 1 2 3 T T T T T T T Figure 10- 15. Schematic view of different stages of tunneling using four trains with two California switches 223 Table 10- 7. Transportation time components of different sections of a tunnel using four trains with two California switches California Switch 2- From To Portal-California Switch 1 California Switch 1-2 TBM T3 T4 tex-cs+2 tbu-cs T4 T5 ta-in tb-out tload T5 T6 ta-out+ tex-p tex-cs+tb-in T3 T6 ta-out+tex-p +ta-in tex-cs+tb-in +tb-out tex-cs+2 tbu-cs+tload Time Item Explanation Remarks Driving from the tunnel portal to the California t t =l /v a-in switch 1 (including one switch passing in portal) a-in cs1 train lcs1 Distance from tunnel portal to California switch 1 vtrain Train speed inside the tunnel Driving from the California switch 1 to the tunnel t t =l /v a-out portal a-out cs1 train tex-cs Train exchange on California switch tex-p Train exchange in tunnel portal tb-in Driving from from California switch 1 to 2 tb-in=lcs12/vtrain lcs12 Distance between California switch 1 and 2 tb-out Driving from California switch 2 to 1 tb-out=lcs12/vtrain tbu-cs Driving distance between BU and California switch tload Loading the train at the BU tload=tstroke Nstroke+tregrip (Nstroke-1) tstroke Excavation stroke time Nstroke Number of strokes per train tregrip Regripping time In order to have continuous excavation, the cycle times of Portal-California 1 and California Switch 1-2 should be less than or equal to the cycle time of California Switch 2-TBM. A cycle time is the total duration to exchange a loaded train with an empty train including driving and loading. When the cycle time of Portal-California Switch 1/California Switch 1-2 exceeds the cycle time of California-TBM, another California switch should be added to the system. The optimal location of California switches can be obtained as follows. ta-out+tex-p+ta-in= tb-in+tex-cs+tb-out= tex-cs+2 tbu-cs+tload (10-4) lcs1 = 8.33 vtrain (tex-cs+2 tbu-cs+tload - tex-p) (10-5) lcs2= lcs1+8.33 vtrain (2 tbu-cs+tload) (10-6) 224 TBM 1 2 2 2 3 3 2 1 3 2 1 2 3 4 1 4 4 1 3 3 5 5 1 4 4 4 5 Dump Station 4 5 0 1 2 3 T T T T T T 225 TBM 3 4 4 4 4 3 5 3 5 2 5 3 5 2 1 2 1 1 1 2 Dump Station 9 6 7 8 T T T T Figure 10- 16. Schematic view of different stages of tunneling using five trains with three California switches 226 Table 10- 8. Transportation time components of different sections of a tunnel using five trains with three California switches California Switch From To Portal-California Switch 1 California Switch 1-2 California-TBM 2-3 T6 T7 tex-cs tex-cs+2 tbu-cs T7 T8 ta-out+tex-p tb-in tc-out tload T8 T9 ta-in tb-out tc-in+tex-cs T6 T9 ta-out+tex-p+tlsh+ta-in tb-in+tex-cs+tb-out tc-in+tc-out+tex-cs tex-cs+2 tbu-cs+tload Time Item Explanation Remarks Driving from the tunnel portal to the California t t =l /v a-in switch 1 (including one switch passing in portal) a-in cs1 train lcs1 Distance from tunnel portal to California switch 1 vtrain Train speed inside the tunnel Driving from the California switch 1 to the tunnel t t =l /v a-out portal a-out cs1 train tex-cs Train exchange on California switch tex-p Train exchange in tunnel portal tb-in Driving from from California switch 1 to 2 tb-in=lcs12/vtrain lcs12 Distance between California switch 1 and 2 tb-out Driving from California switch 2 to 1 tb-out=lcs12/vtrain tc-in Driving from from California switch 2 to 3 tc-in=lcs23/vtrain lcs23 Distance between California switch 2 and 3 tc-out Driving from California switch 2 to 1 tc-out=lcs23/vtrain tbu-cs Driving distance between BU and California switch tload Loading the train at the BU tload=tstroke Nstroke+tregrip (Nstroke-1) tstroke Excavation stroke time Nstroke Number of strokes per train tregrip Regripping time In order to have continuous excavation, the cycle times of Portal-California 1, California Switch 1-2, and California Switch 2-3 should be less than or equal to the cycle time of California Switch 3-TBM. A cycle time is the total duration to exchange a loaded train with an empty train including driving and loading. When the cycle time of Portal-California Switch 1/California Switch 1-2/California Switch 2-3 exceeds the cycle time of California Switch 3-TBM, another California switch should be added to the system. The optimal location of California switches can be obtained as follows. ta-out+tex-p+ta-in= tb-in+tex-cs+tb-out= tc-in+tc-out+tex-cs= tex-cs+2 tbu-cs+tload (10-7) 227 lcs1 = 8.33 vtrain (tex-cs+2 tbu-cs+tload - tex-p) (10-8) lcs2= lcs1+8.33 vtrain (2 tbu-cs+tload) (10-9) lcs3= lcs1+ lcs2+8.33 vtrain (2 tbu-cs+tload) (10-10) Fig. 10-17 shows a typical graph to find the final locations of different California switches along a tunnel. 12 CS1 CS2 CS3 10 8 6 PR (m/hr) PR 4 2 0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 Distance from Tunnel Outlet (km) Figure 10- 17. Final locations of different California switches (CS) for a typical tunnel with regripping time of 5 min and train speed of 15 km/hr In the following section the track-bound system simulation is performed for a tunnel project. For this simulation, Arena software (training/evaluation mode ver. 12) is used. For this purpose, two approaches are developed which will be discussed in detail. 228 Simulation of the Activities in a Tunnel Project The Karaj Water Conveyance Tunnel is designed to transfer 16 m3/s of water from the Karaj (Amir-Kabir) Dam to Tehran City. Lot 1 of this project is a tunnel with a 16-km length. This tunnel was excavated by a double shield TBM manufactured by Herrenknecht company. The main geological unit in this project is the Karaj formation, a well-known formation of the Alborz Mountains. This formation is composed of a variety of pyroclastic rocks, often interbedded with sedimentary rocks (Hassanpour et al., 2009). The site preparation started in 2004 and the TBM arrived at the site in May 2006, and mining commenced in August 2006. Figure 10- 18. Location of Karaj water conveyance tunnel, northwest of Tehran, Iran (Hassanpour et al., 2009) 229 Figure 10- 19. Double shield TBM at the ET portal (Hassanpour et al., 2009) Table 10- 9. Karaj TBM specification (Hassanpour et al., 2009) Parameter Value Machine diameter 4.65 m Cutters diameter 432 mm Number of disc cutters 31 Disc nominal spacing 90 mm Maximum operating cutterhead thrust 16,913 kN Cutterhead power 5*250 = 1,250 kW Cutterhead speed 0 to 11 rpm Cutterhead torque (nominal) 1,723 kNm (6.58 rpm) 1,029 kNm (11 rpm) Thrust cylinder stroke 1,400 mm Conveyor capacity (approx.) 200 m3/h TBM weight (approx.) 170 tons First Simulation Approach In the simulation model, there are three types of delays including delays for each ring erection (such as Boring and Support), daily delays (Maintenance, Other), and TTR and TBF for 230 three subsystems of Cutter, TBM, and BU. The delays of TTR for each subsystem are applied after TBF, which was calculated on the basis of boring time or TBM clock, meaning that after a certain boring time, these subsystems have a certain failure depending on TTR distribution values. Each subsystem has its own boring time counter which is set to zero after reaching the TTR time on the time line. For the TBF calculation, it is assumed that the failure occurred in the middle of the Bore Time (BT) of the corresponding day. In order to include the delays of non- production days (with Advance Rate (AR) of zero) in the distributions sets, the delays of non- production days were combined with the ones from the previous or the next production day. The distribution functions of all these categories are presented in Tables 10-10 to 10-12. The main control criterion for the fitted functions is R2 according to Abd-Al Jalil (1998). In case of R2≥0.9 (or 90%), the fitted function was directly used in the model, while in case of R2<0.9 the Empirical Cumulative Density Function (ECDF) was used. Fitting the distribution functions was performed with the use of the Input Analyzer Package of the Arena software. Table 10- 10. Best fit function for delays for each ring No. Category Name F(x) R2 1 Boring 0.07+LOGN(0.293,0.0737) 95.5 2 Support LOGN(0.618,0.174) 94.2 3 Reset NORM(0.15,0.0586) 83.3 4 Transport -0.001+GAMM(0.101,2.26) 96.8 5 Utility -0.001+WEIB(0.0878,0.928) 99.9 6 Survey EXPO(0.104) 100 Table 10- 11. Best fit function for daily delays No. Category Name F(x) R2 1 Maintenance -0.001+EXPO(3.77) 91 2 Other -0.001+WEIB(0.127,0.354) 99.8 Table 10- 12. Best fit function for TBF and TTR 231 No. Category Name F(x) R2 1 Cutter_TTR GAMM(2.7,2.13) 72.8 2 Cutter_TBF 5+EXPO(57.7) 97.9 3 TBM_TTR LOGN(2.38,3.85) 97.2 4 TBM_TBF GAMM(0.515,12.4) 79 5 BU_TTR LOGN(1.54,3.63) 99.9 6 BU_TBF 3+LOGN(6.92,9.92) 99.9 0.35 0.45 Data Data 0.40 0.30 Fit Fit 0.35 0.25 0.30 0.20 0.25 0.15 0.20 0.15 0.10 0.10 RelativeFrequency/PDF RelativeFrequency/PDF 0.05 0.05 0.00 0.00 1.9 9.2 5.5 16.6 24.0 12.9 20.3 0.11 0.51 0.25 0.38 0.64 0.77 0.90 Boring Time per Ring (hr) Maintenance Time per Day (hr) 0.90 1.00 0.80 Data 0.90 Data Fit Fit 0.70 0.80 0.70 0.60 0.60 0.50 0.50 0.40 0.40 0.30 0.30 0.20 0.20 RelativeFrequency/PDF RelativeFrequency/PDF 0.10 0.10 0.00 0.00 2.6 7.7 17.8 47.3 76.9 12.8 23.0 17.9 106.0 136.0 BU TTR (hr) Boring Time per Ring (hr) Figure 10- 20. Examples of fitted functions on the data for different categories 0 232 Tr ue Tr ue 0 0 En d False False 0 0 Delay Reset Re c o rd SHN BU De l a y ? Cutter Delay? Delay Maintenance T BF En d Assign ResTS Record Ring As s i g n Su T Next TBM TBF Tr ue 0 Up d a te SHT False Delay Support 0 Tr ue 0 Ring No. Control TBM De l a y Delay Survey False 0 End of SHT? Update Ring No. TTCT Assign SuTS Delay Boring Assign TBMTS T i m e Update SHT and Assign Cycle Tr ue 0 False 0 Assign TOSH TBM TTCT MODEL Up d a te CT TBM De l a y ? Delay Utility Up d a te BT S Delay Other_Begin Assign OthT Next BU TBF Assign UtTS Next Cutter TBF De l a y Maintenance_Begin BU De l a y Delay Other Cutter Delay Assign TTCT Delay Transport 0 Sta rt Assign TrTS Assign BUTS Assign OthTS Assign CutterTS Figure 10- 21. Simulation model in Arena software 233 TB F SAMPLING 0 T ru e Create Cutter TBF Assign Cutter Next Cutter TBF? End Cutter_TBF Loop End Cutter_TBF TBF 0 0 0 F a l s e 0 T ru e Create TBM TBF Assign TBM TBF Next TBM TBF? End TBM_TBF Loop End TBM_TBF 0 0 0 F a l s e 0 T ru e Create BU TBF Assign BU TBF Next BU TBF? End BU_TBF Loop End BU_TBF 0 0 0 F a l s e Figure 10- 22. TBF sampler model Data Analysis For analysis, the tunnel data was split into five categories including one category for the start-up phase (chainage 145-650) and four categories for the production phase (ch. 650-4000, 4000-8000, 8000-12000, 12000-16000). The data of the first production reach was used to simulate TTCT for the subsequent reaches. The corresponding chainage for the end of start-up phase is obtained from curve fitting on the utilization data as shown in Fig. 10-23. 