JACKKNIFE STABILITY OF ARTICULATED TRACTOR SEMITRAILER VEHICLES WITH HIGH-OUTPUT BRAKES AND JACKKNIFE DETECTION ON LOW COEFFICIENT SURFACES
DISSERTATION
Presented in Partial Fulfillment of the Requirements for
the Degree Doctor of Philosophy in the
Graduate School of The Ohio State University
By
Ashley Liston Dunn, M.S.M.E.
* * * * *
The Ohio State University
2003
Ph.D. Examination Committee: Approved by Dr. Dennis Guenther , Adviser Dr. Giorgio Rizzoni, Adviser AdviserAdviser Dr. Gary Heydinger Co-Adviser Department of Mechanical Engineering
Copyright 2003
by
Ashley Liston Dunn, M.S.M.E.
ABSTRACT
This dissertation includes a detailed study of the effects on jackknife stability when using brakes that have higher than standard torque output on the prime mover. This research was initiated by the National Highway Traffic Safety Administration (NHTSA) to investigate the long-term effects of reducing the government mandated stopping distances for heavy commercial vehicles (i.e., FMVSS 121 standards). A significant reduction in stopping distance requirements would result in the need for higher torque brakes on the prime mover (tractor) steer and drive axles. The brake type might change from drum to disc, changing not only the torque output in magnitude but the dynamic transfer function characters as well.
To investigate this phenomenon, a sophisticated nonlinear model for a complete pneumatic braking system, typically found on Class 8 heavy trucks in North America, was developed. Modern brake torque measurements were used to develop brake torque output relationships. The modeled pneumatic and anti-lock braking (ABS) systems were verified over a wide variety of load/surface µ conditions using experimental data taken at
NHTSA’s Vehicle Research and Test Center. The nonlinear model was then run in
ii parallel with a proven commercial vehicle simulation package, as a hybrid computer
model.
The rigorous investigations showed no detrimental effects on jackknife stability,
with the prime mover ABS “on” or “off,” due to higher-torque brakes that were
electronically controlled (deleting the time lag of the traditional pneumatic control
system).
To signal a hypothetical vehicle stability system of impending jackknife
instability, a 15-state model was rigorously developed using Lagrange’s method, then
integrated into an Extended Kalman Filter (EKF). The assumed measured states of longitudinal slip ratios, prime mover lateral acceleration and yaw rate are used, in
conjunction with the EKF, to estimate the states of trailer hitch angle and its rate of
change. The hitch angle parameters were used to construct state phase plots, whose
output was shown to agree with sophisticated vehicle modeling programs over a range of
low-load and low-µ conditions.
iii
DEDICATION
Dedicated to my wife, Melissa, and to my family, including my mother, Blanche, for her tireless love and dedication to my success.
Also, to Michael and Malika Drexel for their supporting love and wonderful friendship.
iv
ACKNOWLEDGMENTS
I wish to express deep appreciation to Dr. Don Houser, Dr. Dennis Guenther, Dr.
Giorgio Rizzoni, and Dr. Gary Heydinger for their infinite support and patience in this
team effort. Dr. Houser has helped me since the beginning of my graduate career, in
September 1996. I am quite fortunate for the honor of being a doctoral candidate in the
Department of Mechanical Engineering at The Ohio State University.
The National Highway Traffic Safety Administration (of the United States
Department of Transportation) has funded this research in full. I am eternally grateful for
the support from the knowledgeable management, engineers, and staff at NHTSA’s
Vehicle Research and Test Center, in East Liberty, Ohio. The VRTC has been for me one of the most pleasant work environments one could ask for. Their technical support has greatly complemented the quality of this research.
Thanks also to Dr. Kamel Salaani for his efforts in keeping me focused on the goals of import, meticulous editing of my technical documents, and most importantly, stimulation of technical thought regarding this research. I am deeply grateful for the boundless and enduring support of my wonderful wife, Melissa Wolfe, who herself is a
Ph.D. candidate.
v Finally, many thanks are necessary to the very generous engineering staff at
DANA Corporation for their technical assistance in understanding the mechanisms behind torque generation for heavy commercial vehicle brakes.
vi
VITA
Ashley Liston Dunn, M.S.M.E
March 3, 1961 ...... Born – Morristown, Tennessee
June, 1986 ...... Bachelor of Science in Mechanical Engineering, North Carolina State University, Raleigh, North Carolina
June, 1999 ...... Master of Science in Mechanical Engineering, The Ohio State University, Columbus, Ohio
publications
Research Publications 1. A.L. Dunn, D.R. Houser, and T.C. Lim, “A New Metric for Rating In-Vehicle Gear Whine Levels,” Noise-Con 98, 1998.
2. A.L. Dunn, D.R. Houser, and T.C. Lim, “Methods for Researching Gear Whine in Automotive Transaxles,” SAE 99NV-167, 1999.
3. A.L. Dunn, “The Effects of Cornering Force Variation on Articulated Vehicle Predictions,” NHTSA / VRTC internal report, September 2000.
vii
4. A.L. Dunn, “Brake-In-Turn Study Comparing Disc/Drum to Drum/Drum Truck Brake Combinations,” NHTSA / VRTC internal report, 19 September 2000.
5. A.L. Dunn, “The Effects of Tire Free Rolling Cornering Properties on Analytical Predictions for Heavy Truck Roll and Yaw Predictions”, NHTSA / VRTC internal report, 24 September 2001.
6. A.L. Dunn, G.J. Heydinger, G. Rizzoni, and D.A. Guenther, “New Model for Simulating the Dynamics of Pneumatic Heavy Truck Brakes with Integrated Anti- Lock Control,” SAE 2003-01-1322, 2003.
7. A.L. Dunn, G.J. Heydinger, G. Rizzoni, and D.A. Guenther, “Methods for Modeling Brake Torque from Experimental Data,” SAE 2003-01-1325, 2003.
8. A.L. Dunn, “Heavy Truck Tire Load Transfer Sensitivity,” NHTSA / VRTC internal report, 28 March 2003.
FIELDS OF STUDY
Major Field: Mechanical Engineering
viii
TABLE OF CONTENTS
ABSTRACT...... ii DEDICATION...... iv ACKNOWLEDGMENTS ...... v VITA...... vii TABLE OF CONTENTS...... ix LIST OF FIGURES ...... xv LIST OF TABLES...... xxiv CHAPTER 1 INTRODUCTION TO THE PROBLEM OF MODELING HEAVY TRUCK BRAKING SYSTEMS AND PREDICTING JACKKNIFE STABILITY...... 1 1.1 Abstract and Motivation...... 1
1.2 Problem Background ...... 2
1.2.1 The Operation of Pneumatically Controlled Brakes on Tractor- Semitrailer Vehicles...... 2 1.2.2 Electronic Controlled Braking Systems (ECBS) ...... 3 1.2.3 Compatibility of Vehicles Equipped with ECBS and Those Equipped with Pneumatically Controlled Brakes...... 4 1.2.4 Pneumatic Disc Brakes on Class 8 Trucks in the United States...5 1.3 Intent and Scope of This Research ...... 8
1.3.1 Jackknife Instability During Brake-in-Turn (B.I.T.) Maneuvers..8 1.4 Detection of the Jackknife Event While It Occurs ...... 12
1.5 Chapter 1 References...... 14
ix CHAPTER 2 EMPIRICAL MODELS FOR COMMERCIAL VEHICLE BRAKE TORQUE FROM EXPERIMENTAL DATA...... 16 2.1 Abstract...... 16
2.2 Motivation...... 16
2.3 Model Development in General...... 17
2.3.1 Model Limits and Assumptions...... 17 2.3.2 Linear Brake Torque Model...... 19 2.4 Nonlinear Response of Some Brake Torques...... 22
2.5 Overall View of Brake Torque Test Data...... 30
2.6 Brief Discussion on Dynamic Axle Weights and Their Impact on These Models ...... 30
2.7 Addressing Brake Fade During a Stop ...... 36
2.8 Brake Torque Model in Operation...... 40
2.9 Conclusions...... 45
2.10 Chapter 2 References...... 46
CHAPTER 3 DEVELOPMENT OF AN ANALYTICAL MODEL FOR SIMULATING THE DYNAMICS OF PNEUMATIC HEAVY TRUCK BRAKES WITH INTEGRATED ANTI-LOCK CONTROL ...... 47 3.1 Abstract...... 47
3.2 Motivation...... 48
3.3 Background...... 49
3.4 Model Overview in General ...... 50
3.5 Dynamics of the Pneumatic Brake System...... 51
3.5.1 Modeling the Dynamics of the Pneumatic Brake Chambers...... 51 3.5.2 Torque Properties of the Drum and Disc Brakes...... 53 3.6 Brake System Parameters ...... 58
x 3.6.1 Brake Simulation Physical Parameters ...... 58 3.7 The Hysteresis Element of the Brake Model...... 60
3.7.1 The Influence of Hysteresis on Simulated ABS-Assisted Stopping Distance...... 63 3.8 The Integrated 4s/4m ABS Controller...... 64
3.8.1 Brief Discussion of Tire Traction Theory and Priorities of an ABS System...... 64 3.8.2 The Simulated 4s/4m ABS Controller ...... 65 3.9 The Simulation of ECBS ...... 71
3.10 Comparisons of Simulation to Experimental Data ...... 73
3.11 The Build-Hold-Build Algorithm ...... 83
3.11.1 Additional Data Results...... 87 3.12 Conclusions...... 97
3.13 Chapter 3 References...... 98
CHAPTER 4 THE EFFECTS OF USING ECBS-DISC BRAKE SYSTEMS ON THE JACKKNIFE STABILITY OF TRACTOR-SEMITRAILER VEHICLES IN BRAKE-IN-TURN MANEUVERS...... 99 4.1 Abstract...... 99
4.2 Intent and Scope of This Research ...... 100
4.2.1 Previous Study Results, Summarized ...... 101 4.3 Simulation Conditions for This Study ...... 103
4.3.1 Simulated Variable µ Levels...... 105 4.3.2 Brake Simulation Physical Parameters ...... 106 4.3.3 Simulated Brake Antilock Control...... 106 4.3.4 TruckSim™ v. 5.0 Run Screens and Files...... 108 4.3.5 Pass-Fail Criteria...... 108 4.4 Simulation Findings...... 109
4.5 State Plots and Observed Behaviors ...... 114
4.6 Conclusions...... 117
xi 4.7 Chapter 4 References...... 119
CHAPTER 5 DERIVATION AND VALIDATION OF 3-AXLE AND 5-AXLE PLANAR ARTICULATED VEHICLE MODELS...... 120 5.1 Abstract...... 120
5.2 3-Axle Planar Articulated Vehicle Model Derivation ...... 121
5.3 Kinematic Equations and Constraints...... 122
5.4 Derivation of Forcing Functions via Virtual Work Expressions ...... 132
5.4.1 Alternate Methods for Determining the Virtual Work Expressions ...... 138 5.5 The Equations of Motion for the 3-Axle Planar Model...... 141
5.6 Linearized Tire Forces...... 142
5.7 State-Space Representation of the 3-Axle Planar Cornering Model ....148
5.8 Verification of the 3-Axle Model ...... 149
5.8.1 Initial Verification...... 149 5.8.2 Verification by Comparison of the Planar Model Response with that from TruckSim™ Vehicle Dynamics Software...... 149 5.8.3 Verification of the 3-Axle Planar Model by TruckSim™ Simulation Output via Light Handling Maneuvers...... 152 5.8.4 Discussion of 3-Axle Simulation Outputs ...... 153 5.9 Adapting the 3-Axle Model to the 5-Axle Configuration...... 159
5.10 Validation of 5-Axle Model and Comparisons to the 3-Axle Model ...162
5.11 The Effects of Tire Dynamic Lag on Model Response ...... 171
5.11.1 Linear Model Response to a Jackknife-Producing Input...... 171 5.12 Development and Validation for the 5-Axle Nonlinear Model ...... 178
5.13 Performance of the 5-Axle Planar Nonlinear Model for Simulating the Jackknifing Event ...... 185
5.13.1 Successful Simulation of the Jackknife Event...... 192 5.14 Conclusions and Recommendations ...... 194
xii 5.15 Chapter 5 References...... 195
CHAPTER 6 IMPLEMENTATION OF THE EXTENDED KALMAN FILTER TO AN ARTICULATED VEHICLE MODEL FOR THE PURPOSE OF JACKKNIFE PREDICTION ...... 196 6.1 Abstract...... 196