234 35 30 25 20 15 Weekly Utilization (%) 10 5 0 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 Week No. Figure 10- 23. Curve fitting on the utilization data for obtaining the end of the start-up phase Table 10- 13. The results of simulation model for TTCT and utilization (U) prediction for reaches 2 to 4 TTCT (hr) U (%) Reach No. Length (m) PR (m/hr) Real Simulated Real Simulated 1 3413 3.576 5618 5610 17.0 17.0 2 3940.3 4.103 5208 6026 18.5 16.0 3 3922.1 3.797 7705 6572 13.6 16.0 4 3927.3 3.733 7618 6399 13.6 16.2 U (%) TTCT (hr) Reach No. Real Simulated Traditional method Real Simulated Traditional method 2 18.5 16.0 17 5208 6026 5649 3 13.6 16.0 17 7705 6572 6076 4 13.6 16.2 17 7618 6399 6189 Total 20531 18997 17914 Error Percentage 7.5 13 In table 10-13, the results of simulation modeling are compared with the traditional modeling. In traditional modeling, the TTCT is simply obtained from the real penetration rate and assumed utilization factor from the first reach, while in simulation modeling every activity uses 235 its own time distribution. As can be seen in table 10-13, the results of the simulation model have a lower error. Second Simulation Approach Figure 10-24 to 10-26 show the operation cycles for three different hard-rock TBMs. The process of excavation by a TBM is performed in a cyclic process. It starts with breaking the rock at the heading, followed by collecting and transfer of the broken rock materials out of tunnel. In addition to these two basic operations, there are some other necessary operations for completing an excavation cycle that are performed in series or parallel with the excavation operation. One important issue in downtime analysis is the overlap between the different downtime categories for parallel activities. . 236 Shift change and/or Safety meeting Time Daily TBM Daily cutter Daily Back-Up maintenance inspection/change maintenance Exploration Boring Mucking Rock support Regripping Boring Mucking Rock support Moving full/empty muck trains Movingtrains full/empty muck along the tunnel tunnel Regripping of outsideMuckthe unloading Train exchange Line Utility Completing rock support extension Figure 10- 24. Typical Open TBM operation cycle 237 Shift change and/or Safety meeting Daily TBM Daily cutter Daily Back-Up maintenance inspection/change maintenance Time Boring Mucking Rock support Boring Mucking along the tunnel of outsideMuckthe unloading tunnel Rock support Movingtrains full/empty muck Train exchange Completing rock support (Segment Line extension erection/Backfilling) Figure 10- 25. Typical single shield (SS) TBM operation cycle 238 Shift change and/or Safety meeting Daily TBM Daily cutter Daily Back-Up maintenance inspection/change maintenance Time Boring Mucking Rock support Regripping outside of the of outsidethe Boring Mucking Rock support Moving full/empty muck trains Movingtrains full/empty muck along the tunnel tunnel Regripping Muckunloading Train exchange Completing rock support (Segment Line extension erection/Backfilling) Figure 10- 26. Typical double shield (DS) TBM operation cycle 239 Simulation techniques have the great advantage over other methods to account for the parallel versus linear activities. Fig. 10-27 to Fig. 10-29 show developed models for different cases of having a different number of California switches along a tunnel with a double shield TBM. The different stages of the first model are simplified as follows: 1. A train enters the tunnel, controls the tunneling timeline, and incurs a delay to reach the TBM. 2. A delay is added to the system for the beginning of the shift, in change over of the crew. 3. The TBM controls the timeline and the boring cycle starts. At the same time, ground supporting installation (ring erection) activities are ongoing in parallel. 4. The first cycle ends either by the end-of-course boring or by the end of supporting activities for that course, whichever takes longer. 5. The next cycle is repeated the same as the previous one, while the timeline of the train travelling to the portal is followed and if the trains incurs any delay, the cycle is adjusted accordingly. 6. For the last cycle, as soon as boring is ended, the train goes out, but at the same time, the ground support activities continue. 7. The train goes out and reaches the dump station. 8. The next train goes inside and repeats the cycle. 9. The failures of each sub-system are applied according to the TBF and TTR values. 10. A maintenance delay is applied when it is at the beginning of the day. The cycles for the other models are almost the same with additional delays of train exchanging over the switches. 240 Train In Se i z e De l a y Sta rt Assign Times Se i z e Ra i l Delay Begin Se i z e TBM Re s o u rs e s Tra n s In 0 Re l e a s e 0 L o c o Se p a ra te 1 O r iginal 0 Duplicat e De l a y Bo ri n g 1 De l a y 0 Ba tc h 1 Se p a ra te 2 Re g ri p 1 O r iginal Se i z e De l a y Re l e a s e 0 0 Duplicat e Ere c to r1 Ere c to r1 Ere c to r1 De l a y Bo ri n g 2 De l a y 0 Ba tc h 2 Se p a ra te 3 Re g ri p 2 O r iginal 0 0 Duplicat e Se i z e De l a y Re l e a s e Ere c to r2 Ere c to r2 Ere c to r2 Boring Process Boring De l a y Bo ri n g 3 De l a y 0 Ba tc h 3 Se p a ra te 4 Re g ri p 3 O r iginal Se i z e De l a y Re l e a s e Duplicat e Ere c to r3 Ere c to r3 Ere c to r3 0 0 De l a y Re l e a s e Se i z e L o c o Bo ri n g 4 TBMandCutter Se i z e De l a y Re l e a s e De l a y RingNoupdate Di s p o s e 1 Ere c to r4 Ere c to r4 ErectorandBU Re g ri p 4 0 De l a y Delay Portal 0 Re l e a s e Tra n s Ou t Dri v e to T ra i n No Release Rail Se p a ra te 7 Se i z e O r iginal L o c o 1 Du m p Du m p e r 0 Duplicat e Train Out De l a y Re l e a s e Du m p i n g Du m p e r 0 De l a y Tr ue Seize Loco1 De c i d e 2 Seize Train1 LocoShunting Release Previous Train 0 False Station Dumping 0 Tr ue Re l e a s e De c i d e 1 Tra i n 2 Seize Train2 TimesUpdate Sieze Next Train Di s p o s e 2 0 False Re l e a s e 0 Tra i n 1 0 Tr ue De l a y De c i d e 3 As s i g n T s h Maintenance M i s 0 False Delay Maintenance/Mis Figure 10- 27. Simulation model of tunneling by double shield TBM using two trains without any California switch (case 1) 241 Train In 0 Se i z e De l a y Re l e a s e Se p a ra te 8 Sta rt Assign Times Se i z e CS1 De l a y CS Re l e a s e CS O r iginal Re s o u rs e s Se i z e Ra i l Tra n s In Ra i l 1 0 0 Duplicat e De l a y 0 Se i z e TBM Se p a ra te 1 Be g i n a n d Afte rCS O r iginal 0 Duplicat e Se i z e De l a y Re l e a s e Re l e a s e Seize Rail1 Se i z e CS1 1 De l a y CS1 Tra i n L o c o 2 Tra n s In 1 Ra i l 1 1 CS1 California Switch California De l a y Bo ri n g 1 De l a y 0 Ba tc h 1 Se p a ra te 2 Re g ri p 1 O r iginal Se i z e De l a y Re l e a s e 0 0 Duplicat e Ere c to r1 Ere c to r1 Ere c to r1 De l a y Bo ri n g 2 De l a y 0 Ba tc h 2 Se p a ra te 3 Re g ri p 2 O r iginal Se i z e De l a y Re l e a s e 0 0 Duplicat e Ere c to r2 Ere c to r2 Ere c to r2 De l a y Bo ri n g 3 Boring Process Boring Se i z e De l a y 0 Se p a ra te 4 Ere c to r3 Ere c to r3 De l a y O r iginal Ba tc h 3 Re g ri p 3 Re l e a s e 0 Duplicat e Ere c to r3 0 De l a y Re l e a s e De l a y Re l e a s e Bo ri n g 4 TBMandCutter Se i z e CS2 De l a y CS2 Seize Rail2 1 0 0 m CS2 Se i z e De l a y Re l e a s e De l a y RingNoupdate Di s p o s e 1 Ere c to r4 Ere c to r4 ErectorandBU Re g ri p 4 0 0 Re l e a s e Tr ue De c i d e 4 As s i g n 6 Se i z e 0 De l a y Train Out Ra i l Se p a ra te 7 Du m p e r O r iginal Du m p i n g 0 False 0 Duplicat e De l a y De l a y Portal Drive T ra i n No Tra n s Ou t to Du m p Dumping Station Dumping Re l e a s e Re l e a s e Tra i n 2 0 Du m p e r De c i d e 1 Tr ue Re l e a s e De c i d e 7 L o c o 1 Next Tr ainNo==1 Next Tr ainNo==2 Else Re l e a s e Tra i n 3 0 False Di s p o s e 2 Re l e a s e L o c o 2 0 Re l e a s e Tra i n 1 0 Tr ue Se i z e De c i d e 2 L o c o 1 0 Release Previous Train Tr ue De c i d e 3 De l a y 0 False LocoShunting 0 False Se i z e TimesUpdate L o c o 2 Se i z e De l a y Tra i n 1 As s i g n T s h Maintenance M i s De c i d e 5 Next Tr ainNo==1 Next Tr ainNo==2 Se i z e Else Tra i n 2 Delay Maintenance/Mis Se i z e Tra i n 3 Sieze Next Train Figure 10- 28. Simulation model of tunneling by double shield TBM using three trains with one California switch (case 2) 242 S e i z e T o Start Assign S t a rt Variables 0 Train In-Rail 3 0 Separat e 1 O r igina l 0 D e l a y Tr u e 0 0 Du p lica t e TimesUpdat e D e c i d e 7 Assign Tsh Maintenance S e i z e T B M Delay Rail3 In Separat e 5 M i s Release CS22 O r igina l 0 Du p lica t e 0 Fa lse Delay Boring1 0 Bat ch 1 Delay Regrip1 Separat e 2 O r igina l S e i z e R e l e a s e D e l a y 0 0 Du p lica t e E re c t o r1 E re c t o r1 E re c t o r1 Delay Boring2 0 Bat ch 2 Delay Regrip2 Separat e 3 O r igina l S e i z e D e l a y R e l e a s e 0 0 Du p lica t e E re c t o r2 E re c t o r2 E re c t o r2 Delay Boring3 0 Process Boring Bat ch 3 Delay Regrip3 Separat e 4 O r igina l S e i z e D e l a y R e l e a s e 0 0 Du p lica t e E re c t o r3 E re c t o r3 E re c t o r3 R e l e a s e Delay Rail3 Delay Boring4 S e i z e C S 2 D e l a y C S 2 Seize Rail2 Release CS2 TBM andCutter Ou t S e i z e D e l a y R e l e a s e Delay Regrip4 RingNoupdat e Dispose 1 E re c t o r4 E re c t o r4 ErectorandBU 0 Train Out-Rail 1-2 0 Separat e 6 Delay Rail2 Release CS1 S e i z e C S 1 Release Rail2 D e l a y C S 1 Seize Rail1 O r igina l Ou t 0 Du p lica t e Delay Portal 0 Delay Rail1 S e i z e Tr u e D ri v e to Release Rail1 TrainLocoNo D e c i d e 1 Ou t D u m p e r D u m p Assign 4 0 Fa lse 0 Tr u e D e c i d e 2 Assign 5 R e l e a s e L o c o 1 0 Fa lse R e l e a s e L o c o 2 0 D e l a y R e l e a s e D e c i d e 3 Separat e 7 O r igina l D u m p i n g D u m p e r Ne x t L o c o No= = 1 Ne x t L o c o No= = 2 0 Du p lica t e Station Dumping Else R e l e a s e L o c o 3 R e l e a s e D e c i d e 4 T ra i n 2 Ne x t Tr a inNo= = 1 Release Previous Loco Previous Release Ne x t Tr a inNo= = 2 Ne x t Tr a inNo= = 3 Else R e l e a s e T ra i n 3 Seize Loco1 Dispose 4 D e l a y R e l e a s e D e c i d e 5 0 LocoShunting T ra i n 4 Ne x t L o c o No= = 1 Ne x t L o c o No= = 2 Else R e l e a s e Seize Loco2 T ra i n 1 Release Previous Train Previous Release Seize Train1 Seize Loco3 D e c i d e 6 Ne x t Tr a inNo= = 1 Ne x t Tr a inNo= = 2 Seize Train2 Ne x t Tr a inNo= = 3 Sieze Next Loco Else Seize Train3 Seize Train4 Train Next Sieze Train In-Rail 1-2 Release Rail1 Release CS1 Seize Rail1 In Delay Rail1 In Seize CS1 In Delay CS1 In Seize Rail2 In Delay Rail2 In In In Release Rail2 Seize CS2 In In Delay CS2 In Figure 10- 29. Simulation model of tunneling by double shield TBM using four trains with two California switches (case 3) 243 Case 1 Data Analysis The model for this case was run for 2625 rings of support (3412 m) and the results of a typical simulation are presented in Table 10-14 and Fig. 10-30. The simulated time and TBM utilization value for this case are 5610 and 46%, respectively. These values are in good agreement with the real values (5592 and 0.45). Table 10- 14. Results of simulation model for case 1 Resource Instant Utilization Number busy Number Scheduled Number Seized BU 0.49 0.49 1 1313 Cutter 0.44 0.44 1 1313 Dumper 0.10 0.10 1 1313 Erector 0.47 0.47 1 5252 Loco 0.46 0.46 1 2626 Rail 0.98 0.98 1 1313 TBM 0.44 0.44 1 1313 Train 1 0.43 0.43 1 657 Train 2 0.45 0.45 1 656 5 Sample Size=20 Mean=5610 hr COV=1.9% 4 3 Count 2 1 0 5500 5550 5600 5750 5800 5850 5450 5650 5700 ------ 5450 5500 5550 5700 5750 5800 5400 5600 5650 Time to Complete Tunnel (hr) Figure 10- 30. Simulation model results of tunneling by double shield TBM using two trains without any California switch (case 1) 244 Case 2 Data Analysis The model for this case was run for 2283 rings of support (2968 m) using three trains and one intermediate California switch, and the results of a typical simulation are presented in Table 10-15 and Fig. 10-31. The simulated time and TBM utilization value for this case are 3590 and 0.40. These values are in good agreement with the real values (3600 and 0.42). Table 10- 15. Results of simulation model for case 2 Resource Instant Utilization Number busy Number Scheduled Number Seized BU 0.44 0.44 1 1143 CS1 0.02 0.02 2 2285 Cutter 0.44 0.44 1 1143 Dumper 0.13 0.13 1 1141 Erector 0.37 0.37 1 4568 Loco1 0.99 0.99 1 572 Loco2 0.99 0.99 1 571 Rail1 0.33 0.33 1 2284 TBM 0.40 0.40 1 1143 Train 1 0.44 0.44 1 381 Train2 0.52 0.52 1 381 Train3 0.44 0.44 1 381 6 Sample Size=20 5 Mean=3590 hr 4 COV=2.2% 3 Count 2 1 0 3400-3440 3440-3480 3480-3530 3530-3580 3580-3630 3630-3680 3680-3730 3730-3780 3780-3830 Time to Complete Tunnel (hr) Figure 10- 31. Simulation model results of tunneling by double shield TBM using three trains with one California switch (case 2) 245 Case 3 Data Analysis The model for this case was run for 4109 rings of support (5342 m) with four trains and two California switches and the results of a typical simulation are presented in Table 10-16 and Fig. 10-32. The simulated time and TBM utilization value for this case are 7870 and 0.33. These values are in good agreement with the real values (7920 and 0.30). Table 10- 16. Results of simulation model for case 3 Resource Instant Utilization Number busy Number Scheduled Number Seized BU 0.33 0.33 1 2055 CS1 0.03 0.03 2 4110 CS2 0.78 0.78 2 4110 Cutter 0.35 0.35 1 2055 Dumper 0.10 0.10 1 2054 Erector 0.31 0.31 1 8220 Loco1 0.99 0.99 1 686 Loco2 0.99 0.99 1 686 Loco3 0.99 0.99 1 686 Rail1 0.24 0.24 1 4110 Rail2 0.57 0.57 1 4110 TBM 0.33 0.33 1 2055 Train 1 0.77 0.77 1 514 Train2 0.77 0.77 1 514 Train3 0.77 0.77 1 515 Train4 0.77 0.77 1 514 6 Sample Size=20 Mean=7870 hr COV=1% 5 4 3 Count 2 1 0 7690-7730 7730-7770 7770-7810 7810-7850 7850-7890 7890-7930 7930-7970 7970-8010 Time to Complete Tunnel (hr) Figure 10- 32. Simulation model results of tunneling by double shield TBM using four trains with two California switches (case 3) 246 Discussions In this chapter, the main methods for simulation of TBM performance were reviewed. Two new approaches of simulation of the TBM operation were proposed based on detailed activity time distributions and rail-bound transportation schemes. New formulas were proposed to find the final location of California switches for different scenarios of having a different number of trains. The models are also calibrated for different numbers of trains on the basis of detailed information of a recent tunneling project using a double shield machine in a medium-strength rock. 247 Chapter 11 Conclusions and Recommendations Conclusions In this thesis some new empirical models and formulas were derived for the purpose of the prediction of TBM utilization, downtime components, and advance rate on the basis of gathered field data from several tunnel projects from around the world. The primary goal of this study was to develop new models/formulas which can be used in the calculation of the time to complete tunnel (TTCT) and cost to complete tunnel (CTCT). For this purpose, a comprehensive database of TBM field performance was compiled to allow for development of new formulas. The proposed models could improve previous TBM performance prediction models. The database of TBM field performance includes 300 tunnel projects and pertinent information including rock properties, TBM specifications, and TBM operational parameters. Statistical analyses were used to seek correlations between rock/ground characteristics and TBM specifications / operational parameters and the resulting TBM performance. For some tunnel cases, the data was collected directly on site and the detailed TBM performance information was available, while for the other records, the information was limited to average values over geological zones or the entire tunnel length. Some detailed databases were also compiled from the information shared graciously by some researchers and contractors. To make a consistent database for all of the tunnel cases, the detailed data were averaged over geological zones and were added to the general data for other tunnel cases. After organizing the data in a spreadsheet, linear regression analysis was performed to establish correlations and 248 formulas for target and independent parameters. This includes relationships among tunnel specifications, rock properties, and TBM operational parameters. One of the most important problems in the compiled databases is the heterogeneity of the data sets. Despite this problem, linear regression analysis including stepwise or best sub-set methods were used to develop the correlations among various parameters. Many individual multiple regression analyses were performed to reach the best set of formulas and, a subset of the data was used for verification of the derived formulas. Although, the results of the primary modeling on training subset showed R-sq of more than 60%, the results of the verification phase on the testing subset were not satisfactory. Many trials were performed on the testing subset to eliminate the bias with no improvement. For this reason, a different methodology was used to develop a new rating system to increase the efficiency of the models. This led to driving the best models using common geological and machine parameters which could offer more accurate results. The results of this study provide better understanding of the trends of TBM performance parameters of different tunnels in different conditions. The new formulas can be used in the process of planning of a tunnel project during the design phase. The following summarizes the most important findings of this study: Comparisons between predicted and actual TBM penetration indicated that most of the existing predictive models, especially the simple ones, cannot offer accurate estimates of TBM performance for new projects. A study of these models indicates that using more parameters in a model will not always guarantee improved results due to the lack of inherent limits of the initial models. The penetration rate analyses of the available data indicate that a certain set of available parameters including tunnel diameter, UCS, RPM, and rock type, account for approximately half of the PR and PRev variation. In order to best evaluate the penetration rate, the field penetration index (FPI) was finally used as the response 249 variable. The analysis results show that UCS, CAI, and RQD are the most influential parameters in FPI evaluation. The predicted utilization values of the existing models of EMI and NTH do not have good agreement with the real data compiled in the database and they often overestimate the machine performance. The new utilization model is proposed to improve the previous models, prediction capabilities. The new TBM utilization model distinguishes the machine type and accounts for cutter change, ground support, water inflow and TBM/Backup downtimes. Results of the cutter downtime analysis show that, on average, cutter change time is 66 minutes and there is a strong relationship between cutter change time and boring time. While it is anticipated that cutter size might have an inverse relationship with cutter change downtime, the result of the analysis of data from a few projects showed that the differences of cutter change downtime for different cutter sizes are not very high. Although rock abrasiveness is recognized as the most important parameter in cutter wear prediction, the performed analysis does not show a strong correlation between CAI and cutter consumption. This might be related to the fact that the average CAI values were considered for the whole length of tunnel drives. However, the results show that there is a need for more in-depth study on the application of CAI in prediction of cutter consumption. The results of the statistical analysis show that the impacts of cutter size or penetration per revolution are less than the effects of rock type, UCS, and quartz content for cutter life prediction. 