6.2 Linear Observer Theory and How It Applies to Kalman Filters ...... 196
6.2.1 Linear State Observers...... 197 6.3 The Linear Observer as the Kalman Filter...... 201
6.4 Background...... 202
6.5 The Kalman Filter Equations and Operation Philosophy ...... 202
6.6 Extrapolation of the Kalman Filter into the Extended Kalman Filter for Nonlinear Systems ...... 207
6.7 Application of the Linear and Extended Kalman Filter to the Planar Articulated Vehicle Model...... 211
6.7.1 Stated Goals...... 211 6.7.2 Outline of the Extended Kalman Filter Algorithm ...... 211 6.8 Tuning the Extended Kalman Filter to Optimize Its Performance ...... 214
6.8.1 Comparisons of Three Estimates for Error Covariance...... 215 6.9 Various Runs – Discussion...... 228
6.9.1 Medium µ (µ=0.55), ½ GVW Load ABS ON, Resulting in a Stable Stop...... 228 6.9.2 Medium µ (µ=0.55), ½ Load ABS OFF, Resulting in a Jackknife ...... 235 6.9.3 Low µ (µ=0.30), ½ GVW Load ABS ON, Resulting in Controlled Stop...... 239 6.9.4 Low µ (µ=0.30), ½ GVW Load ABS OFF, Resulting in a Jackknife ...... 243 6.9.5 Medium µ (µ=0.55), 0 Payload ABS ON, Resulting in a Stable Stop ...... 247 6.9.6 Sensitivity of the EKF to Improper Load Estimates...... 251 6.9.7 Medium µ (µ=0.55), 0 Load ABS OFF, Resulting in a Jackknife ...... 256
xiii 6.9.8 Low-µ (µ = 0.30), 0-Load, ABS ON, Resulting in a Stable Stop ...... 260 6.9.9 EKF Sensitivity to Improper Estimation of Surface µ...... 265 6.9.10 Low-µ (µ = 0.30), 0-Payload, ABS OFF, Resulting in a Jackknife ...... 273 6.9.11 Operation of the EKF in a Double Lane Change While in a 152.4 m (500-ft) Diameter Turn, No Braking, µ = 0.55 ...... 278 6.10 Jackknife Detection Warning Lead Time Estimates...... 283
6.11 Conclusions...... 286
6.12 Chapter 6 References...... 289
CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS...... 290 7.1 Summary of Fundamental Contributions to the Engineering Community ...... 290
7.2 Summary of Dissertation Topics ...... 291
7.3 Recommended Actions...... 294
7.4 Chapter 7 References...... 296
APPENDIX A TruckSim™ v.5.0 Parsfile Numbers and import / expport variables used in Chaper 5 ...... 297 APPENDIX B State-space coefficients for 3-axle planar model...... 303 APPENDIX C State-space coefficients for 5-axle planar model...... 307 APPENDIX D Runge-kutta integration routine in matlab® ...... 313 BIBLIOGRAPHY...... 315
xiv
LIST OF FIGURES
Figure 1.1 Schematic of the actuation mechanism of a pneumatically operated s-cam drum brake ...... 3 Figure 1.2 Modeled and experimental data for drive axle drum brakes. Broken lines (with icons) indicate experimental data at a simulated axle load of 21,000 lb...... 7 Figure 1.3 Modeled and experimental data for drive axle disc brakes at a simulated GAWR of 23,000 lb. Color and type schemes are identical to those of Figure 1.2...... 7 Figure 2.1 Drum brake torque measurements at 20, 50 (2 reps), and 60 mph, for a wide variety of s-cam brakes used for drive axle applications...... 18 Figure 2.2 Brake torque model output shown as a function of chamber pressure for various hub speeds. Solid lines (without symbols) indicate linear brake model output for 20, 25, 30, 40, 50, and 60 mph. Dashed lines (with symbols) indicate experimental data at an equivalent axle load of 13,200 lb for three test speeds (20, 50, 60 mph)...... 20 Figure 2.3 Comparisons of trailer axle brake torque models to experimental data. Output from the linear model is shown (with experimental data) in the top panel; output from the quadratic model (with the same experimental data) is shown in the bottom panel. The low-pressure linear fit is included in the quadratic model, but not the linear model. Dashed lines with symbols indicate the experimental data...... 24 Figure 2.4 Trailer axle drum brake model error, in percent of test output, for linear (solid lines) and quadratic (broken lines) models at 20, 50 and 60 mph. . 25 Figure 2.5 Steer axle drum brake model error, in percent of test output, for linear (indicated by solid lines) and quadratic (dashed lines) models at 20, 50, and 60 mph...... 25 Figure 2.6 Quadratic model output for drum brakes. Steer axle brake model is in the top panel; drive axle brake model is in the bottom panel. Dashed lines with symbols indicate the experimental data...... 27 Figure 2.7 Quadratic model output for disc brakes. Steer axle brake model (based on experimental data at 12,000-lb GAWR, actuated by a 20-in2 brake chamber) is in the top panel; drive axle brake model (based on
xv experimental results at 23,000-lb GAWR, actuated by a 30-in2 brake chamber) is in the bottom panel. Dashed lines with symbols indicate the experimental data...... 29 Figure 2.8 Steer axle drum brake torque measurements at 20 (top panel), 50 mph (middle panel), and 60 mph (bottom panel)...... 32 Figure 2.9 Drive axle drum brake experimental measurements 20 (top panel), 50 mph (middle panel), and 60 mph (bottom panel)...... 33 Figure 2.10 Air disc brake torque measurements at 20 (top panel), 50 mph (middle panel), and 60 mph (bottom panel)...... 35 Figure 2.11: Brake dynamometer experimental output showing step input response to the brake pressure command for a steer axle s-cam drum brake and a drive axle s-cam drum brake. Traces are brake chamber pressure (right axis) and measured torque (left axis)...... 40 Figure 2.12 Brake dynamometer simulation and experimental data: response to a step input in command pressure. Top panel: pressures, simulated command (treadle), simulated chamber, experimental chamber, with simulated speed output. Bottom panel: simulated torque output, experimental dynamometer torque output...... 42 Figure 2.13 Brake dynamometer simulation. Top panel: pressures, simulated command (treadle), simulated chamber, simulated dynamometer speed (ft/s). Bottom panel: simulated brake torque output...... 43 Figure 3.1 Schematic illustration of s-cam drum brake assembly showing crucial working parts [12]...... 53 Figure 3.2 Brake Torque model output shown as a function of chamber pressure for various application speeds. Solid lines (without symbols) indicate brake model output for 20, 25, 30, 40, 50, and 60 mph. Dashed lines (with symbols) indicate experimental data at an equivalent axle load of 20,000 lbs for three test speeds (20, 50, and 60 mph). The thicker gray line models the low-pressure torque response for all wheel speeds (Pc < 20 psi)...... 55 Figure 3.3 Time response of the pressure chamber model [top panel] and brake torque model [bottom panel] to a theoretical step input in P(treadle) at t=0 seconds to 80 psi, then back to 0 psi at t=2 seconds. Experimental dynamometer torque data are included in the bottom panel for comparison...... 57 Figure 3.4 Brake Dynamometer Simulation. Top panel: Simulated command (treadle) and chamber pressures (psi), and simulated dynamometer speed (ft/s). Bottom panel: Simulated brake torque output (lb-ft)...... 58 Figure 3.5 Dynamometer output and estimated hysteresis loop resulting from application and release of an S-cam brake...... 62 Figure 3.6 Simulation output showing the effect of hysteresis level on the response to a theoretical step input in treadle (reference) pressure...... 62 Figure 3.7 Increase in simulated stopping distance on a wet surface (broken line) and dry pavement (solid line) (µ=0.40 and µ=0.75, respectively), showing a
xvi significant effect on stopping distance resulting from adjusting the level of hysteresis in the drum brake models...... 63 Figure 3.8 Normalized lateral and longitudinal traction coefficients versus longitudinal slip ratio magnitude for a steer axle truck tire at rated load. Experimental measurements were taken at 0°, 2°, and 4° lateral slip angles...... 65 Figure 3.9 ABS control logic demonstration, showing (from top) wheel slip, “slip optimum” trigger, “slip high” trigger, wheel tangential acceleration, “accel. high” trigger, ABS logic output, and chamber pressure (Pc)...... 70 Figure 3.10 Experimental brake pressures (top panel) and longitudinal slip levels (bottom panel) for one wheel position from the steer, drive, and trailer axle. Note the successive time delays for the brake signal to reach each axle. All data have been low-pass filtered at 2 Hz...... 72 Figure 3.11 Simulation output showing treadle pressure with wheel chamber pressure (top panel), wheel slip (middle), and ABS control outputs (bottom) for steer axle (L). Maneuver is a straight-ahead stop from 30 mph on a wet Jennite surface, loaded to GVW...... 74 Figure 3.12 Experimental data showing treadle pressure with wheel chamber pressure (top) and wheel slip (bottom) for steer axle (L). Maneuver is a straight- ahead stop from 30 mph on a wet Jennite surface, loaded to GVW...... 75 Figure 3.13 Simulation output showing treadle pressure with wheel chamber pressure (top panel), wheel slip (middle), and ABS control outputs (bottom) for the leading drive axle (L). Maneuver is a straight-ahead stop from 30 mph on a wet Jennite surface, with no load...... 76 Figure 3.14 Experimental data showing treadle pressure with wheel chamber pressure (top) and wheel slip (bottom) for the leading drive axle (L). Maneuver is a straight-ahead stop from 30 mph on a wet Jennite surface, with no load. 77 Figure 3.15 Showing recent testing with an unloaded tractor with 48-ft. van trailer on a wet Jennite surface, stopping from 30 mph. This experimental data shows the different control characteristics for the tractor drive axle ABS controllers (Meritor-Wabco D-Type 4s/4m)...... 78 Figure 3.16 Simulation output showing treadle pressure with wheel chamber pressure (top panel), wheel slip (middle), and ABS control outputs (bottom) for the leading trailer axle (L). Maneuver is a straight-ahead stop from 30 mph on a wet Jennite surface, with no load...... 80 Figure 3.17 Experimental data showing treadle pressure with wheel chamber pressure (top) and wheel slip (bottom) for the leading trailer axle (L). Maneuver is a straight-ahead stop from 30 mph on a wet Jennite surface, with no load...... 81 Figure 3.18 Showing recent testing with an unloaded tractor with 48-ft. van trailer on a wet Jennite surface, stopping from 30 mph. The experimental data shows the different control characteristics for the tractor and trailer ABS controllers (here, Meritor-Wabco D-Type 4s/4m, Eaton 2000 4s/2m,