250 Results of UST analysis for case histories using small open type TBMs show that if the Austrian rock mass classification is used, reasonable estimates for ground support-related downtime can be obtained. While more competent rocks (RMR>70) rarely need any ground support, the amount of required support will increase with decreasing RMR and related downtime can be estimated. However, support installation time in stretches of very poor rock (F7 in Austrian classification) might need more than 40 hr/m supporting time and is difficult to estimate since it involves extreme measures including the use of ground improvement methods which are unique to each project. Analysis of SIT shows that when shotcrete is applied near the tunnel face, it can severely hinder excavation production, taking 40 hr/m on average. This is unless shotcrete is applied in larger diameter tunnels with special provisions for its applications in the L1 area. Obviously, this is explains why in application of smaller TBMs, installation of shotcrete immediately behind the head is not required and shotcrete is only applied in L1 area under extreme conditions. The results of support downtime analyses show that rock bolting is the fastest ground support method in TBM tunneling. Use of rock bolts in more competent rock coincides with their low impact on machine utilization. Comparisons between TBM utilization as predicted by the currently available models and the actual field data indicate that existing predictive models cannot offer accurate estimates for new projects. In some cases, these models give extremely high values for the utilization factor. Overall, it is noted that the existing models tend to overestimate advance rate The results of analyses indicate that PR, UCS, water condition, and tunnel diameter are the most influential parameters for utilization factor evaluation. 251 The results of the analyses indicate that tunnel diameter and UCS are among the primary parameters in advance rate prediction. The division of the database for different tunnel diameters yielded different trends indicating different behaviors for different tunnel sizes. This might be due to different interactions amongst the various tunneling categories, such as tunnel haulage and ventilation restrictions, available working space, and TBM technical specifications. A new methodology for estimation of the Learning Period Performance (LPP) was introduced. A linear function is used to obtain the LPP parameters for simplicity and improved accuracy, especially when an estimate of anticipated advance rates during the normal period are available. Simulations techniques can be very useful in terms of accounting for parallel activities and different probable scenarios. In this regard, new models are generated to account for these issues and to improve the accuracy of the prediction. The results of simulation modeling using new models show a very good agreement with the actual values. Recommendations Based on the findings and conclusions of this study, the following recommendations are offered for further study of performance prediction of hard-rock TBMs in the future. In the current study, many attempts were made to correlate between TBM performance parameters and rock mass properties; however, the results of linear regression analysis either for bivariate or multivariate regressions do not indicate strong correlations with geological variability or CFF. The reason for lack of success could be wide ranges of different categories for characterization of ground conditions. To improve the results, a close collaboration of many parties including 252 researchers, engineers, and contractors is needed to compile a more detailed and more reliable database from different tunnel cases around the world to be used as a basis to adjust the proposed models for given ground conditions. Certainly, the results of this study would help understanding the most important parameters which need more focus for data compilation. Analyses of existing TBM utilization and AR prediction models indicate that some of them tend to overestimate TBM performance. Some of these models have limited variability for different conditions. The results of the analysis indicate that improved accuracy of the models depends on their ability to account for additional machine and ground properties. It is recommended that a universal recording system be used for TBM back-mapping to record ground conditions. This can lead to generation of a consistent set of data and a uniform database which can be used in the future studies. The same is applicable for recoding of the downtime, which would allow for more efficient analysis of downtime among different projects in future. A time study of the various activities is warranted to distinguish the difference between activity time associated with various TBM types, sizes, and tunnel lengths. One of the most important findings of this research is that rock type would be one of the best parameters that can be used for distinction of various job sites, and it could contribute in development of more accurate performance estimation formulas. There is a need to investigate this parameter in more detail, either by updating the rock type categorization or by averaging the performance of TBMs on the basis of different rock types. Further study would be needed to come up with a cutter wear and cutter consumption evaluation using more detailed data and to account for the impact of rock structure on cutter consumption/life. 253 An improvement on the future downtime analysis can be useful for estimation of project completion time. It offers more flexibility for separating different time items. 254 REFERENCES Askilsrud, O.G., 1996. Hard Rock Tunnel Boring Machine-Mechanical Tunneling, Short Course in Colorado School of Mines, Atlas Copco Robbins Inc., Colorado, USA. Abd Al-Jalil, Y., 1998. Analysis of performance of tunnel boring machine-based systems, The University of Texas at Austin, 427 p. Alber, M., 2000. Advance rates of hard rock TBMs and their effects on projects economics, Tunnelling and Underground Space Technology 15 (1) (2000), pp. 55–84. Alvarez Grima, M., P.A., 2000a. Bruines and P.N.W. Verhoef, Modeling tunnel boring machine performance by neuro-fuzzy methods, Tunnell. Undergr. Space Technol. 15 (3), pp. 259– 269. Alvarez Grima, M., 2000b. Neuro-fuzzy modeling in Engineering Geology: applications to mechanical rock excavation, rock strength estimation and geological mapping, PhD.Thesis, Delft University of Technology, 252 pp, Balkema, Rotterdam. arez rima M ab a, R., 1999. Fuzzy model for the prediction of unconfined compressive strength of rock samples, Int. J. Rock Mech. and Min. Sci 36 (3), pp. 339– 349. Alvarez Grima, M., P.A. Bruines, P.A., and P.N.W. Verhoef, P.N.W., 2000a. Modeling tunnel boring machine performance by neuro-fuzzy methods, Tunnell. Undergr. Space Technol. 15 (3), pp. 259–269. Atlas Copco, 2005. Rapid Support Behind the TBM, Rock and Soil Reinforcement, 3rd Edition, pp. 61-64. 255 Bamford, W.F., 1984. Rock test indices are being successfully correlated with tunnel boring machine performance. In: Proceedings of the 5th Australian Tunneling Conference, Melbourne, vol. 2, pp. 9–22. Barla, G., Pelizza, S., 2000. TBM tunneling in difficult ground conditions, Proceedings of GeoEng 2000, Proceedings of the International Conference on Geotechnical & Geological Engineering, Melbourne, November 19–24, 2000, Technomic Publishing Company, Lancaster, pp. 329–354. Barton, N., 2000. TBM tunneling in jointed and faulted rock. A.A. Balkema Publishers, Rotterdam. Barton, N., 1999. TBM performance in rock using QTBM, Tunnels Tunnell. Int. 31, pp. 41–48. Barton, N., 2000. TBM tunneling in jointed and faulted rock. A.A. Balkema Publishers, Rotterdam. Barton, N., 2011. TBM prognosis for open-gripper and double-shield machines tunneling through hard jointed rock with weakness zones, Proceedings of World Tunneling Congress (WTC), May 20-26, Helsinki, Finland. Beyer, F., 1982. Zum mittleren Kluftabstand aus der Anzahl von Kluftanschnittlinien, Rock Mechanics, No. 14, pp. 235-251. Bieniawski, Z. T., Celada, B., Galera, J.M., Tardáguila, I., 2009. Prediction of cutter wear using RME, ITA-AITES World Tunnel Congress, Budapest. Bieniawski, Z.T., 1989. Engineering rock mass classifications. John Wiley & Sons, New York, 251 pp. Bieniawski, Z.T., Celada B, Galera, JM., 2007a. TBM excavability: prediction and machine– rock interaction. In: Proceedings of the Rapid Excavation and Tunneling Conference (RETC), Toronto, Canada, pp 1118–1130. 