xvii respectively). Note also the longer release time for the trailer brake, due to lag and higher system restriction...... 82 Figure 3.19 Simulation (left) versus experimental (right) output illustrating the build- then-hold algorithm during a stop from 30 mph (48.3 kph) on wet pavement (µ=0.375) at GVW load. Brake pressure, wheel slip, and tangential velocity are shown for the steer axle...... 84 Figure 3.20 Simulation (left) versus experimental (right) output illustrating the build- then-hold algorithm during a stop from 30 mph (48.3 kph) on wet pavement (µ=0.375) at GVW load. Brake pressure, wheel slip, and tangential velocity are shown for the drive axle...... 85 Figure 3.21 Simulation (left) versus experimental (right) output illustrating the build- then-hold algorithm during a stop from 30 mph (48.3 kph) on wet pavement (µ=0.375) at GVW load. Brake pressure, wheel slip, and tangential velocity are shown for the semitrailer axle...... 86 Figure 3.22 Simulation (left panel) and experimental data (right panel) for drive axle (L), 30 mph (48.3 kph), on wet Jennite, loaded to GVW. ABS control is via a Meritor Version C 4s/4m controller...... 88 Figure 3.23 Repeat of simulated drive axle (L), 30 mph (48.3 kph), on wet Jennite, loaded to GVW (left panel) compared to experimental data (right panel) for leading drive axle (L & R sides) at 30 mph (48.3 kph), on wet Jennite, loaded to GVW. Tractor ABS control is via a Wabco-Meritor Version D 4s/4m controller. Note that , with Wabco-Meritor controller, Pc behavior is closer to the simulation...... 89 Figure 3.24 Simulated steer axle (L), 30 mph (48.3 kph), on wet Jennite, 0 payload (left panel), compared to experimental data (right panel) for steer axle (L), 30 mph (48.3 kph), on wet Jennite, 0 payload...... 90 Figure 3.25 Simulated (left panel) lead drive axle (L), 50 mph (80.5 kph), on dry concrete, loaded to GVW compared to experimental data (right panel) for lead drive axle (L), 50 mph (80.5 kph), on dry concrete, loaded to GVW...... 91 Figure 3.26 Simulated leading semitrailer axle (left panel), 30 mph (48.3 kph), on wet Jennite, loaded to GVW compared to experimental data (right panel) for leading semitrailer axle (L), 30 mph (48.3 kph), on wet Jennite, loaded to GVW. ABS control is via a Midland-Grau (currently Haldex) 2s/1m controller...... 92 Figure 3.27 Simulated leading drive axle (left panel), 50 mph (80.5 kph), on dry concrete, loaded to GVW compared to experimental data (right panel) for leading drive axle (L), 50 mph (80.5 kph), on dry concrete, loaded to GVW...... 93 Figure 3.28 Simulink® model for the enabled subsystem representing each of the modulator controlled brake chambers. Treadle (reference) pressure and previous chamber pressure are imported. Chamber pressure is exported to the model. This module cannot be multiplexed...... 94
xviii Figure 3.29 Simulink® model for brake torque generation. Multiplexed chamber pressures are imported. Multiplexed brake torques are exported. Coefficients for brake torque calculation are adjusted for brake application speed...... 95 Figure 3.30 Simulink® model for the 2s/2m ABS controller used on the trailer tandem for the simulations. Multiplexed wheel velocities and acceleration rates, with vehicle CG longitudinal velocity and acceleration, are imported. ABS logic commands of –1 (dump pressure), 0 (hold pressure constant), or +1 (allow pressure to build following Pt) are exported to the model. .. 96 Figure 4.1 Top panel: µ-level output for the variable coefficient simulator. Bottom panel: vehicle speed in kph during the same maneuver...... 107 Figure 4.2 Close-up of variable µ computer from the simulation...... 108 Figure 4.3 Excerpt from TruckSim™ animator showing a typical jackknife event with the ABS off for the tractor and ABS on for the semitrailer. The trailer appears to remain on the tangent for the curve. Simulation: file 776, “Disc/Drum, ½ treadle, ½ GVW load, low µ, ABS OFF.”...... 112 Figure 4.4 Excerpt from TruckSim™ animator showing a typical non-jackknife event with the ABS off for both the tractor and semitrailer. The entire vehicle appears to remain on the tangent for the curve. Simulation: file 775, “Disc/Drum, ½ treadle, ½ GVW load, low µ, ABS OFF.”...... 112 Figure 4.5 Excerpt from TruckSim™ animator showing a typical non-jackknife event with the ABS on for both the tractor and semitrailer. The entire vehicle remained within the prescribed lane during the stop. Simulation: file 774, “Disc/Drum, ½ treadle, ½ GVW load, low µ, ABS ON.” ...... 113 Figure 4.6 Another excerpt from same TruckSim™ animator file shown above, which shows a typical non-jackknife event with the ABS on for both the tractor and semitrailer. This view better shows the maintenance of lane position by the entire vehicle. Simulation: file 774, “Disc/Drum, ½ treadle, ½ GVW load, low µ, ABS ON.”...... 113 Figure 4.7 Phase plane diagram showing relationships between model state variables during a maneuver in which the hitch articulation angle remained under control, i.e., no jackknife occurred. The conditions for this simulation were: full treadle brake application, no ABS (either vehicle), medium traction (µ=0.55), and air-drum configuration. The boxes on the state plots are arbitrary operation limits on the respective states...... 115 Figure 4.8 Phase plane diagram showing relationships between model state variables during a maneuver in which the hitch a jackknife occurred. The conditions for this simulation were: half treadle brake application, no ABS (either vehicle), medium traction (µ=0.55), and air-drum configuration. All state traces exceed the same limits applied to the non- jackknife example in Figure 4.6...... 116 Figure 5.1 Schematic of the 3-axle planar model as derived ...... 122 Figure 5.2 Virtual displacement of the model in the direction δψ...... 138 Figure 5.3 Virtual displacement of the model in the direction δγ ...... 139
xix Figure 5.4 Measured normalized cornering forces for a drive axle truck tire in free rolling (i.e., no braking slip) ...... 150 Figure 5.5 3-axle planar model response to sinusoidal steer input...... 155 Figure 5.6 3-axle model response to 1 degree step input...... 156 Figure 5.7 3-axle planar model response to a 2-degree step input...... 157 Figure 5.8 3-axle planar model response to a slow ramp input...... 158 Figure 5.9 Schematic of the 5-axle planar truck model...... 160 Figure 5.10 Comparison of 3- and 5-axle linear model responses to the reference for sinusoidal steer input...... 165 Figure 5.11 Comparison of 3- and 5-axle linear model responses to the reference for a 1-degree step input...... 166 Figure 5.12 Comparison of 3- and 5-axle linear model responses to the reference for a 2-degree step steer input...... 167 Figure 5.13 Comparison of 3- and 5-axle linear model responses to the reference for a 3-degree step steer input...... 168 Figure 5.14 Comparison of 3- and 5-axle linear model responses to the reference for a 4-degree step steer input...... 169 Figure 5.15 Comparison of the 3- and 5-axle linear model responses to a ramp steer input...... 170 Figure 5.16 Linear model response compared to the reference (for 2 values of tire response lag) for a ± 2-degree amplitude sinusoidal steer input...... 172 Figure 5.17 Linear model response compared to the reference (for 2 values of tire response lag) for a 2-degree step input to steer angle...... 173 Figure 5.18 Linear model response compared to the reference (for 2 values of tire response lag) for a 4-degree step input to steer angle...... 174 Figure 5.19 Linear model response compared to the reference (for 2 values of tire response lag) for a 4.5-degree step input to steer angle...... 175 Figure 5.20 State-model response to the same path-following steering input showing inaccuracies at the point of jackknife (occurs at t ≈ 6.5 seconds). The “input” and TruckSim™ comparison data are from Run #836...... 176 Figure 5.21 Comparison of linear gain (left) and linear gain with saturation (right). 177 Figure 5.22 Experimentally measured and modeled truck tire free rolling cornering force for all five axle applications at a single load. The model is labeled “combinator” in the legend...... 184 Figure 5.23 Tire lateral force at various lateral slip angles, modeled as a function of longitudinal wheel slip...... 185 Figure 5.24 Nonlinear model response to sinusoidal steer input. The tire vertical load scale factors have been adjusted ± 10% for comparison...... 187 Figure 5.25 Nonlinear model response to 1-degree step steer input. The tire vertical load scale factors have been adjusted ± 10% for comparison...... 188 Figure 5.26 Nonlinear model response to 2-degree step steer input. The tire vertical load scale factors have been adjusted ± 10% for comparison...... 189 Figure 5.27 Nonlinear model response to 4-degree step steer input. The tire vertical load scale factors have been adjusted ± 10% for comparison...... 190 xx Figure 5.28 Nonlinear model response to 4.4-degree ramp steer input. The tire vertical load scale factors have been adjusted ± 10% for comparison...... 191 Figure 5.29 5-axle nonlinear model response to steering input that resulted in a jackknife...... 193 Figure 6.1 Schematic of a linear state observer...... 199 Figure 6.2 Flow diagram illustrating the implementation of the Kalman Filter...... 206 Figure 6.3 Schematic for the Extended Kalman Filter operation...... 210 Figure 6.4 Optimal R: initial 2.75 seconds shown for each state...... 217 Figure 6.5 Optimal R: initial 2.75 seconds shown for each state, along with model parameters and values for R...... 218 Figure 6.6 Optimal R: tire lateral slip angles forces shown for the entire run...... 219 Figure 6.7 Phase plot for case of optimal R...... 220 Figure 6.8 Low R – initial 2.75 seconds shown for each state...... 221 Figure 6.9 Low R – initial 2.75 seconds shown for each state, along with model parameters and values for R...... 222 Figure 6.10 Low R – tire lateral slip angles forces shown for the entire run...... 223 Figure 6.11 Phase plot for condition of very low R...... 224 Figure 6.12 High R...... 225 Figure 6.13 High R...... 226 Figure 6.14 High R. tire lateral slip angles forces shown for the entire run...... 227 Figure 6.15 State phase plot for very high values of R...... 228 Figure 6.16 Output showing EKF states and steering input signals...... 231 Figure 6.17 Showing model states and configuration parameters...... 232 Figure 6.18 Showing EKF states of lateral slip and tire forces. Note that the bold lines are for the EKF, and the lighter lines are of the TruckSim™ simulation...... 233 Figure 6.19 Vehicle forward speed and hitch angle phase plot, showing very good agreement in the state phase plane plot for hitch articulation angle, γ. .. 234 Figure 6.20 Steer angle input along with EKF states, showing very good agreement with the comparison TruckSim™ model...... 235 Figure 6.21 Steer angle input, model states of lateral velocity and yaw rate, along with model parameter settings...... 236 Figure 6.22 Note the good agreement for model tire lateral slip angles, αi, and tire lateral forces, Fyi...... 237 Figure 6.23 Vehicle forward speed and hitch angle phase plot. Note the good agreement in state phase plots for hitch articulation angle, γ...... 238 Figure 6.24 Steer input and model states. Note that the EKF had some difficulty tracking on this low-µ surface with the high-frequency forces of the ABS system...... 239 Figure 6.25 Same run, note change in load scale factors and slight increase in µ level used by EKF to get optimal agreement...... 240 Figure 6.26 Tire lateral slip and lateral force output from EKF and TruckSim™. Note the unusually high lateral slip angles that the EKF modeled for the trailer, in spite of the lateral forces for the trailer being modeled quite accurately.