256 Bieniawski, Z.T., Celada, B., Galera Fernandez, J.M., 2007b. Predicting TBM Excavability- Part I. Tunnels & Tunneling International, p. 25. Bieniawski, Z.T., Celada, B., Galera, J.M., Álvares, M., 2006. Rock Mass Excavability (RME) index: a New Way to Selecting the Optimum Tunnel Construction Method, Proc. ITA World Tunneling Congress, Seoul. Bieniawski, Z.T., Celada, B., Galera, J.M., Tardáguila, I., 2008. New applications of the excavability index for selection of TBM types and predicting their performance, ITA World Tunneling Congress, Agra, India. Bieniawski, Z.T., Grandori R., 2007c. Predicting TBM Excavability-Part II. Tunnels & Tunneling International, pp. 15-18. Boyd, R.J., 1986. Hard rock continuous mining machine: Mobile Miner MM-120. In: D.F. Howarth et al., Editors, Rock excavation engineering seminar, Dept. Mining and Met. Eng, University of Queensland. Bruland, A., 1998a. Hard rock tunnel boring. Ph.D. Thesis, The Norwegian University of Science and Technology, Trondheim. Bruland, A., 1998b. Prediction Model for Performance and Costs, The Norwegian University of Science and Technology, Trondheim, Norwegian Tunneling society, NFF, Publication No. 11, pp. 29-34. Büchi, E., 1984. Influence of geological parameters on the propulsion achievement of a tunnel drilling machine. Thesis, University of Berne. Büchi, E., 1993. 20 in cutters cost analysis for Klippen project and cost estimate for future projects. Report No. A 9319, Robbins Co., GEOTEST. Cassinelli, F., Cina, S., Innaurato, N., Mancini, R., Sampaolo, A., 1982. Power consumption and metal wear in tunnel-boring machines: analysis of tunnel-boring operation in hard rock. Tunneling ‘82, London, Inst. Min. Metall., pp. 73–81. 257 Cheema, S., 1999. Development of a rock mass boreability index for the performance of tunnel boring machines. Doctoral dissertation, Department of Mining Engineering, Colorado School of Mines, Golden, CO, USA. Einstein, H.H., Dudt, J.P., Halabe, V.B., Descoeudres, F., 1992. Decision aided tunneling, principals and practical application, Report prepared in conjunction with the research project for the Swiss Federal office of transportation, November. Farmer, I.W., Glossop, N.H., 1980. Mechanics of disc cutter penetration, Tunnels Tunnell. Int. 12 (6), pp. 22–25. Farrokh, E., Mortazavi, A., Shamsi, Gh., 2006. Evaluation of ground convergence and squeezing potential in the TBM driven Ghomroud tunnel project , Tunnelling and Underground Space Technology, Volume 21, Issue 5, pp. 504-510. Farrokh, E., Rostami, J., 2007. The relationship between tunnel convergence and TBM operational parameters and chip size for double shield TBMs, RETC, Canada, pp. 1094– 1108 (Chapter 88). Farrokh, E., Rostami, J., 2008. Correlation of tunnel convergence with TBM operational parameters and chip size in the Ghomroud tunnel, Iran. Tunneling and Underground Space Technology, 23 6, pp. 700–710. Farrokh, E., Rostami, J., 2009. Effect of adverse geological condition on TBM operation in Ghomroud tunnel conveyance project, Tunnelling and Underground Space Technology, Vol. 24, Issue 4, pp. 436-446. Farrokh, E., Rostami, J., Laughton, C., 2011. Analysis of unit supporting time and support installation time for Open TBMs, Rock Mechanics and Rock Engineering, Volume 44, Issue 4, pp.431-445. Frenzel, Ch., Käsling, H., Thuro, K., 2008. Factors Influencing Disc Cutter Wear. Geomechanik und Tunnelbau 1, pp. 55-60. 258 GEOTEST. 1993. Job Site Report on Alassio Tunnel, Italy. Goel, R. K., 2008. Evaluation of TBM performance in a Himalayan tunnel, World Tunnel Congress 2008 - Underground Facilities for Better Environment and Safety – India, 1522-1532. Gong, Q.M., 2005. Development of a rock mass characteristics model for TBM penetration rate prediction, PhD thesis, Nanyang Technology University. Gotthard TBM safely across the Piora Mulda. 2008, Oct 2010). Graham, P.C., 1976. Rock exploration for machine manufacturers. In: Z.T. Bieniawski, Editor, Exploration for Rock Engineering, Balkema, Johannesburg, pp. 173–180. Grandori, R., 1996a. Avanzamento meccanico in condizioni estreme. Scelta del tipo di TBM e sue caratteristiche, 3° Symp. Européen de la construction des tunnels, Berna, 21-22 Mars, pp. 50-57. Grandori, R., 1996b. The universal TBM in the year 2000. Technical aspects and contractor consideration. Gallerie e grandi opere sotterranee, SIG, No.50, November, 38-48. Grandori, R., 2007. TBM Performances and RME Classification System. Jornada sobre 'Experiencias recientes con tuneladoras'. CEDEX, Madrid. Grandori, R., Bieniawski, Z.T., Vizzino, D., Lizzadro, L., Romualdi, P., Busillo, A., 2011. Hard Rock Extreme Conditions in the First 10 km of TBM-Driven Brenner Exploratory Tunnel, Rapid Excavation and Tunneling Conference (RETC), San Francisco, California, pp. 667–685. Grandori, R., Jaeger, M., Antonini, F., Vigl, L., 1995. Evino-Mornos Tunnel Grece. Construction of a 30 km long hydraulic tunnel, in less than three years under the most adverse geological conditions. RETC, San Francisco, 18-21 June, pp. 747-767. 259 Grimstad, E., and Barton, N., 1993. Updating the Q-System for NMT. Proc. int. symp. on sprayed concrete - modern use of wet mix sprayed concrete for underground support, Fagernes. 46-66. Oslo: Norwegian Concrete Assn. Gripper TBM. Hamilton, W.H., and, Dollinger, G.L., 1979. Optimizing tunnel boring machine and cutter design for greater boreability, RETC Proceedings, Atlanta 1, pp. 280–296. Hassanpour J., Rostami J., Zhao J., 2011. A new hard rock TBM performance prediction model for project planning, Tunneling and Underground Space Technology, Elsevier, pp. 11. Hassanpour, J., 2009. Investigation of the effect of engineering geological parameters on TBM performance and modifications to existing prediction models. Ph.D. Thesis, Tarbiat Modares University, Tehran, Iran. Hassanpour, J., 2009a. Investigation of the effect of engineering geological parameters on TBM performance and modifications to existing prediction models. Ph.D. Thesis, Tarbiat Modares University, Tehran, Iran. Hassanpour, J., Rostami, J., Khamehchiyan, M., Bruland, A., 2009b. Development new equations for performance prediction. Geo Mechanics and Geoengineering: An International Journal 4 (4), 287-297. Hassanpour, J., Rostami, J., Khamehchiyan, M., Bruland, A., and Tavakoli, H.R., 2009a. TBM performance analysis in pyroclastic rocks: a case history of Karaj water conveyance tunnel, Rock Mech. Rock Eng. Hassanpour, J., Rostami, J., Khamehchiyan, M., Bruland, A., Tavakoli, H.R., 2009b. TBM performance analysis in pyroclastic rocks: a case history of Karaj water conveyance tunnel, Rock Mech. Rock Eng. Herrenknecht web 260 Hoek, E., and Brown, E.T., 1980. Underground Excavations in Rock. London: Instn Min. Metall., 527 pages. Home, L., 2010. TBM tunneling in difficult rock, latest developments, ITACET Foundation short course on Tunneling, Vancouver, BC, May 14-15. Howarth, D.F., 1986. Review of rock drillability and boreability assessment methods, Trans IMM Sect A Min Ind 9, The Institution of Mining and Metallurgy, London, pp. A191– A202. Hughes, H.M., 1986. The relative cuttability of coal-measures stone, Min. Sci. Technol. 3 (2), pp. 95–109. Innaurato, N., Mancini, R., Rondena, E., Zaninetti, A., 1991. Forecasting and effective TBM performances in a rapid excavation of a tunnel in Italy. In: Proceedings of the Seventh International Congress ISRM, Aachen, pp. 1009–1014. ITA/AITES WG. 14, 2000. Recommendations and Guidelines for Tunnel Boring Machine, Lausanne. Johannessen, O., 1994. NTH Hard Rock Tunnel Boring. Project Report 1–94, NTH/NTNU Trondheim, Norway. Johannessen, O., 1988. NTH Hard Rock Tunnel Boring. Project Report 1–88, NTH/NTNU Trondheim, Norway, 1988. Johannessen, O., 1994. NTH Hard Rock Tunnel Boring. Project Report 1–94, NTH/NTNU Trondheim, Norway. Khademi Hamidi, J., Shahriar, K., Rezai, B., Rostami J., 2010. Performance prediction of hard rock TBM using Rock Mass Rating (RMR) system, Tunnell. Undergr. Space Technol. 25 (4), pp. 333-345. Khademi Hamidi, J., Shahriar, K., Rezai, B., Bejari, H., 2009. Application of Fuzzy Set Theory to Rock Engineering Classification Systems: An Illustration of the Rock Mass 261 Excavability Index, Rock Mechanics and Rock Engineering, Volume 43, Number 3, pp. 335-350. Kim, T., 2004. Development of a Fuzzy Logic Based Utilization Predictor Model for Hard Rock Tunnel Boring Machines. PhD Thesis Dissertation, Colorado School of Mines. Kim, T., 2010, Private communication and data exchange. Klein, S., Schmoll, M., Avery, T., 1995. TBM performance at four hard rock tunnels in California. In: Proceedings of the Rapid Excavation & Tunnelling Conference, pp. 61–75 (Chapter 4). Kwiatkowski, Thomas P., Bhore, Joginder S., Velez, Jose, Orlandi, Christopher D., 2007. Successful Tunneling for Water Tunnel No. 3, New York City, New York, Rapid Excavation and Tunneling Conference Proceedings, Chapter 89 - TBM Case Studies, pp. 1109-1117. Laughton, C., 1998. Evaluation and Prediction of Tunnel Boring Machine Performance in Variable Rock Masses. Ph.D. Dissertation, The University of Texas at Austin, 315 pages. Laughton, C., 1998. Evaluation and prediction of tunnel boring machine performance in variable rock masses. Ph.D. Thesis, The University of Texas, Austin, USA. Laughton, C., 2005. Geotechnical Problems Encountered by Hard Rock Tunnel Boring Machines. MPES, 17 pages. Laughton, C., Nelson, P.P., Abd Al-Jalil Y., 1995. The probabilistic prediction of tunnel boring machine performance in practice. NSF Workshop on Transferring Research into Practice, ASCE Press. Lislerud et al., 1983. Hard rock tunnel boring, Project report 183, Norwegian Institute of Technology, Trondheim. Maidl, B., Schmid, L., Ritz, W., Herrenknecht, M., 2008. Hard Rock Tunnel Boring Machines. 347 pages. 