xxi This behavior indicates that saturation in Fy has been exceeded by the semitrailer axles, i.e., they are sliding...... 241 Figure 6.27 Vehicle forward speed and state phase plot for hitch articulation angle, γ...... 242 Figure 6.28 Steer angle input and model states shown below steering input...... 243 Figure 6.29 Model states and run conditions...... 244 Figure 6.30 Lateral slip angles and tire lateral forces for both EKF and TruckSim™ models; bold lines indicate EKF output...... 245 Figure 6.31 Vehicle longitudinal speed and state phase plot for hitch angle, γ...... 246 Figure 6.32 Steer angle input and EKF model states...... 247 Figure 6.33 Model states and EKF running parameters...... 248 Figure 6.34 Tire lateral slip angles and tire lateral forces...... 249 Figure 6.35 Simulation vehicle longitudinal speed and state phase plot for hitch angle (γ)...... 250 Figure 6.36 Same run as previous, but EKF run @ ½ load instead of 0 load...... 252 Figure 6.37 Same run, but EKF run @ ½ GVW load, not 0-payload...... 253 Figure 6.38 Same run, but EKF run @ ½ GVW load, instead of 0-payload...... 254 Figure 6.39 Same run, but EKF run @ ½ -GVW load instead of proper 0-payload.. 255 Figure 6.40 Steer angle input and EKF states...... 256 Figure 6.41 Steer angle input, model states, and EKF parameters...... 257 Figure 6.42 Tire lateral slip angles and lateral forces...... 258 Figure 6.43 Vehicle forward speed and hitch angle phase plot. Note the good agreement in the state phase plot for γ...... 259 Figure 6.44 Steer angle input and EKF states...... 261 Figure 6.45 Showing steer input, model states, and EKF parameters...... 262 Figure 6.46 Tire lateral slip angles and forces...... 263 Figure 6.47 Vehicle forward speed and state phase plane plot for hitch articulation angle, γ, showing good agreement between EKF and TruckSim™...... 264 Figure 6.48 Same run, load and inertias scaled to 1.0...... 266 Figure 6.49 Same run, load and inertia scaled to 1.0...... 267 Figure 6.50 Same run, using load and inertia scales at 1.0...... 268 Figure 6.51 Similar conditions as previous, lower speed entry speed...... 270 Figure 6.52 Similar conditions, but slightly lower speed. Note the very high dynamic loading (high frequency dynamics) in the tractor tire lateral forces, as simulated by TruckSim™...... 271 Figure 6.53 Similar conditions, slightly lower speed...... 272 Figure 6.54 Steer angle input and EKF states...... 274 Figure 6.55 Model states and EKF parameter settings...... 275 Figure 6.56 Showing the EKF outputs (BOLD lines) and TruckSim™ (narrow lines) for tire lateral slip angles and forces...... 276 Figure 6.57 Vehicle forward speed and phase plane plot for hitch articulation angle, γ...... 277 Figure 6.58 Steer angle input and EKF states during a DLC while in a curve. The vertical line @ t=6 seconds indicates the onset of system instability..... 279 xxii Figure 6.59 Model states and EKF parameters for DLC in a curve...... 280 Figure 6.60 Tire lateral slip angles and forces during DLC in a curve...... 281 Figure 6.61 State phase plot and model speed...... 282 Figure 6.62 Time traces and state phase plot for hitch angle (γ) and its rate of change, showing the points of EKF detection of instability (“A”) and loss of control (“B”). The data were presented in Section 6.10...... 284 Figure 6.63 Time traces and state phase plot for hitch angle (γ) and its rate of change, showing the points of EKF detection of instability (“A”) and loss of control (“B”). The data were presented in Section 6.12...... 285
xxiii
LIST OF TABLES
Table 1.1 Initial Brake-in-Turn Simulation Study Results ...... 10 Table 2.1 Dynamic Steer Axle Loads from Best-Effort Braking Simulations ...... 34 Table 2.2 Total Work and Average Power Consumed per Brake During a Stop ..... 38 Table 2.3 Experimental Brake Fade, in Terms of Torque Decrease, for Various Brake Types ...... 41 Table 2.4 Model Fit Coefficients for the Steer Axle Brake Torque Model...... 44 Table 2.5 Model Fit Coefficients for the Drive Axle Brake Torque Model ...... 44 Table 2.6 Model Fit Coefficients for the Trailer Axle Brake Torque Model ...... 44 Table 2.7 Model Fit Coefficients for the ADB Brake Torque Model...... 44 Table 3.1 Hysteresis and Delay Time Parameters for Brake Models, ECBS-Disc and Air-Drum Configurations...... 60 Table 3.2 Brake Chamber Time Constant Values for Brake Model, ECBS-Disc and Air-Drum Configurations...... 60 Table 3.3 ABS Logic Truth Table ...... 67 Table 4.1 Initial Brake-In-Turn Simulation Study Results...... 102 Table 4.2 Simulation Entry Speeds Determined from Maximum Drive–Through Speeds ...... 104 Table 4.3 Jackknife Stability Simulation Result Matrix – Phase I: Full ABS and No ABS, Either Vehicle...... 110 Table 4.4 Jackknife Stability Simulation Result Matrix – Phase II: No ABS on Tractor, ABS Fully Functional on Semitrailer...... 110 Table 5.1 Physical Parameters Used for the 3-Axle Articulated Vehicle Model at the ½ GVW Load Condition...... 152 Table 5.2 Physical Parameters Used for the 5-Axle Articulated Vehicle Model at the ½ GVW Load Condition...... 164
xxiv CHAPTER 1
INTRODUCTION TO THE PROBLEM OF MODELING HEAVY TRUCK BRAKING SYSTEMS AND PREDICTING JACKKNIFE STABILITY
1.1 Abstract and Motivation
The motivation for this dissertation comes primarily from questions regarding heavy truck stability, rising from interest in government and private industry in reducing the stopping distances for heavily loaded commercial Class 8 trucks.
Reducing the stopping distance of a given mass at a given speed involves simply increasing the stopping forces – and therefore power dissipated during the stop. This study sought to answer two questions:
1. would significant increases in braking forces result in increased jackknife propensity for articulated tractor-semitrailer rigs, and
2. would such an event be detectible using currently available on-board vehicle dynamics measurement instrumentation, without the need to monitor articulation angle explicitly?
This extensive study shows that the simulated presence of Electronically Controlled Braking System with pneumatically actuated disc brakes (ECBS-disc) on the tractor results in no significant degradation of the performance of the rig in terms of jackknife stability while braking in a turn. Furthermore, elaborate simulations of vehicles
1 equipped with disc brakes and electronically controlled brakes systems (ECBS) show significant reduction in the tractor maximum yaw rate and hitch articulation angle.
These studies were conducted by simulating a brake-in-turn maneuver for a tractor-semitrailer at the speed corresponding to 90% of the theoretical maximum lateral acceleration in the constant radius curve that would allow a drive-through without the vehicle leaving the 3.66 m (12-ft) lane. All simulations were conducted on a 152.4 m (500 ft) radius corner with the mean friction coefficient levels set at either µ=0.30 or µ=0.55.
1.2 Problem Background
1.2.1 The Operation of Pneumatically Controlled Brakes on Tractor-Semitrailer Vehicles
The vast majority of heavy trucks on the road today have simple but proven pneumatic drum brakes. Most – or all – use a pneumatic low-flow-volume control line that delivers the command signal (from the “treadle valve” at the brake pedal) to activate the brakes. This command signal is usually terminated at each axle or tandem into a modulator valve, which allows flow of the actuating air to the brake assembly. The brake is then actuated via pneumatics from a local reservoir. One problem that exists with pneumatic control is that for today’s longer trailers, up to 16.15 m (53 ft.) in length, changes in the pneumatic brake control signal could take 400 ms or longer, even for a properly maintained system, to reach the semitrailer tandem. That delay alone translates to an additional 27 m (89 ft) covered at 96.6 kph (60 mph) before the trailer brakes – approximately 40% of the braking power on a five-axle rig – actuate. Figure 1.1 is a schematic of the most prevalent heavy truck brake used in North America today [1].
2
Figure 1.1 Schematic of the actuation mechanism of a pneumatically operated s- cam drum brake
1.2.2 Electronic Controlled Braking Systems (ECBS)
One solution to the problems presented by pneumatic control is the advent of Electronic Controlled Braking Systems (ECBS), which use electrical signals to “fire” solenoids that are located near the brake chamber, thus minimizing delays associated with the pneumatic control signal. The utility and flexibility of using electronic control of pneumatic braking in heavy trucks (ECBS) is well documented [2, 3] and has been proven by the general acceptance, from industry as well as regulatory entities, of these systems in Europe [4]. Market penetration in North America, however, remains sparse.
Although the implementation of ECBS in some tractor/trailer combinations has enjoyed good press, its full contribution to the dynamic stability of heavy trucks is yet to be realized. Benefits currently appearing in some prototypes and a few production vehicles in North America are:
1. the ability to control each brake separately for the purposes of traction control and vehicle stability control,
3 2. the ability for dynamic roll stability control, especially for double and triple semitrailer combinations [5], and
3. the ability of brake-by-wire systems to provide the vehicle operator with realistic, repeatable force feedback through the brake pedal, similar to that of a properly designed hydraulic braking system.
The research in this dissertation addresses a primary concern in the area of vehicle dynamics that accompanies the implementation of ECBS on new tractors. One question that this research was intended to answer is how the elimination of the “treadle signal” delay to the tractor, while substantial delay remains on the traditionally controlled semitrailer braking system, will affect jackknife stability for brake-in-turn maneuvers.
1.2.3 Compatibility of Vehicles Equipped with ECBS and Those Equipped with Pneumatically Controlled Brakes
Fleet operated tractors have a typical useful life of about three years, versus ten or more years for trailers. Fleet owners do not purchase or operate these rigs as single vehicles, but “mix and match” tractors and trailers of infinite variety. Therefore, fleet consumers of heavy trucks need interchangeability between “modern” ECBS-controlled tractors or semitrailers with “traditional” tractors or semitrailers, equipped with pneumatically controlled braking systems, and vice-versa.
The treadle signal, which traditionally controls the modulator valves on tractor- semitrailer rigs, is a pneumatic signal, having a pressure range of 0 to 7.58 bar (110 psi ). For an EBS system, the pneumatic signal is replaced by an electronic control signal, which varies line voltage, current, or transmission frequency to signal the brake modulator valve to control the amount of braking effort demanded by the driver and/or brake controller.
ABS control for tractor-semitrailer vehicles is currently required by law in the U.S. (the roll-in began for prime movers in March 1997) [6]. In order to be flexible for tractor-semitrailer interchangeability, ABS controls for current production tractors and semitrailers are not electronically coupled. Hence, the ABS unit for the tractor is
4 separate, and it operates separately from that for the semitrailer. The same is true for semitrailer ABS systems, in that they measure wheel speeds, estimate semitrailer velocities, and operate the brakes independently of the ABS unit on the tractor. The only ABS interconnection between the tractor and semitrailer is the ABS power circuit, which provides electrical power only to the semitrailer ABS controller. Therefore, both ABS controllers not only work independently, but can fail independently as well.
1.2.4 Pneumatic Disc Brakes on Class 8 Trucks in the United States
Although offered by some heavy truck brake manufacturers, pneumatically actuated disc brakes (ADB) have enjoyed a very slow infiltration into the North American market, now accounting for only a few percent of the brakes supplied with new prime movers. However, disc brakes are standard equipment on all Class 8 trucks currently sold in Europe [7]. Although passenger cars are federally required to come to a stop in 65.8 m (216 ft) from 96.6 kph (60 mph) [8], most can accomplish 45.7 m (150 ft) stops on dry pavement. The current requirement for heavy trucks, loaded to tractor GVW, is 108.2 m (355 ft) [9]. However, most modern rigs can stop from 96.6 kph (60 mph) within 91.4 m (300 ft). The long-term goals for NHTSA include revising the FMVSS 121 requirements such that heavy trucks will be required to stop from (96.6 kph) 60 mph within 76.2 m (250 ft), optimally as low as 61.0 m (200 ft) [10].
A significant contributor to these improvements in stopping distance compatibility between loaded tractor-semitrailers and light vehicles is the implementation of modern air disc brakes (ADB). One physical advantage of disc brakes is that the inevitable expansion resulting from thermal stress causes the ferrous brake surface to expand; this is disadvantageous for drums, taking the braking surface away from the pad, but better for discs, forcing the metal disc into the pad. Another performance improvement may result from the significant reduction in operating hysteresis seen with the less flexible air disc brake design. Recent studies have shown that for a given level of ABS-controlled braking power, a system hysteresis level of 10% of full scale capability can result in a 20% increase in wet stopping distances for a loaded truck versus a system
5 with negligible hysteresis [11]. This dissertation presents a theory suggesting that reduction of this phenomenon is an important contributor to improved ABS-controlled stopping distances.
Manufacturers of commercially available disc brakes tout a potential increase in effective braking torque of 20% over existing drum brakes. Other manufacturer’s claims include:
1. Meritor claims 16% shorter stopping distance (40 ft. off of a 250-ft. stop) with their “disc Plus” disc brake system,
2. The Mercedes-Benz Actros claims a 30% improvement in stopping performance due to the use of a prototype ECBS/disc system at all brake positions, and
3. Bendix claims a 10% weight savings and 30% stopping distance improvement using their disc brakes, as well as 1,000,000-mile disc life and 500,000-mile pad life for highway use [12].
My study at NHTSA’s Vehicle Research and Test Center (VRTC) of contemporary heavy truck brakes revealed that the performance gains from disc brakes are not necessarily from the availability of significantly higher ultimate torque, but instead from the consistency of their brake torque generation with respect to application speed. Figures 1.2 and 1.3 illustrate the differences in brake torque generation for pneumatic drums and ADB, respectively. The drive axle drum brakes shown in Figure 1.2 show much more sensitivity to application speed than do the drive axle disc brakes shown in Figure 1.3. For the drum brake, the slope of the relationship of brake torque
(Tb) with respect to chamber pressure (Pc) decreases significantly, as does the maximum torque, with increasing application speed. The same is not true for the ADB results shown in Figure 1.3. Further information on these evaluations and the models resulting from them can be found in [13]. Other concerns regarding disc brake systems include initial and maintenance costs, which will eventually improve due to economies of scale.
The analytical models used in this study have been in use at NHTSA since 2000, when Ed Milich reported on the correlation between elaborate rigid-body simulation program output and experimental vehicle data for a broad series of maneuvers [14]. 6
Figure 1.2 Modeled and experimental data for drive axle drum brakes. Broken lines (with icons) indicate experimental data at a simulated axle load of 21,000 lb.
Figure 1.3 Modeled and experimental data for drive axle disc brakes at a simulated GAWR of 23,000 lb. Color and type schemes are identical to those of Figure 1.2.
7 1.3 Intent and Scope of This Research
1.3.1 Jackknife Instability During Brake-in-Turn (B.I.T.) Maneuvers
The purpose of the B.I.T. study was two-fold. The first goal was to develop advanced analytical models for electro-pneumatic disc braking systems (“ECBS-disc”) and traditional pneumatically controlled drum braking systems (“air-drum”) that surpass current public domain models for accuracy in simulating dynamic brake behaviors. The second goal was to use these in-house developed brake system models, in parallel with a sophisticated vehicle simulation package, to analyze the potential impact of advanced high-torque ECBS-disc equipped tractors on the jackknife stability of tractor-trailer rigs.
The vehicle simulation package used for this research was TruckSim™ version 5.0, by Mechanical Simulation Corporation of Ann Arbor [15]. The TruckSim™ software is a nonlinear solver that treats the vehicle chassis, suspension, and drivetrain masses as rigid bodies. This software package uses linear and nonlinear force and moment relationships to simulate the applied forces to the vehicle and internal forces between the vehicle components. The TruckSim™ software simulates the dynamics of the vehicle, including highly nonlinear aspects such as the tire force models, suspension deflection models, leaf spring models, and the hitch model. The brake system dynamics, brake torque outputs, and brake hysteresis were modeled with a nonlinear Simulink® model developed for use within NHTSA’s Vehicle Research and Test Center (VRTC) [16]. The brake simulation software received the brake system command from the vehicle simulation, then calculated brake chamber pressures and torques accordingly. The brake simulation software also simulated a 4s/4m ABS system for the tractor, and a separate 2s/2m system for the semitrailer.
The term “4s/4m” refers to the number of sensors (4) and modulator valves (also 4 for this system) that the ABS system uses to sense wheel speed, and then control brake pressure. A typical “4s/4m” system senses the front wheel speeds (2) and the leading drive axle hub speeds (2), for a total of 4 sensed wheel speeds. The designation
8 following the slash refers to the four modulators, which in this configuration control the steer axle brake chambers independently (requiring two modulators), then control each side of the drive tandem independently (requiring the other two modulators). Thereby both the leading and trailing axle brake position on each side of the drive tandem is controlled via one modulator valve.
This study, therefore, compares the ECBS-disc equipped tractor to an otherwise identical air-drum tractor, each coupled to the same simulated air-drum equipped semitrailer. The rigs are evaluated for jackknife stability under brake-in-turn maneuvers. Brake actuation is simulated to occur at either the onset of negotiating a curve, or after the vehicle is fully established in the curve, depending on simulation setup. The test parameters were derived by analyzing a prior study that used similar vehicle comparisons on simulated wet and dry surfaces at load conditions of GVW and no payloads [17]. The previous study concluded that:
1. disc brake equipped tractors should not exacerbate jackknife problems for tractor-semitrailer rigs in brake-in-turn maneuvers, regardless of ABS functionality, and
2. for disc or drum equipped tractors, the combination of vehicle configuration / simulation condition that proved most troublesome for jackknife stability was that of vehicles with low trailer loads maneuvering on low coefficient (µ) pavement surfaces.
Table 1.1 shows the results of the initial study.
9 No Load GVW Load Hi µ Low µ Hi µ Low µ Drum Disc / Drum Disc / Drum Disc / Drum Disc / / drum drum / drum drum / drum drum / drum drum Full ABS Slow ABS No
Half Treadle Half Treadle ABS Full ABS Slow ABS No
Full Treadle ABS
Indicates near jackknife (high hitch articulation angle), and/or high hitch forces Indicates jackknife
Table 1.1 Initial Brake-in-Turn Simulation Study Results
10 The following are explanations of the terms used in Table 1.1.
“Full ABS” indicates functioning ABS for both tractor and semitrailer, as per the vehicle simulation software, TruckSim™ version 4.6.
“Slow ABS” indicates simulated partial failure of the ABS system, on both the tractor and semitrailer, by increasing the slip threshold at which the ABS signals the brake chamber to dump pressure, and increasing the slip threshold at which the ABS signals the brake chamber to allow pressure to increase again.
“No ABS” indicates no simulated ABS operation on either the tractor (prime mover) or the semitrailer.
“Full Treadle” and “Half Treadle” correspond to the magnitude of control line pressure demanded by the driver after the vehicle was fully established in the 152.4 m (500 ft) radius corner. It is the experience and observation of this researcher that not all drivers apply full braking power after realizing that they may be going too fast for the traction level in a corner. Slow, cautious brake application is arguably the best tactic under some emergency situations.
“Drum/drum” refers to the traditional configuration of air-drum brakes on the tractor, and air-drum brakes on the trailer. “Disc/drum” refers to ECBS-disc brakes on the tractor, and air-drum brakes on the trailer.
Although experimental studies with disc brake equipped tractors and semitrailers are ongoing at NHTSA, the simulations conducted as part of this dissertation were intended to expand upon the limited number of test conditions available to test engineers. Not only is this dissertation interested in the effects that significantly increased braking torque would have on tractor-trailer stability (i.e., Will the increase in braking force on the tractor make it more susceptible to jackknifing due to higher kingpin force magnitudes?), but how will controllability be influenced by the response time improvements promised with ECBS and disc brakes?
11 1.4 Detection of the Jackknife Event While It Occurs
Detection – or even prediction – of the jackknife as it occurs would be quite useful for an on-board vehicle stability system, which could use the incoming information about an impending jackknife to initiate execution of a procedure to correct for the instability. Such on-board vehicle stability systems are becoming quite common in many passenger vehicles at the time of this dissertation. Vehicle stability systems currently available on production passenger vehicles mitigate vehicle yaw instability by manipulation of the vehicle’s brakes (independently) and/or throttle. A fully operational system for heavy Class 8 trucks is not only inevitable, but certain to save fleet owners millions of dollars in repair, medical, and legal costs each year.
Detection of the jackknife event would be quite easy if hitch articulation angle could be directly measured. However, such measurements become difficult when employing expensive and delicate angular measurement systems. More difficulty is encountered when the engineer tries to find a way to have the device fit and work on many different tractor-semitrailer rigs, without causing great inconvenience to the driver or maintenance crews.
A more elegant solution is to estimate hitch angle and rate, using other state variables that we can assume are measured by a vehicle stability / ABS system. In Chapter 5, a complete planar model for an articulated vehicle is developed “from the ground up,” then linearized to produce a four-state constant-speed model, having the states of:
v = lateral velocity
r = yaw rate
q = hitch articulation angle
γ = hitch articulation rate
12 The four-state model serves as the basis for a far more elaborate fifteen-state model that is necessary to accurately model the vehicle dynamics during the onset of instability.
The Kalman Filter [18] has proven to be a popular and very flexible state observer since its introduction by Kalman in 1961 [19]. Further utility with highly nonlinear systems exists in the Extended Kalman Filter, which expands the linear Kalman Filter equations into differential equations that are integrated with each time step. Inspiration for this approach, as well as a thorough discussion on the history and implementation of the Kalman and Extended Kalman Filters are covered by Chrstos in [20].
Chapter 6 of this dissertation discusses the development of an Extended Kalman Filter for the purpose of estimating hitch articulation angle and rate, based on state measurements that are assumed to be available in any vehicle stability system – steer angle (the input), vehicle lateral acceleration, and vehicle yaw rate.
13 1.5 Chapter 1 References
1. Figure from D. Yanakiev, J. Eyre, I. Kannellakopoulos, “Longitudinal Control of Heavy Duty Vehicles: Experimental Evaluation,” California PATH research report, UCLA Electrical Engineering, 1998. 2. Scania Trucks press release, (www.scanio.com/ms/events/press/wwwtxt/n97111en.htm). 3. Heavy Truck Magazine, Roemer Insurance, Inc., (http://www.roemer- insurance.com/rr22000.html). 4. “Technology Review for Electronically Controlled Braking Systems,” NHTSA, 1998. 5. “Mechanics of Heavy Truck Systems,” short course material, University of Michigan Transportation Research Institute (UMTRI), July 2000. 6. Q&A: ANTILOCK BRAKES: CARS, TRUCKS, MOTORCYCLES, Insurance Institute for Highway Safety, (http://www.hwysafety.org/safety_facts/qanda/antilock.htm). 7. Land Line Magazine (April 11, 2000). 8. FMVSS 135 braking standards, Oct. 2001. 9. FMVSS 121 braking standards, Oct. 2001. 10. NHTSA Vehicle Safety Rulemaking priorities 2002-2005, (http://www.nhtsa.dot.gov/cars/rules/rulings/PriorityPlan/Index.html#secIV). 11. A.L. Dunn, G. J. Heydinger, G. Rizzoni, and D. Guenther, “New Model for Simulating the Dynamics of Pneumatic Heavy Truck Brakes with Integrated Anti- Lock Control,” SAE 2003-01-1323. 12. Bendix Truck Brake System Advertisement, (http://www.bendix.com/products/SellSheet?p=ADB225). 13. A.L. Dunn, G. J. Heydinger, G. Rizzoni, and D. Guenther, “Empirical Models for Commercial Vehicle Brake Torque from Experimental Data,” SAE 2003-01-1325. 14. E. Milich, “An Evaluation of VDM-Road and VDANL Vehicle Dynamics Software for Modeling Tractor-Trailer Dynamics” (master’s thesis, The Ohio State University, 1999). 15. TruckSim version 5.0 User Manual (Ann Arbor, MI: Mechanical Simulation Corporation, 2003); and (http://www.trucksim.com). 16. A.L. Dunn, “Simulink Heavy Truck Brake Model Simulation Manual,” NHTSA / VRTC internal report, 2003.