262 Martin, D., 1988. TBM Tunnelling in Poor and Very Poor Rock Conditions. Tunnels and Tunneling, Vol. 20, No. 3, pp. 22–27. Minitab Inc., 2010. Minitab 16 Software. Morgan, J. M., Barrat, D. A., Hudson, J. A., 1979. Tunnel boring machine performance and ground properties: Report on the initial 1.5 km of the North Wear Drive, Kielder Aqueduct. Transport and Road Research Laboratory, UK, technical report. Movinkel, T., Johannessen, O., 1986. Geological parameters for hard rock tunnel boring. Tunnels Tunneling 4: 45-48. Nelson, P. P., Abd Al-Jalil, Y., Laughton, C., 1994a. Tunnel boring machine project data bases and construction simulation, Geotechnical Engineering Center Report GR 94-4. The University of Texas at Austin, 1994a. Nelson, P. P., Abd Al-Jalil, Y., Laughton, C., 1994b. Probabilistic Estimates of Time to Complete Tunnel for Six Inland Feeder Project Tunnel Drives using Level 1 Data Base Searching and Simulation Methodologies. Report to the Metropolitan Water District. Geotechnical Engineering Center, University of Texas at Austin, TX. Ne son P O’Ro r e T D K hawy F H 1983. Factors affecting TBM penetration rates in sedimentary rocks. In: Proceedings, 24th US Symposium on Rock Mechanics, Texas A&M, College Station, TX, pp. 227–237. Nelson, P.P., Abad-Aljali, Y., and, Laughton, C., 1994. Tunnel Boring Machine project databases and construction simulation, Geotechnical Engineering Report GR94-4, Geotechnical Engineering Center, Department Civil Engineering, The University of Texas at Austin. Nelson, P.P., Abad-Aljali, Y., Laughton, C., 1999. Improved strategies for TBM performance prediction and project management. RETC, pp. 963–979. 263 Ne son P P O’Ro r e T D aser S D 1985. TBM System Downtime – Causes, Frequency and Duration on Six Tunnel Projects. Rapid Excavation and Tunneling Conference Proceedings, pp. 751-770. Ozdemir, L., Miller, R., Wang, F.D., 1978. Mechanical Tunnel Boring Prediction and Machine Design. Final Project Report to NSF APR73-07776-A03, Colorado School of Mines, Golden, Co. Peila, D., Pelizza, S., 2009. Ground Probing and treatment in Rock TBM tunnel to overcome limiting conditions, Journal of Mining Science, pp. 18, 2009, Vol. 45, pp. 602-619. Pelizza, S., 1998. Selection of TBMs. Workshop on Selection of Tunneling Methods, World Tunnel Congress '98, Sao Paulo (Brazil). Ramezanzadeh, A., 2005. Performance analysis and development of new models for performance prediction of hard rock TBMs in rock mass, Ph.D. thesis, INSA, Lyon, France, p. 333. Ramezanzadeh, A., Rostami, J., Kastner, R., 2005. Influence of Rock Mass Properties on Performance of Hard rock TBMs. RETC, June 27-29, Seattle, Washington, USA. Redja, G.E., 1992. Principles of risk management and insurance, Harper Collins, 673 p. Ribacchi, R, Lembo-Fazio, A., 2005. Influence of rock mass parameters on the performance of a TBM in a Gneissic formation (Varzo tunnel). Rock Mech Rock Eng., 38 (2):105–127 Robbins Co. Website, 2010. industry/ground-support-for-high-cover-tunnels-what-is-the-best-method> (visited 15 Jan 2011). Robbins, R.J., 1992. Large diameter hard rock boring machines: State-of-the-art and development in view of alpine base tunnels, Felsbau, Vol. 10, No. 2, 56-62. Roby, J., Sandell, T., Kocab, J., Lindbergh, L., 2008. Current state of disc cutter design and development directions. North American Tunneling 2008 proceedings, pp. 36-44. 264 Rostami, J., 1993. Design optimization, performance prediction and economic analysis of tunnel boring machine for the construction of the proposed Yucca Mountain nuclear waste repository. Ms. Thesis, Colorado School of Mines, Golden, Colorado, USA. Rostami, J., 1997. Development of a force estimation model for rock fragmentation with disc cutters through theoretical modeling and physical measurement of crushed zone pressure. Ph.D. Thesis, Colorado School of Mines, Golden, Colorado, USA, p. 249. Rostami, J., 2008. Hard Rock TBM cutterhead modeling for design and performance prediction, Geomechanik und Tunnelbau, Ernst & Sohn, (Austrian Journal of Geotechnical Eng.), pp. 18-28. Rostami, J., Ozdemir, L., 1993. A new model for performance prediction of hard rock TBM. In: Bowerman, L.D., et al. (Eds.), Proceedings, Rapid Excavation and Tunnelling Conference, pp. 793–809. Roxborough, F.F., Phillips H.R., 1975. Rock excavation by disc cutter, Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 12, pp. 361–366. Sanio, H.P., 1985. Prediction of the performance of disc cutters in anisotropic rock, Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 22, pp. 153–161. Sapigni, M., Berti, M., Behtaz, E., Busillo, A., Cardone, G., 2002. TBM performance estimation using rock mass classification, Int. J. Rock Mech. Min. Sci. 39, pp. 771–788. Sato, K., Gong, F., Itakura, K., 1991. Prediction of disc cutter performance using a circular rock cutting ring. In: Proceedings, The first International Mine Mechanization and Automation Symposium, Colorado School of Mines, Golden, Colorado, USA. Schmid, L., 2004. The development of the methodology of the shield tunneling in Switzerland. Geotechnique 27, No. 2, 193-200. Schubert, W., 2010. Rock Excavation by Drill & Blast, Lecture No. 10, Institute for Rock Mechanics and Tunneling, 25 pages. 265 Scolari, F., 1995. Open-face borers in Italian Alps. World Tunnelling 1995:8;361–6. Sharp, W., Ozdemir, L., 1991. Computer Modelling for TBM Performance Prediction and Optimization. Proceedings of the International Symposium on Mine Mechanization and Automation, CSM/USBM, vol. 1(4), pp. 57–66. Snowdon, R.A., Ryley, M.D., Temporal, J., 1982. A study of disc cutting in selected British rocks, Int. J. Rock Mech. Min. Sci. 19, pp. 107–121. Stevenson Garry, W., 1999. Empirical estimates of TBM performance in hard rock; Proc. RETC,. AIME, pp. 993–1006. Sundin, N.O., Wanstedt, S., 1994. A boreability model for TBM’s. In: Nelson, P., Laubach, S.E. (Eds.), Rock Mechanics Models and Measurements Challenges from Industry. Proceedings of the 1st North American Rock Mechanics Symposium, The University of Texas at Austin, Balkema, Rotterdam, pp. 311–318. Swiss Engineers and Architects (SIA), 1993. Norm 198: Underground construction. Takagi, T., Sugeno, M., 1985. Fuzzy identification of systems and its applications to modeling and control, IEEE transactions on systems, man, and cybernetics, vol. 15,no. 1, pp. 116- 132. Tarkoy, P.J., 1973. Predicting TBM penetration rates in selected rock types. In: Proceedings, 9th Canadian Rock Mechanics Symposium, Montreal. Tayfur Altiok, and Benjamin Melamed, 2007. Simulation Modeling and Analysis with ARENA, Elsevior Inc., p. 440. Thuro, K., Plinninger, R.J., 2003. Hard rock tunnel boring, cutting, drilling and blasting: rock parameters for excavatability. In: Proceedings of the 10th ISRM International Congress on Rock Mechanics, Johannesburg, South Africa, 1227-1234. Tunnel Boring Machine < http://www.youtube.com/watch?v=yFHrI3HFMDg> (visited 15 Oct 2010). 266 US Army Corps of Engineers. 1997. Engineering and Design Tunnels and Shafts in Rock, Appendix C: Tunnel Boring Machine Performance-Concepts and Performance Prediction, Washington, D.C. Wachter, R., 2001. Training Effect in Mechanical Tunneling. AITES-ITA, Italy, Vol. II, pp. 501-511. Wais, A., 2002. Possibilities and Limitations of Predicting Tunnel Boring Machine Advance Rates for the Assesment of Tunnel Project Costs, Master Thesis, University of Stuttgart, Institute of Hydraulic Engineering. Wais, A., Wachter, R., 2009. Predicting the Learning Curve in TBM Tunneling. In Book "Tunnel Construction - Contributions from Research and Practice," Germany, Ch. 15, pp. 143-154. Wallis, S., 1993. Old Marathon at Alassio, World Tunneling, May, 47-51. Wallis, S., 2009. McConnell Dowell securing hard-won and award winning successes in New Zealand and Australia. WBI-PRINT 6. 2007. Stability Analysis and Design for Mechanized Tunneling, Prof. Dr.-Ing. W. Wittke Consulting Engineers for Foundation Engineering and Construction in Rock Ltd. Yagiz, S., 2002. Development of rock fracture and brittleness indices to quantifying the effects of rock mass features and toughness in the CSM Model basic penetration for hard rock tunneling machines. PhD Thesis. T-5605, p. 289, Colorado School of Mines, Co., USA. Yagiz, S., 2008. Utilizing rock mass properties for predicting TBM performance in hard rock condition, Tunnell. Undergr. Space Technol. 23 (3), pp. 326–339. 267 APPENDIX A Database Projects Excavated Excavated Name of project Country No. Diameter (m) Length (m) 1 Stanley Canyon tunnel US 3.47 5105 2 Cowles Mountain Tunnel US 3.43 2073 Forks of Butte power tunnel US 3.25 3660 3 Goodwin tunnel US 4.3 4800 4 Grizzly power tunnel US 3.35 3600 5 New York City Water Tunnel US 5.76 8784.95 6 No. 3 New York City Water Tunnel US 7.06 7658.71 7 No. 4 River mountains tunnel No. 2 US 4.34 6070 8 River mountains tunnel No. 1 US 3.66 6087 9 Evinos-Mornos Project Greece 4.2 849 10 11 Evinos-Mornos Project Greece 4.2 2823 12 Evinos-Mornos Project Greece 4.2 1566 13 Evinos-Mornos Project Greece 4.2 2194 14 Evinos-Mornos Project Greece 4.2 664 15 Evinos-Mornos Project Greece 4.04 449 Evinos-Mornos Project Greece 4.04 2083 16 Evinos-Mornos Project Greece 4.04 2546 17 Evinos-Mornos Project Greece 4.04 3430 18 Evinos-Mornos Project Greece 4.04 1189 19 Evinos-Mornos Project Greece 4.12 355 20 Evinos-Mornos Project Greece 4.12 443 21 Evinos-Mornos Project Greece 4.12 1869 22 23 Evinos-Mornos Project Greece 4.12 2212 24 Evinos-Mornos Project Greece 4.12 1866 25 Evinos-Mornos Project Greece 4.2 298 26 Evinos-Mornos Project Greece 4.2 1055 27 Evinos-Mornos Project Greece 4.2 128 Evinos-Mornos Project Greece 4.2 1781 28 268 29 Evinos-Mornos Project Greece 4.2 923 Blue mountains Sewerage-Katoomba carrier tunnel Australia 3.4 13081 30 Blue mountains Sewerage-Lawson carrier tunnel Australia 3.4 2769 31 Hazelbrook Carrier Section 3 Australia 3.35 9520 32 Rapid transit subway tunnel US 6.55 4800 33 Yucca Mountains tunnel US 7.62 7600 34 Honkong Cable tunnel China 4.8 5200 35 Calument city leg US 5.28 2858 36 37 Marhkam and CDS-48 leg US 3.71 4862 38 Indiana Ave. 140th street legs US 9.85 11565 39 Cowles Mountain Tunnel US 3.43 1950 40 Yindaruqin Irrigation project(30A) China 5.54 11649 41 Yindaruqin Irrigation project(38) China 5.54 4880 Svartisen Storglomvatn Norway 4.3 5954.5 42 Svartisen Storglomvatn Norway 5 7816 43 Svartisen Storglomvatn Norway 4.3 11861 44 Svartisen Staupaga Norway 3.5 8219 45 Meraaker Hydro tunneling Norway 3.5 10120 46 Clermont -Inanda Wiggins-Umgeni In-W2 SA 3.5 5372 47 IVAR project Norway 3.52 6130.3 48 49 IVAR project Norway 3.52 776.5 50 IVAR project Norway 3.52 1085.3 51 Syar tunnel US 3.61 9200 52 Los Rosales water tunnel US 3.54 9088 53 San Manual Mine Magma Copper US 4.62 2050 San Manual Mine Magma Copper US 4.62 10750 54 Montana Mine US 4.6 5650 55 Upper Collierville power tunnel -headrace US 5.5 6261 56 Lower Collierville power tunnel -headrace US 4.3 2145 57 Buffalo LRRT C11 Outbound US 5.64 3111 58 Buffalo LRRT C11 Inbound US 5.64 2972.9 59 Buffalo LRRT C31 Outbound US 5.66 2069.6 60 61 Buffalo LRRT C31 Inbound US 5.64 2152 62 Culvar goodman US 5.87 5058 63 Tarp Contract 73-287-2H US 6.48 7364 269 The Genesee River Interceptor Southwest (GRIS) US 5.64 1547 64 tunnel project The Genesee River Interceptor Southwest (GRIS) US 5.64 944 65 tunnel project 66 Coal Mine access tunnel US 7.6 237.3 Coal Mine access tunnel US 7.6 1171.5 67 Coal Mine access tunnel US 7.6 1129 68 Coal Mine access tunnel US 7.6 983.3 69 Port Huron project US 5.6 9618 70 Boston outfall tunnel US 8.08 15289 71 Subway tunnel at Nuremberg Germany 9.17 8458 72 73 Inter Island Tunnel Boston harbor US 4.2 5300 74 TARP US 5.56 11954 75 Tolo Effluent Export Tunnel Hong Kong 3.56 7350 76 Silvermine Bay Aqueduct Hong Kong 3.56 6954 77 Maen 4.2 1750 78 Cogolo 3.9 500 Metro Alpine 4.2 1580 79 Zermatt Sunnegga 3.7 1700 80 Clauson Dixence Section F8/F6 4.7 1600 81 Clauson Dixence Section F6/F5 4.7 700 82 Clauson Dixence Section F5/Verruccano 4.4 450 83 Silz 3.2 1995 84 85 Alassio Italy 3.6 2310 86 Cardano-Prato Tires Italy 3.9 2878 87 Cardano-Prato Tires Italy 3.9 577 88 Cardano-Prato Tires Italy 3.9 1007 89 Cardano-Prato Tires Italy 3.9 629 90 Val D’arzino (Valley of Arzino) Italy 4.5 5663 Prato Isarco 3.5 12500 91 Pre Saint Didier (Left Tube) 3.9 2145 92 Avise (Left Tube) 4.5 1285 93 Avise (Right Tube) 4.5 2640 94 Leverogne (Left Tube) 3.9 1630 95 Leverogne (Right Tube) 3.9 1650 96 97 Arvier (Left Tube) 3.9 2360 98 Arvier (Right Tube) 3.9 2355 270 99 Villeneuve (Left Tube) 3.9 2750 Villeneuve (Right Tube)-1 4.7 570 100 Villeneuve (Right Tube)-2 3.9 2200 101 Varzo Tunnel Italy 3.84 4500 102 Isola headrace tunnel Italy 3.5 6187 103 Corniolo diversion tunnels Italy 3.5 4582 104 Tarabya tunnel Turkey 2.915 13,270 105 Tarabya tunnel-Ch. 981-2260 1,279 106 107 Tarabya tunnel-Ch. 2500-7700 5,200 108 Inuyama water conveyance tunnel Japan 4.3 109 Kagawa water utilization, Asan tunnel Japan 4.3 110 Distribution-reservoir tunnel for Yasumiyama tunnel Japan 3.5 111 Hokuriku line, Koura tunnel-pilot tunnel Japan 2.3 New Sanyo line, Saijo tunnel-pilot tunnel Japan 4.5 112 Ou line, Taiheiyama tunnel-pilot tunnel Japan 4.5 113 New Tohoku line, Oka tunnel-pilot tunnel Japan 5 114 Route 45, Hamada pedestrian tunnel Japan 3.3 155 115 Route 6, Hattachi second pedestrian tunnel Japan 3.5 116 Route 6, Hattachi first pedestrian tunnel Japan 3.5 117 Ogouchi water conveyance tunnel Japan 3.6 118 119 Hiraya water conveyance tunnel Japan 2.6 120 Nikengoya water conveyance tunnel Japan 2.7 5,503 121 Doushi water conveyance tunnel Japan 3 122 Shinyuyama water conveyance tunnel Japan 3.8 123 Maiko road tunnel (pilot tunnel) Japan 5 2,331 Tsukui water conveyance tunnel Japan 5.4 5,222 124 Takisato water conveyance tunnel Japan 8.3 125 Bergen motorway system-Floyfjell tunnel-South Norway 7.8 7000 126 bound Bergen motorway system-Floyfjell tunnel-South Norway 7.8 3,200 127 bound-tube no. 1 Bergen motorway system-Floyfjell tunnel-South Norway 7.8 3,800 128 bound-tube no. 2 129 headrace tunnel of Amsteg power station Switzerland 5.08 7,300 130 The Inanda-Wiggins Aqueduct-Emolweni SA 3.5 5,328 131 Intrenchment Creek US 7.95 2,878 132 Juam Korea 4.5 8,600 The Kemano Completion Project (KCP) Canada 5.73 780 133 271 134 Kerckhoff-2 Hydroelectric Powerplant US 7.34 6,706 Midmar Water Tunnel SA 3.5 6,424 135 Nyset-Steggje Norway 3.2 136 Paute Hydroelectric Project Ecuador 7.8 6,000 137 Saltsjo Sweden 3.5 7,500 138 Sand Bar Hydroelectric Project US 3.71 5,486 139 Seabrook Nuclear Station-Intake tunnel 6.7 4,960 140 Seabrook Nuclear Station-Discharge tunnel 6.7 4,752 141 142 Seoul Ring 6.5 143 Stillwater Mine-5000E US 4.1 1,156 144 Stillwater Mine-5700W US 4.1 1,462 145 Stillwater Mine-5900W US 4.1 3,444 146 Stillwater Mine-5500W US 4.1 514 Terror Lake Hydroelectric Project -headrace Alaska 3.35 7,622 147 Three Rivers Tunnel US 3.2 8,199 148 Tjodan hydropower plant Norway 3.2 5,100 149 Ulla-Forre Norway 3.5 8,022 150 BI-county-West US 3.81 4,595 151 BI-county-East US 3.81 5,500 152 Drassnitz Austria 3.5 153 154 E 63rd St 6.7 headrace tunnel of the Hayakido Hydro Power 2.6 1,146 155 Station-headrace Jostedal Stegagerdet Norway 4.5 9,001 156 Klippen Hydropower Sweden 6.5 10,315 157 Klippen Hydropower- Headrace Sweden 6.5 6,914 158 Klippen Hydropower- Tailrace Sweden 6.5 3,401 159 160 Kobbelv headrace 6.25 9,332 161 Lamnitz 3.5 162 Mosvik Hydro Electric Power Plant-headrace Norway 3.5 5,390 163 The Ormen (The Snake) Sweden 3.5 3,200 164 Rogers Pass Canada 6.8 8,199 165 Rogers Pass 2,240 Rogers Pass 305 166 Rogers Pass 335 167 Rogers Pass 351 168 272 169 Rogers Pass 2,164 Rogers Pass 396 170 Rogers Pass 640 171 Rogers Pass 762 172 Rogers Pass 1,006 173 Stobie 3.66 174 the Queens Water Tunnel # 3, Stage 2 US 7.06 7,500 175 W-80 US 2.29 1,189 176 177 Wolla Austria 3.5 6,700 178 Woodside interceptor US 2.19456 366 179 Zillergrundl Austria 4.74 7,300 180 Carhuaquero Hydroelectric-headrace Peru 3.8 8,500 181 Faeroe 3.35 Gossensass Italy 3.5 5,128 182 Kiena gold mine-S.E. Drift Canada 2.1844 514 183 Kiena gold mine N.W. Drift Canada 2.1844 862 184 Lesotho-Muela Adit South SA 5.03 16,148 185 Lesotho-Hlotse Adit South SA 5.03 17,040 186 Lesotho-Katse Intake North SA 5.018 13,047 187 Lesotho-Katse Intake North SA 5.018 2,131 188 189 Lesotho-Katse Intake North SA 5.018 5,225 190 Lesotho-Katse Intake North SA 5.018 5,691 Lesotho-(Hololo-Muela, Ngoajane-Hololo, Ngoajane- SA 5.18 13,047 191 Vent. Sh. 5) Lesotho-(Hololo-Muela, Drive 1) SA 5.18 2,131 192 Lesotho-(Ngoajane-Hololo, Drive 2) SA 5.18 5,225 193 Lesotho-(Ngoajane- Vent. Sh. 5, Drive 3) SA 5.18 5,691 194 195 Mount Etna 5.86 196 Ponte Gardena Italy 3.5 13,159 197 Portland Reach B US 6.49 1,432 198 Portland Reach C US 6.49 1,533 199 Transfare 4.3 200 State-Mt Hope 5 Vinstra hydropower plant-headrace Norway 4.75 16,562 201 Clermont 3.5 202 Midmar SA 3.5 6,499 203 273 204 Sherwood SA 3 3,300 Thomson-Yarra Australia 4.12 205 Zurich LIMESTONE 3 206 Amlach hydro electric headrace Austria 3.9 13,350 207 Camporosso Italy 3.9 6,559 208 Knoxville 1.63 209 Presenzano Italy 6.6 210 Rovereto Hydraulic tunnel Italy 3.5 211 212 Yellow River 6.1 213 Mont Cenis France 2.2 300 214 Star Mine US 2.74 125 215 Libanon Gold Mine SA 3.35 216 Tunjita Columbia 4.3 10,500 Selkirk Canada 6.8 8,350 217 Selby Coalfiel UK 5.8 6,470 218 Zurichberg Railroad Tunnel Switzerland 11.52 4,356 219 The Quebec Water Treatment Board US 4 5,683 220 The Quebec Water Treatment Board US 4 1,782 221 The Quebec Water Treatment Board US 4 2,208 222 The Quebec Water Treatment Board US 4 1,693 223 Bafokeng Rasimone SA 2.1 224 Platinum Mine (BRPM) TARP-North Branch Tunnel US 9.83 14,075 225 Kilvik Headrace Norway 8.5 7,334 226 Storjord Roofgutter South Norway 3.5 9,273 227 Govalle Segment B (from FARM to PIZZA) US 3.3 4,511 228 Govalle Segment B US 3.3 3,695 229 230 Govalle Segment B US 3.3 556 231 Govalle Segment C (from PIZZA to CATERBURY) US 3.2 5,541 232 Govalle Segment C US 3.2 1,681 233 Govalle Segment C US 3.2 3,860 234 Milwaukee Project No. 1 US 9.9 6,245 235 Milwaukee Project No. 2-1 US 5.8 870 Milwaukee Project No. 2-2 US 5.8 870 236 Milwaukee Project No. 2-3 US 2 2,054 237 Cleveland Southwest Intercept 3 3.4 4,122 238 274 239 Chicago Tarp Contract 72-049-2H 9.1 6,209 Chunnel Landward Service-Britain, OCWI, Section II 5.76 7,888 240 Chunnel Landward-Section B-2 Station 4+77-122+16 2.79 3,578 241 Chunnel Landward-Section L 2.79 707 242 OCWI, Section IV 2.79 4,706 243 OCWI, Section IV 3 4,194 244 Longuevil 4.67 18,646 245 Zurichberg Switzerland 11.52 14,300 246 247 Navajo No. 3 US 6.25 15,264 248 Donkin Morien Canada 7.6 3,522 The Sultan River Blue Mountain Tunnel- 4.32 6,289 249 HEADRACE Lemont interceptor US 1.9812 5,364 250 Northeast relief sewer US 2.5908 2,682 251 Dul Hasti hydroelectric project India 8.33 8,500 252 253 Mamquam hydroelectric project Canada 4.1 1,346 254 Jangdae Korea 3.3 5,285 255 Guinza Italy 3.63 5,930 Blue mountains Sewerage transfer scheme-Katoomba Australia 3.4 13,400 256 Carrier Alpe Devero Delivery Tunnel (Devero to Agaro lake) Italy 3.5 3,577 257 Alpe Devero Delivery Tunnel (Bodolero to Cairasca) Italy 3.5 6,552 258 Alpe Devero Delivery Tunnel (Devero to Bodolero) Italy 3.5 3,400 259 260 Firenze Italy 3.9 2,890 261 Rieti-1 Italy 3.9 4,412 262 Castellammare Italy 3.62 2,820 263 Rockville Section A6a US 5.79 5,486 264 Rockville Section A9a US 5.79 4,267 265 Shimizu I-Drive 1 Japan 2.6 2,711 Akaishizawa Power Plant Project Japan 2.6 6,122 266 Kitamatado Japan 2.6 2,588 267 Pipehead tunnel Australia 3.93 7,500 268 CSOAP-Lyell Avenue (LY) US 4.27 3,081 269 CSOAP-Jay-Arnett US 5.3 2,679 270 CSOAP-Saxton-colvin US 5.5 1,800 271 272 CSOAP-Tiger-Carlisle W US 5 1,659 273 CSOAP-Tiger-Carlisle E US 5 363 275 274 Kielder Water Scheme-Tees to Sharnbey UK 3.55 8,224 Kielder Water Scheme-(Derwent-Wear drive) UK 3.55 3,790 275 Robbins 2nd Dr Kielder Water Scheme-Wear to Sharnberry (South UK 3.5 6,030 276 wear) or Wear to Tees Kielder Water Scheme-Wear to Waskerley (North UK 3.5 8,491 277 wear) or Wear to Derwent 278 Tucson