14 17. A.L. Dunn, “Brake-In-Turn Study Comparing Disc/Drum to Drum/Drum Truck Brake Combinations,” NHTSA / VRTC internal report, 19 September 2000. 18. G. Welch and G. Bishop, “An Introduction to the Kalman Filter,” TR95-041, Department of Computer Science, University of North Carolina at Chapel Hill. 19. R.E. Kalman, “A New Approach to Linear Filtering and Prediction Problems,” ASME Journal of Basic Engineering, 1960. 20. Chrstos, Jeffrey P., “Use of Vehicle Dynamics Modeling to Quantify Race Car Handling Behavior” (Ph.D. Dissertation, The Ohio State University, 2000).
15 CHAPTER 2
EMPIRICAL MODELS FOR COMMERCIAL VEHICLE BRAKE TORQUE FROM EXPERIMENTAL DATA
2.1 Abstract
This chapter introduces a new series of empirical mathematical models developed to characterize brake torque generation of pneumatically actuated Class-8 vehicle brakes. The brake torque models, presented as functions of brake chamber pressure and application speed, accurately simulate steer axle, drive axle, and trailer tandem brakes, as well as air disc brakes (ADB). The contemporary data that support this research were collected using an industry standard inertial brake dynamometer, which is routinely used for verification of FMVSS 121 commercial vehicle brake standards.
2.2 Motivation
The goal for these analyses and model development was to accurately characterize brake torque versus chamber pressure and application speed at the hub, for use with multi-body vehicle dynamics simulations. Information on the topic of pneumatic brake torque output has not been published in this form since 1975 [1]. Since then, new developments, such as improved lining materials and air disc brakes (ADB), along with the availability of numerous advanced PC-based vehicle dynamics modeling 16 programs, have resulted in a renewed demand for accurate, publicly available information regarding brake torque output of pneumatically operated commercial vehicle brakes.
The data discussed herein are used to develop empirical mathematical models, which are implemented in the form of multiple quadratic relationships that relate torque output to a real-time dynamic brake chamber pressure model and tire tangential speed. These torque outputs are part of newly developed nonlinear mathematical models that use a systems modeling approach to simulate a pneumatic commercial heavy vehicle braking system (discussed in Chapter 3 and [2]). The newly developed brake system models covered in Chapter 3 and [2] use the torque relations discussed here, as well as additional relevant source information regarding hysteresis and response lag times inherent in the actuation of pneumatic-over-mechanical Class-8 vehicle brakes [1, 3-5].
2.3 Model Development in General
2.3.1 Model Limits and Assumptions
These models were intended to very accurately describe the dynamic behavior of commercial braking systems during stops when the temperature limits dictated in the FMVSS 121 test procedures were not exceeded. Also, these models were developed from experimental brake dynamometer data at application speeds from 20 to 60 mph. Hence, valid simulation results cannot be guaranteed when tire tangential speeds significantly exceed this speed range.
As seen in Figure 2.1, different linings or test conditions produce vastly different brake torque (Tb) versus chamber pressure (Pc) relationships. For each condition of speed
and test load (inertia), the curves representing Tb versus Pc can be estimated using a number of methods, such as a linear, quadratic, third order, or an exponential approximation.
17 Drive Axle Air Drum Brake Torques by Speed for Linings 'A,B,C,D,E' 25000
20000
15000
10000 Torque (ft-lbs)
5000
0 020406080100120 chamber pressure (psi) 20 mph 50 mph 60 mph Lining E 26k GAWR drive-1 20 mph 50 mph 60 mph Lining D 29k GAWR drive-2 20 mph 50 mph 60 mph Lining C 22k GAWR drive-3 20 mph 50 mph 60 mph Lining B 21k GAWR drive-3 20 mph 50 mph 60 mph Lining B 20k GAWR drive-4 20 mph 50 mph 60 mph Lining B 20k GAWR drive-4 20 mph 50 mph 60 mph Lining A 20k GAWR drive-3
Figure 2.1 Drum brake torque measurements at 20, 50 (2 reps), and 60 mph, for a wide variety of s-cam brakes used for drive axle applications.
For the purpose of modeling brake torque versus chamber pressure and speed, experiments conforming to the “Brake Retardation Force test” (NHTSA §5.4.1.1, FMVSS 121 standards) were conducted from initial speeds of 20, 50, and 60 mph (32.2, 80.5, and 96.6 kph). Brake chamber pressure was applied at a constant value of 20 psi (1.38 bar), and brake torque output measured, and averaged, during the stop. The stops were repeated as the constant chamber pressure was increased in 10-psi increments, up to 100 psi. Also, data from a modified version of the FMVSS “Brake Retardation Force test,” having an initial brake chamber pressure of 7 psi (0.48 bar), then 10 psi (0.69 bar), followed by 10-psi increments to 100 psi (6.89 bar), were used to verify the data at 50 mph (80.5 kph). The nature by which the supporting data were gathered should make it clear that these models are not intended for use in simulating the brake fade phenomenon
18 that occurs during long descents or repeated hard stops that would cause brake temperatures to exceed those mandated in FMVSS 121.
2.3.2 Linear Brake Torque Model
In previous formulations of the quasi-static relationship of torque to brake chamber pressure and speed, the models were postulated from experimental data to have
piecewise linear relationships of brake torque (Tb) versus chamber pressure (Pc) [1, 6]. The nominal slope of these relationships differed for each brake application speed, and generally decreased with increasing speed.
Beginning with steer axle brake torques, various mathematical models were applied, then compared, to determine which model most efficiently provided an acceptable estimate of the available experimental data. For the steer axle brakes, the relationship between Tb and Pc was relatively linear in nature, and a series of linear
equations were derived using linear regression of Tb with respect to Pc. at each application speed. With the assumption that the linear coefficients for each line fitting a
Tb versus Pc relationship at a single application speed (i.e., tire-wheel tangential speed at which the brake was applied) may have a quantifiable relationship to brake application speed, an empirical mathematical model was developed using linear regression to
describe the interaction of the coefficients of each Tb versus Pc linear relationship, as
functions of speed. For the steer axle brakes, the linear model of Tb versus Pc, whose coefficients vary linearly with respect to application speed, provided good results, with model error being within 6% over the entire range of application pressures. This linear model response is co-plotted in Figure 2.2 with actual brake torque data.
19 Steer Axle Air Drum Linear / Linear Brake Torque vs. Pressure for Various Speeds
9000
Vhub = 20 mph Vhub = 25 mph 8000
Vhub = 30 mph
7000 Vhub = 40 mph
Vhub = 50 mph 6000
Vhub = 60 mph 5000
4000
Brake Torque (ft-lb) Torque Brake 3000
2000
1000
0 0 20406080100 Pc = chamber pressure (psi)
20 25 30 40 50 60 20 mph test 50 mph test 60 mph test
Figure 2.2 Brake torque model output shown as a function of chamber pressure for various hub speeds. Solid lines (without symbols) indicate linear brake model output for 20, 25, 30, 40, 50, and 60 mph. Dashed lines (with symbols) indicate experimental data at an equivalent axle load of 13,200 lb for three test speeds (20, 50, 60 mph).
The linear torque versus pressure relationships for the entire pressure operating range are expressed in equation (2.1). For the pressure range of [7 < Pc ≤ 20 psi], torque output for all application speeds has been modeled using a single linear model between 0 psi and the convergent speed-averaged torques at 20 psi (P0). Brake torque is usually negligible below the “push-out pressure” (Ppo, usually around 7 psi), where brake chamber motion is prevented by stiction.
20 ≤≤ ⎧⎫0 0 PPcpo ⎪⎪ ⎪⎪TbP =<≤0 TPbc⎨⎬ PPP poc0 (2.1) ⎪⎪P0 ⎪⎪+<≤ ⎩⎭CCP01ccres PPP 0
TCP==+C TPP ( ) (2.2) bP0 010 b c 0
=+ ≤ Ccii01 cV ihub PP c po ≤≤ (2.3) 20Vmphhub 60
For equations (1-3):
Tb = torque available at the brake
C0, C1 = coefficients for linear torque relationship
Pc = brake chamber pressure (psi)
Pres = brake pneumatic pressure reservoir pressure, taken as constant, usually under 110 psi.
P0 = pressure at which the torque relationships converge, with respect to speed, here P0 = 20 psi.
TbPo = speed averaged torque at P0
Ppo = pop-out pressure, here Ppo = 7 psi. cij = speed regressed coefficients, i.e., Ci = f[cij, Vhub]
Vhub = tangential wheel speed (mi/hr)
21 2.4 Nonlinear Response of Some Brake Torques
After application of the aforementioned linear model to the trailer brake torque outputs, the agreement was less than satisfactory. In fact, for the trailer brakes for which information was available, the Tb versus Pc relationship was far less linear, and actually more quadratic in character.
Applying the procedure described above to a higher degree of equation, the Tb versus Pc relationship was first modeled at each test speed (20, 50, and 60 mph) as a
quadratic function having a non-zero offset at Pc = 0. The three coefficients for each
“constant-application-speed” Tb versus Pc relationship model were then trended with respect to speed, resulting in what is called here a quadratic model. The trailer brake models, showing both the linear and the quadratic relationships for comparison, are presented in Figure 2.3.
The quadratic model is described in equations (2.4) through (2.6), and is similar to equations (2.1) through (2.3) except for the additional higher degree term. Comparison of the two panels in Figure 2.3 shows the improvement in the model accuracy (at the speeds for which data exist) when the quadratic model is employed, as opposed to a linear model or combinations of “linear /quadratic” models (not shown). Note that the phrase “quadratic / quadratic” refers to both sets of the empirical model coefficients being modeled as quadratic responses (of pressure, then speed) instead of linear. The description “quadratic/quadratic” is used here interchangeably with “quadratic” for simplicity. The same usage applies to “linear/linear” models.
2 =≤j adjust only if : CcVi∑ ij hub PP c po (2.4) j=0
22 ⎧⎫ ⎪⎪0 0 ≤≤PP ⎪⎪cpo ⎪⎪T =<≤bP0 TPbc⎨⎬ PPP poc0 (2.5) ⎪⎪P0 ⎪⎪2 i <≤ ⎪⎪∑CPic( ) P0 P c P res ⎩⎭i=0
2 TCP==( )i TPP ( ) (2.6) bP0 ∑ i00 b c i=0
Consistent with figures comparing modeled versus experimental brake torque in this report, the dashed lines in Figure 2.3 represent brake dynamometer output (for the specific brake and lining type modeled) at 20, 50, and 60 mph. The solid lines represent the model output at 20, 25, 30, 40, 50, and 60 mph. On the bottom panel, the quadratic nature of the Tb versus Pc relationship for the trailer brakes is clearly shown.
Figure 2.4 contains the error curves (model error versus Pc) for both the linear and quadratic models for the trailer brakes. Error for the best fit linear model (shown using solid lines) ranges from –15% at low pressure to +7% at medium chamber pressure. The error for the quadratic model (shown using dashed lines) is much friendlier, at ±2.5%.