Tunnel US 3.6576 2,540 279 Tucson Tunnel US 3.6576 228 280 Tucson Tunnel US 3.6576 151 281 Tucson Tunnel US 3.6576 1,789 282 Tucson Tunnel US 3.6576 141 283 Tucson Tunnel US 3.6576 231 Chiotas pedestrian tunnel Italy 2.57 1,898 284 Echaillon France 5.8 4,362 285 La Coche France 3 5,287 286 RER Chat elet-Gare de Lyon France 7 5,100 287 Belledonne France 5.88 9,998 288 Bramefarine France 8.1 3,700 289 290 Grand Maison-Eau Dolle-headrace France 3.6 849 291 Western Oslofjord Norway 3 10,500 292 Brevon France 3 4,150 293 Grand Maison-Penstock and service shaft France 3.6 5,466 294 Super Bissorte France 3.6 2,975 295 Pouget France 5.05 3,999 Grand Maison-Vaujany France 7.7 5,400 296 Vieux Pre France 2.9 1,257 297 Haut e Romanche Tunnel France 3.6 2,860 298 Cilaos Reunion 3 5,701 299 Monaco-tunnel No.6 Monaco 5.05 183 300 Ferrieres France 5.9 4,313 301 302 Durolle France 3.4 2,139 303 Mont fermy-1 France 3.55 2,467 304 Mont fermy-2 France 3.55 2,573 France- CERN LEP (machine1 sec. 1) 4.5 2,480 305 Suisse France- CERN LEP (machine1-sec. 2) 4.5 6,100 306 Suisse France- CERN LEP (machine 2) 4.5 6,100 307 Suisse 276 France- CERN LEP (machine 3-sec. 2) 4.5 2,480 308 Suisse France- CERN LEP (machine 3-sec. 1) 4.5 2,226 309 Suisse 310 Val d' Isère funicular France 4.2 1,689 Takamaka II France 3.2 4,803 311 Villejust tunnel(machine1-N) France 9.25 2,618 312 Villejust tunnel(machine1-S) France 9.25 2,159 313 Villejust tunnel(machine2) France 9.25 4,798 314 Sèvres - Achères: Package 3 France 4.05 3,550 315 Channel Tunnel-T1 Undersea service France 5.77 15,618 316 317 Channel Tunnel-T2 Undersea running north France 8.78 20,009 318 Channel Tunnel-T3 Undersea running south France 8.78 18,860 319 Channel Tunnel-T4 Underland service France 5.61 3,162 320 Channel Tunnel-T5 Underland running north France 8.62 3,265 321 Channel Tunnel-T6 Underland running south France 8.62 3,265 322 Channel Tunnel-T7 Undersea service UK 5.38 22,298 Channel Tunnel-T8 Undersea running north UK 8.36 18,532 323 Channel Tunnel-T9 Undersea running south UK 8.36 17,793 324 Channel Tunnel-T10 Underland service UK 5.76 8,152 325 Channel Tunnel-T11 Underland running north UK 8.72 8,157 326 Channel Tunnel-T12 Underland running south UK 8.72 8,140 327 Yamada sewage trunk line laying construction Australia 2 3,342 328 329 Lesotho-Delivery Tunnel North- Ash Tunnel SA 5.39 10,981 330 Lesotho-Delivery Tunnel North- Caledon Tunnel SA 5.39 8,071 331 Hsuehshan Tunnel-Pilot China 4.8 5,168 332 Hsuehshan Tunnel-Main China 11.74 3,870 277 APPENDIX B Typical Minitab Outputs Regression Analysis: ARw versus 15 independent variables Stepwise Regression: : ARw versus PR, UCS, ... Alpha-to-Enter: 0.15 Alpha-to-Remove: 0.15 Response is ARw on 15 predictors, with N = 52 N(cases with missing observations) = 300 N(all cases) = 352 Step 1 2 Constant 40.03 26.13 UCS -0.157 -0.104 T-Value -6.27 -3.44 P-Value 0.000 0.001 PR 2.47 T-Value 2.78 P-Value 0.008 S 7.07 6.64 R-Sq 44.00 51.61 R-Sq(adj) 42.88 49.64 Mallows Cp 1.6 -3.1 Best Subsets Regression: ARw versus PR, UCS, ... Response is ARw 52 cases used, 300 cases contain missing values U R Q G W E S S P T T R D L Mallows P C Q t v a v u u o o h P i e Vars R-Sq R-Sq(adj) Cp S R S D z a t a p R w q r M a n 1 44.0 42.9 1.6 7.0676 X 1 39.9 38.7 5.3 7.3218 X 2 51.6 49.6 -3.1 6.6365 X X 2 51.1 49.1 -2.7 6.6717 X X 3 53.3 50.4 -2.6 6.5865 X X X 3 52.9 49.9 -2.3 6.6166 X X X 4 54.6 50.7 -1.8 6.5635 X X X X 4 54.5 50.6 -1.7 6.5727 X X X X 5 56.4 51.7 -1.4 6.5022 X X X X X 5 55.8 51.0 -0.9 6.5446 X X X X X 6 57.0 51.3 0.1 6.5291 X X X X X X 278 6 56.8 51.0 0.3 6.5431 X X X X X X 7 58.0 51.3 1.2 6.5272 X X X X X X X 7 57.4 50.7 1.7 6.5689 X X X X X X X 8 58.5 50.7 2.8 6.5638 X X X X X X X X 8 58.2 50.4 3.1 6.5884 X X X X X X X X 9 58.7 49.8 4.6 6.6246 X X X X X X X X X 9 58.7 49.8 4.6 6.6262 X X X X X X X X X 10 58.9 48.9 6.4 6.6850 X X X X X X X X X X 10 58.8 48.8 6.5 6.6941 X X X X X X X X X X 11 59.1 47.8 8.3 6.7543 X X X X X X X X X X X 11 59.1 47.8 8.3 6.7545 X X X X X X X X X X X 12 59.3 46.8 10.0 6.8206 X X X X X X X X X X X X 12 59.1 46.5 10.2 6.8386 X X X X X X X X X X X X 13 59.4 45.5 12.0 6.9069 X X X X X X X X X X X X X 13 59.3 45.4 12.0 6.9088 X X X X X X X X X X X X X 14 59.4 44.0 14.0 6.9989 X X X X X X X X X X X X X X 14 59.4 44.0 14.0 6.9990 X X X X X X X X X X X X X X 15 59.4 42.4 16.0 7.0948 X X X X X X X X X X X X X X X Regression Analysis: Uw versus 15 independent variables Stepwise Regression: Uw versus PR, UCS, ... Alpha-to-Enter: 0.15 Alpha-to-Remove: 0.15 Response is Uw on 15 predictors, with N = 52 N(cases with missing observations) = 300 N(all cases) = 352 Step 1 2 3 Constant 44.78 67.44 66.76 PR -4.21 -6.64 -6.30 T-Value -4.71 -6.51 -6.16 P-Value 0.000 0.000 0.000 UCS -0.131 -0.108 T-Value -3.79 -2.95 P-Value 0.000 0.005 Max Support Installed -2.2 T-Value -1.71 P-Value 0.094 S 8.57 7.61 7.46 R-Sq 30.72 46.45 49.52 R-Sq(adj) 29.33 44.26 46.36 Mallows Cp 8.8 -2.1 -2.6 Best Subsets Regression: Uw versus PR, UCS, ... Response is Uw 52 cases used, 300 cases contain missing values U R Q G W E S S P T T R D L 279 Mallows P C Q t v a v u u o o h P i e Vars R-Sq R-Sq(adj) Cp S R S D z a t a p R w q r M m n 1 30.7 29.3 8.8 8.5662 X 1 12.9 11.2 23.3 9.6027 X 2 46.4 44.3 -2.1 7.6075 X X 2 40.4 38.0 2.8 8.0267 X X 3 49.5 46.4 -2.6 7.4627 X X X 3 47.7 44.4 -1.1 7.5993 X X X 4 50.4 46.2 -1.4 7.4721 X X X X 4 50.4 46.1 -1.3 7.4779 X X X X 5 51.8 46.6 -0.5 7.4467 X X X X X 5 51.4 46.1 -0.2 7.4785 X X X X X 6 52.7 46.4 0.7 7.4607 X X X X X X 6 52.2 45.8 1.1 7.4988 X X X X X X 7 53.2 45.7 2.3 7.5054 X X X X X X X 7 53.1 45.6 2.4 7.5125 X X X X X X X 8 53.7 45.1 3.9 7.5511 X X X X X X X X 8 53.6 45.0 4.0 7.5555 X X X X X X X X 9 54.3 44.5 5.4 7.5917 X X X X X X X X X 9 54.2 44.4 5.5 7.5980 X X X X X X X X X 10 55.1 44.1 6.8 7.6175 X X X X X X X X X X 10 54.7 43.6 7.1 7.6505 X X X X X X X X X X 11 55.4 43.1 8.5 7.6841 X X X X X X X X X X X 11 55.3 43.0 8.6 7.6904 X X X X X X X X X X X 12 55.7 42.0 10.3 7.7577 X X X X X X X X X X X X 12 55.6 42.0 10.4 7.7634 X X X X X X X X X X X X 13 55.9 40.9 12.1 7.8355 X X X X X X X X X X X X X 13 55.9 40.8 12.2 7.8430 X X X X X X X X X X X X X 14 56.1 39.4 14.0 7.9308 X X X X X X X X X X X X X X 14 56.0 39.3 14.1 7.9400 X X X X X X X X X X X X X X 15 56.1 37.7 16.0 8.0402 X X X X X X X X X X X X X X X Regression Analysis: LnFPI versus RQD, CAI, UCS The regression equation is LnFPI = 1.97 + 0.00630 RQD + 0.103 CAI + 0.00685 UCS 136 cases used, 17 cases contain missing values Predictor Coef SE Coef T P VIF Constant 1.96950 0.08382 23.50 0.000 RQD 0.006303 0.001231 5.12 0.000 1.218 CAI 0.10322 0.01407 7.33 0.000 1.400 UCS 0.0068522 0.0004589 14.93 0.000 1.477 S = 0.259565 R-Sq = 84.4% R-Sq(adj) = 84.0% Analysis of Variance Source DF SS MS F P Regression 3 48.029 16.010 237.63 0.000 Residual Error 132 8.893 0.067 Total 135 56.923 280 Source DF Seq SS RQD 1 16.649 CAI 1 16.358 UCS 1 15.023 Unusual Observations Obs RQD LnFPI Fit SE Fit Residual St Resid 21 90 2.7384 3.3866 0.0375 -0.6482 -2.52R 32 90 2.9172 3.4292 0.0446 -0.5120 -2.00R 33 80 2.6895 3.2863 0.0374 -0.5967 -2.32R 34 90 2.6616 3.1766 0.0424 -0.5150 -2.01R 103 83 5.1127 4.5668 0.0534 0.5459 2.15R 124 100 3.1421 3.6636 0.0678 -0.5215 -2.08R R denotes an observation with a large standardized residual. Residual Plots for LnFPI Regression Analysis: Ln(ARw) versus UCS, D, Dc, M The regression equation is Ln(ARw) = 3.67 - 0.00589 UCS - 0.0851 D + 0.0285 Dc + 0.0988 M Predictor Coef SE Coef T P Constant 3.66814 0.09071 40.44 0.000 281 UCS -0.0058864 0.0001436 -41.00 0.000 D -0.085117 0.003197 -26.62 0.000 Dc 0.028475 0.005345 5.33 0.000 M 0.09881 0.02733 3.62 0.000 S = 0.113635 R-Sq = 93.1% R-Sq(adj) = 93.0% Analysis of Variance Source DF SS MS F P Regression 4 34.4945 8.6236 667.84 0.000 Residual Error 198 2.5567 0.0129 Total 202 37.0513 Source DF Seq SS UCS 1 25.2747 D 1 8.7448 Dc 1 0.3063 M 1 0.1688 Unusual Observations Obs UCS Ln(ARw) Fit SE Fit Residual St Resid 66 100 3.08649 2.82568 0.02688 0.26080 2.36R 96 100 3.08649 2.82568 0.02688 0.26080 2.36R 123 100 3.08649 2.82568 0.02688 0.26080 2.36R 130 100 3.08649 2.82568 0.02688 0.26080 2.36R 199 100 3.08649 2.82568 0.02688 0.26080 2.36R R denotes an observation with a large standardized residual. Residual Plots for Ln(ARw) 282 283 VITA Ebrahim Farrokh EDUCATION 2009-2013: PhD at Penn State University. Dissertation: “TBM utilization and advance rate prediction.” 2001–2004: M.Sc. in Mining Engineering: Tehran University/ Tehran, Iran/Thesis Subject: “Evaluation of Ghomroud Tunnel Convergence and Studying its Effect on TBM Performance.” 1997–2001: B.Sc. in Mining Engineering: Yazd University/ Yazd, Iran. WORK EXPERIENCE 2011: Internship at Brierley Associates. 2010-2011: Administrator of Geomechanic Lab for rock properties testing at PSU. 2007-2008: Administrator of engineering group in Parts III & IV of Ghomroud Water Conveyance Tunnel Site (TBM driven tunnel) in charge of performing engineering services. TEACHING EXPERIENCE 2010-2012: Teaching Assistant for Rock Mech. Lab/course at PSU. 2002: Advanced Underground Mining Teaching for Sub-level caving method. 2000: Geotech. Eng. Teaching Assistant at Yazd University. BOOK Farrokh, E. “Concrete Segmental Lining, Procedure of design, production, and erection of segmental lining in mechanized tunnelling”, 2006, Publication of Jihad Amirkabir University, Tehran, 310 pages in Persian. ACADEMIC HONORS 2012: NAT Student Conference Scholarship Award sponsored by Society AND AWARDS for Mining, Metallurgy, and Exploration (SME). 2011: IAAP student scholarship award recipient for year 2011 at PSU for being recognized in the fields of mining and tunneling. 2010: Hardy Memorial Award recipient for academic year of 2010/2011 at PSU for excellence in mining eng. PROFESSIONAL 2010-2012: Society for Mining, Metallurgy, and Exploration (SME) MEMBERSHIPS 2010: International Explosive Engineering Society.