A quadratic/linear model (where the coefficients of the quadratic relationship between Tb and Pc are modeled linearly with speed) was tried, but the model fit did not improve satisfactorily (error was generally 5%, up to 15% for some speed-chamber pressure combinations).
23 Lin e ar M odels for T raile r Axle Air D rum B rake T orque v s. Pre ssure for Various S pe ed s
14000
12000
10000
8000
6000 Brake Torque (ft-lbi) 4000
2000
0 020406080100 Pc = chamber pre ssure (psi)
20 25 30 40 50 60 20 m ph test 50 mph test 60 m ph tes t
Quad ratic /Qua dratic M od els fo r T ra iler Axle Air D rum B rake To rq ue v s. P res sure for V ario us Sp eeds
14000
12000
10000
8000
6000 Brake Torque (ft-lbi) 4000
2000
0 020406080100 Pc = chamber pressure (psi)
20 25 30 40 50 60 20 m ph tes t 50 mph test 60 mph test 0-20 psi, all speeds
Figure 2.3 Comparisons of trailer axle brake torque models to experimental data. Output from the linear model is shown (with experimental data) in the top panel; output from the quadratic model (with the same experimental data) is shown in the bottom panel. The low-pressure linear fit is included in the quadratic model, but not the linear model. Dashed lines with symbols indicate the experimental data.
24 Quadratic and Linear Model Error Plot for Trailer Axle Brake (sheet: trailer_6)
10.0%
5.0%
0.0%
-5.0% Brake Torque ERROR (%) Brake Torque ERROR -10.0%
-15.0% 0 20406080100 Pc = chamber pressure (psi)
quad-20 quad-50 quad-60 linear-20 linear-50 linear-60
Figure 2.4 Trailer axle drum brake model error, in percent of test output, for linear (solid lines) and quadratic (broken lines) models at 20, 50 and 60 mph.
Quadratic and Linear Model Error Plot for Steer Axle Drum Brake for Steer Axle @ 13.2 k# GAWR
6.0%
4.0%
2.0%
0.0% 0 20 40 60 80 100 Brake Torque (%)) ERROR -2.0%
-4.0%
-6.0% Pc = chamber pressure (psi)
quad-20 quad-50 quad-60 linear-20 linear-50 linear-60
Figure 2.5 Steer axle drum brake model error, in percent of test output, for linear (indicated by solid lines) and quadratic (dashed lines) models at 20, 50, and 60 mph.
25 Figure 2.5 shows the improvement in model error when the steer axle model was converted to a quadratic model. The error magnitude for the steer axle drum brakes dropped from a high of over 6% for the “linear/linear” model to 2.5% for the “quadratic/quadratic” model. The improvement is significant despite the much more linear character of the steer axle brakes than that of the trailer brakes.
The results displayed in Figures 2.3, 2.4, and 2.5 show that the “quadratic/quadratic” model provides superior fit, as compared to the “linear/linear” model. The “quadratic/quadratic” model was therefore applied to each of the brake types analyzed, including drum brakes for steer (15” diameter by 4” wide drum – 381 x 102 mm), drive, and trailer axles (both drums at 16.5” x 7” – 419 x 178 mm), and 430 mm (16.9 inch) diameter disc brakes for steer and drive axles. Note that air disc brake models were regressed from data taken under two vastly different applications. To simulate use on the steer axle, the brake was tested at an equivalent gross axle weight rating (GAWR) of 12,000 lb (5,443 kg) and applied using a 20-in2 (129.0-cm2) brake chamber; for drive axle applications, the brake was tested at 23,000 lb (10,433 kg) GAWR and applied using a 30-in2 (193.5-cm2) brake chamber. An additional benefit of using one model for all brake types was convenience of implementation. Hence, the standard model equations can be applied to any brake type for a tractor-semitrailer vehicle, by varying only the nine coefficients for different brake types.
The two panels in Figure 2.6 show the quadratic model output for drum brakes on the steer and drive axles. The quadratic steer axle brake model is shown in the top panel; the quadratic drive axle brake model is shown in the bottom panel.
26 Quadratic /Quadratic Models for Steer Axle Air Drum Brake Torque vs. Pressure for Various Speeds (13.2 k# GAWR)
14000
12000
10000
8000
6000 Brake Torque (ft-lbi)
4000
2000
0 0 20406080100 Pc = chamber pressure (psi)
20 25 30 40 50 60 20 mph test 50 mph test 60 mph test 0-20 psi, all speeds
Quadratic /Quadratic Models for Drive Axle Air Drum Brake Torque vs. Pressure Model for Various Speeds
14000
12000
10000
8000
6000 Brake Torque (ft-lbi)
4000
2000
0 020406080100 Pc = chamber pressure (psi) 20 25 30 40 50 60 20 mph test 50 mph tes t 60 mph test 0-20 psi, all speeds Figure 2.6 Quadratic model output for drum brakes. Steer axle brake model is in the top panel; drive axle brake model is in the bottom panel. Dashed lines with symbols indicate the experimental data.
27 The two panels in Figure 2.7 show the output from the quadratic model for disc brakes. A disc brake model for the steer axle (based on experimental data taken at 12,000 lb GAWR, using a 20-in2 chamber) is shown in the top panel; a disc brake model for the drive axle (based on experimental data taken at 23,000 lb GAWR, using a 30-in2 chamber) is shown in the bottom panel. Note in Figure 2.7, that although the ultimate torque output (at Pc = 100 psi) for the disc brake (drive axle application) is no higher than that of the drive axle drum brakes of equivalent power capacity (both are near 14,000 lb- ft – 18.98 kN), sensitivity of the torque-pressure slope to hub speed is vastly reduced with the disc brake. Furthermore, it might be appropriate to apply a single second-order model for all speeds of the 23,000-lb GAWR disc brake application.
28 Quadratic /Quadratic Models for Air DISC Brake Torque vs. Pressure for Various Speeds on the Steer Axle (12 k# GAWR)
14000
12000
10000
8000
6000 Brake Torque (ft-lbi)
4000
2000
0 020406080100 Pc = chamber pressure (psi)
20 25 30 40 50 60 20 mph test 50 mph test 60 mph test 0-20 psi, all speeds
Quadratic /Quadratic Models for Air DISC Brake Torque vs. Pressure for Various Speeds on the Drive Axle (23 k# GAWR) 14000
12000
10000
8000
6000 Brake Torque (ft-lbi) 4000
2000
0 020406080100 Pc = chamber pressure (psi)
20 25 30 40 50 60 20 mph test 50 mph test 60 mph test 0-20 psi, all speeds
Figure 2.7 Quadratic model output for disc brakes. Steer axle brake model (based on experimental data at 12,000-lb GAWR, actuated by a 20-in2 brake chamber) is in the top panel; drive axle brake model (based on experimental results at 23,000-lb GAWR, actuated by a 30-in2 brake chamber) is in the bottom panel. Dashed lines with symbols indicate the experimental data.
29 2.5 Overall View of Brake Torque Test Data
Figures 2.8, 2.9, and 2.10 show torque outputs for ADB, steer axle drum, and drive axle drum brakes, respectively. Although the amount of information can be overwhelming, the figures are presented to illustrate the vast possibility of torque outputs from a properly functioning brake assembly. It can be concluded that the brake size, air chamber (actuator) size, and pad lining material have the most significant effects (at a given speed) on torque output, whereas the assembly type produces secondary effects on torque output for a new, properly operating brake assembly. Figure 2.1 is a compilation of the drive axle drum brake data, and again is presented only to illustrate the spectrum of response character and magnitude available for the same size and type of brake. In the interest of brevity, the trailer brake output is not presented in this form. Note that Figures 2.1, 2.8, 2.9, and 2.10 show two individual tests for the same conditions at 50 mph (80.46 kph).
2.6 Brief Discussion on Dynamic Axle Weights and Their Impact on These Models
Dynamic axle loads can vary significantly around the static load during braking and cornering maneuvers. This natural response of the vehicle dynamics might suggest that the brake torque model output be adjusted to the changes in load that the axle experiences as a result of dynamic forces on the entire vehicle.
Thus, vehicle dynamics simulation outputs were reviewed to summarize the range of vertical loads that could be experienced on an axle during straight-ahead braking, and high-speed cornering under braking. The following results were extracted from simulations on dry asphalt (µ ≈ 0.75). The simulations have been vigorously verified using experimental vehicle data [7, 8], and were run using the TruckSim™ version 4.6 vehicle dynamics analysis package. Verification was against experimental vehicle data
30 taken at NHTSA’s Vehicle Research and Test Center (VRTC) in support of vehicle models developed for the National Advanced Driving Simulator [9].
The information in Table 2.1 gives some interesting insight into the effects of braking dynamics on vertical tire load for heavy trucks. The first two rows in Table 2.1 show an increase of 24% to 33% in vertical tire load under straight-ahead braking, regardless of whether the tractor/trailer is unloaded or loaded to GVW.
The situation becomes more complex as three-dimensional vehicle dynamics are considered. The bottom two rows of Table 2.1 contain results from a simulated best- effort stop while in a 500-ft radius corner and loaded to GVW. For the inside tire there is roughly a 3,600-lb (76%) maximum increase in vertical load while braking. For the outside tire, which experiences a higher dynamic load before braking, there is an increase of roughly 1,700 lbs (or 23%) during braking. As expected, the vertical loads on the inside and outside steer axle tires converge near the end of the stop, as lateral acceleration approaches zero.
From this summary, it is clear that the loads on the axle change dramatically during braking. This fact raises the consideration of how the model should treat the load effect revealed by the simulation, based on an assumption that the amount of load being arrested affects torque generation during one stop. The subject of brake fade is addressed in the following section.
31 Steer Axle Air Drum Brake Torques at 20 mph for Lining 'A,B,C, D'
9000
8000
7000
6000
5000
4000
Torque (ft-lbs) Torque 3000
2000
1000
0 0 20406080100 chamber pressure (psi)
Lining A 12k GAW R Lining B 13.2k GAW R Lining C 13.2 k GAW R Lining D 12k GAW R
S te er A xle Air D ru m B rake To rq ues at 50 m p h fo r L inin g 'A ,B ,C , D '
9000
8000
7000
6000
5000
4000
Torque (ft-lbs) 3000
2000
1000
0 020406080100120 chamber pressure (psi)
50 mph Lining A 12k GAW R 50 mph Lining B 13.2k GAW R 50 mph Lining C 13.2 k GAW R 50 mph Lining D 12k GAW R
S te er A xle Air D ru m B rake T o rq u es at 6 0 m p h fo r L in in g 'A ,B ,C , D ' 9000
8000
7000
6000
5000
4000
Torque (ft-lbs) 3000
2000
1000
0 0 20406080100120 chamber pressure (psi)
Lining A 12k GA W R Lining B 13.2k GAW R Lining C 13.2 k GAW R Lining D 12k GAW R
Figure 2.8 Steer axle drum brake torque measurements at 20 (top panel), 50 mph (middle panel), and 60 mph (bottom panel).
32 Drive Axle Air Drum Brake Torques at 20 m ph for Linings 'A,B,C,D,E' 18000
16000
14000
12000
10000
8000
Torque (ft-lbs) 6000
4000
2000
0 0 20406080100120 chamber pressure (psi)
Lining E 26k GAWR drive-1 Lining D 29k GA W R drive-2 Lining C 22k GA WR drive-3 Lining B 21k GAWR drive-3 Lining B 20k GAWR drive-4 Lining B 20k GA W R drive-4 Lining A 20k GA W R drive-3
Drive Axle Air Drum Brake Torques at 50 mph for Linings 'A,B,C,D,E'
18000
16000
14000
12000
10000
8000
Torque (ft-lbs) 6000
4000
2000
0 0 20406080100120 chamber pressure (psi)
Lining E 26k GAWR drive-1 Lining D 29k GA W R drive-2 Lining C 22k GA WR drive-3 Lining B 21k GAWR drive-3 Lining B 20k GAWR drive-4 Lining B 20k GA W R drive-4 Lining A 20k GA W R drive-3
Drive Axle Air Drum Brake Torques at 60 m ph for Linings 'A,B,C,D,E' 18000
16000
14000
12000
10000
8000
Torque (ft-lbs) 6000
4000
2000
0 0 20406080100120 cham ber pressure (psi)
Lining E 26k GAWR drive-1 Lining D 29k GA W R drive-2 Lining C 22k GA WR drive-3 Lining B 21k GAWR drive-3 Lining B 20k GAWR drive-4 Lining B 20k GA W R drive-4 Lining A 20k GA W R drive-3 Figure 2.9 Drive axle drum brake experimental measurements 20 (top panel), 50 mph (middle panel), and 60 mph (bottom panel).
33 Maximum Tire Vertical dynamic tire Change in Percent Vehicle load before load under load change in Load Road braking (lbs) braking (lbs) (lbs) load 0 load Straight 5,665 7,025 1,360 + 24 % GVW “ 6,182 8.206 2,024 + 33 %
GVW 500-ft 4,721 8,318 3,597 + 76% (inside tire) radius GVW (outside “ 7,194 8,880 1,686 + 23 % tire)
Table 2.1 Dynamic Steer Axle Loads from Best-Effort Braking Simulations
Although the results presented in Table 2.1 show that the load effect may be quite significant, it is unclear that adding more complexity to the brake model would contribute significantly to its accuracy. In straight-ahead braking, axle vertical load changes about 30%. Even during a braking-in-corner maneuver, where the wheel load can vary as much as 76%, this difference is the maximum difference during the stop, and it does not occur over a wide range of speeds. Note also, referring to Figure 2.10, that for a 15% change in test load for the ADB, the effect of load change on brake torque production appears minimal.
34 Air Disc Drum Brake Torques @ 20 mph for Lining 'A' 23,000 GAW R 14000 30 in2 air chamber
20,000 GAW R 12000 30 in2 air chamber
10000 12,000 GAW R 2 8000 20 in air chamber
6000 Torque (ft-lbs) Torque 4000
2000
0 0 20 40 60 80 100 120 chamber pressure (psi)
Lining A 23k GAW R Lining A 12k GAW R Lining A 20 k GAW R
Air Disc Brake Torques @ 50 mph for Lining 'A' 23,000 GAW R (3 tests) 14000 30 in2 air chamber
12000 20,000 GAW R (2 tests) 30 in2 air chamber 10000
8000 12,000 GAW R (2 tests) 20 in2 air chamber 6000 Torque (ft-lbs) 4000
2000
0 0 20 40 60 80 100 120 chamber pressure (psi)
Lining A 23k GAW R 50 mph Lining A 23k GAW R 50 mph Lining A 12k GAW R 50 mph Lining A 20 k GAW R
Air Disc Brake Torques by Speed for Lining 'A' 14000 23,000 GAW R 30 in2 air chamber 12000 20,000 GAW R 30 in2 air chamber 10000
8000 12,000 GAW R 20 in2 air chamber 6000 Torque (ft-lbs) Torque
4000
2000
0 0 20 40 60 80 100 120 chamber pressure (psi)
Lining A 23k GAW R Lining A 12k GAW R Lining A 20 k GAW R
Figure 2.10 Air disc brake torque measurements at 20 (top panel), 50 mph (middle panel), and 60 mph (bottom panel).
35 Since in this study there were no tests of drum brakes designed to show the effects of significant changes in dynamometer load (inertia) on Tb, we cannot determine precisely the effect of load on Tb. However, this topic should be explored in future research.
2.7 Addressing Brake Fade During a Stop
Brake fade that occurs during a single hard stop is also of concern to this study. In general, brake fade results from many phenomena including thermal expansion of the brake drum (away from the pad / mechanism), thermal or mechanical distortion of the mounting brackets and actuation mechanisms, and, probably most important, heat-related degradation of the friction between the brake pad (or shoe) and the (usually) ferrous surface that the brake pad or shoe works against [10].
For the purposes of analyzing vehicle stability during a stop, it was initially assumed that modeling the amount of brake fade (torque degradation) for both disc and drum brakes during a single stop would be of significant importance. To that end, real- time torque data during a stop would be most useful. In the absence of real-time data, the use of the FMVSS 121 standard “Brake Power” Test data was explored to quantify brake fade. As stated in FMVSS § 5.4.2, the test is performed by executing ten snubs that decelerate the inertial load from 50 to 15 mph (80.5 to 24.1 kph), at an average tangential (tire-wheel) rate of 9 ft/s2 (2.75 m/s2) on a 72-second duty cycle. Brake torque degradation (or fade) can be determined by the change (usually an increase) in brake chamber pressure required to maintain the prescribed 9-ft/s2 average deceleration rate for each of the ten consecutive snubs.
A second test, the “Brake Recovery” Test, utilizes twenty stops that decelerate the test inertia from 30 to 0 mph, at an average rate of 12 ft/s2 (3.66 m/s2) repeating on a 60- second duty cycle (FMVSS §5.4.3). A factor in deciding to use Brake Power Test, as opposed to the Brake Recovery Test to quantify brake fade was the amount of energy
36 consumed by the brake assembly during either test, compared to the total work done by an axle, during a typical hard stop executed by an actual vehicle.
The first two rows in Table 2.2 apply to simulated work and average horsepower per brake during a 0.37-G (12 ft/s2) stop from 60 mph, and a “snub” from 60 to 30 mph. The following row applies to a 0.37-G stop from 30 to 0 mph. The next two rows give the same data for one brake during the Brake Power and Brake Recovery Tests conducted at the same GAWR. The final three rows repeat the comparison at a simulated 23,000-lb GAWR. Here, we set our goal of using experimental data wherein the energy dissipated by the brakes, as computed in equation (2.7), is as close as possible to that dissipated by an actual vehicle brake at similar GAWR. Given the stated goal, it would appear that the work and average power for the Brake Power Test (0.28-G snub from 50 to 15 mph) are somewhat more applicable than those from the Brake Recovery Test for estimating the fade for a heavy truck brake during an actual vehicle stop.
1 ∆=WmVV()22 − (2.7) 2 f i
37
Work GAWR Speed during stop Average (lbs) Decel level range (k-lbs-ft) horsepower Comment 12 ft/s2 Typical vehicle 12,000 60-0 mph 721 179 (0.37 G) stop “ “ 60-30 mph 541 268 “ Dyno Recovery “ “ 30-0 mph 180 89 Test 9 ft/s2 Dyno Power “ 50-15 mph 455 145 (0.28 G) Test 12 ft/s2 Typical vehicle 23,000 60-0 mph 1263 313 (0.37 G) stop “ “ 60-30 mph 947 470 “ 9 ft/s2 Dyno Power “ 50-15 mph 798 254 (0.28 G) Test
Table 2.2 Total Work and Average Power Consumed per Brake During a Stop
Furthermore, two snubs from 50 to 15 mph (from the Brake Power Test conducted at 12,000-lb GAWR) result in a total work performed of 910 klb-ft (= 2 x 455) (1.234 kJ), which corresponds fairly closely to the 721 klb-ft (0.977 kJ) of work done by a typical steer axle brake during a 0.37-G vehicle stop (see the first row). Therefore, for this exercise, the first several repetitions of all Brake Power Test data available for each brake type were used to roughly estimate the amount of degradation of brake torque that one brake might experience during one stop from 50 mph to rest.
The method for determining the torque change (usually degradation for a given pressure) during a stop was straightforward. For the first three applications (snubs) of the Brake Power Test, the required increase in pressure (if any) was converted to a decrease in torque at the vicinity of chamber pressure necessary to perform the test. In other
38 words, a linear interpolation was used to estimate the amount of torque change near the operating point experienced by the brake during those first applications of the Brake Power Test. From the estimated torque loss during one Brake Power Test snub, a linear brake degradation relationship was formulated in the terms given in equation (2.8). The estimates of brake fade expected during a hard stop are listed in Table 2.3.
∆T R = b (2.8) fade ∆W
where
Rfade = brake fade ratio (% / 100 ft-lb)
The results from Table 2.3 reveal that steer axle drum brakes experience the most fade during a stop, the drive and trailer axle drum brakes (of similar design and identical dimension) show similar rates of fade, and the disc brakes (on either axle) show negligible fade. However, no brake demonstrated enough torque fade during a series of FMVSS 121 compliant stops or snubs to justify additional model complication to account for fade during one or two stops from highway speed.
Figure 2.11 shows time-based torque data from a dynamometer test, resulting from an application at constant brake chamber pressure of 80 psi (5.515 bar). The ordinate of Figure 2.11 gives the torque output in terms of the peak measured for that brake application. The data shown in Figure 2.11 confirm the results from the previous, more involved method, in which it is shown that is that there is little appreciable brake fade during a single stop. The 10% deviation in measured brake torque levels (after peak) in Figure 2.11 compare favorably to the predicted fade of around 6% in Table 2.3.
39 50 MPH Brake Power Test at 80 psi 16.5x7 (drive axle S-cam drum brake) 15x4 (steer axle S-cam drum brake)
100% 100
90% 95 80% 90 70% brake torques 85 60%
50% 80
40% chamber pressures 75 30%
16.5x7 Torque 70 Chamber Pressure [psi]
% of Maximum Brake Torque 20% 15x4 Torque 16.5x7 Chamber Pressure 65 10% 15x4 Chamber Pressure 0% 60 0.02 0.42 0.82 1.22 1.62 2.02 2.42 2.82 3.22 3.62 4.02 4.42 Time [Sec]
Figure 2.11: Brake dynamometer experimental output showing step input response to the brake pressure command for a steer axle s-cam drum brake and a drive axle s-cam drum brake. Traces are brake chamber pressure (right axis) and measured torque (left axis).
2.8 Brake Torque Model in Operation
Figures 2-12 and 2-13 show dynamic output of the brake torque model, as coupled to the pneumatic model discussed at length in reference [2].
Figure 2.12 shows the response of a dynamic model of the brake dynamometer to a step input in the command signal at t=0 seconds. The top panel shows both simulated and actual dynamometer brake chamber time responses, along with the theoretical step command. The bottom panel shows torque output from the simulation containing the models developed herein, along with experimental dynamometer torque output. Note
40 that the deviations in torque after stabilization of chamber pressure (and torque) were not modeled in the simulation.
Rfade Percentage Work done Brake type Decrease in Fade Ratio Comments during snub Tb (% per 100 ft- lbs of work) Steer axle 6.2 % 455 k-ft-lbs 1.363 drum Drive axle 6.5 % 797.9 k-ft-lbs 0.814 drum Trailer Axle 6.0 % “ 0.752 Drum * = negligible Steer axle disc 0 % * 455 k-ft-lbs 0 increase in Pc Drive axle * = negligible 0 % * 797.9 k-ft-lbs 0 disc increase in Pc
Table 2.3 Experimental Brake Fade, in Terms of Torque Decrease, for Various Brake Types
41 DRIVE Axle Brake Simulation: Treadle & Chamber Pressure
100
step command 80
simulated dyno speed (ft/s) 60
40 simulated treadle (or command) simulated brake chamber experimental DYNO brake chamber dyno speed (ft/s) pressure (psi) and speed (ft/s) 20
0 -1 0 1 2 3 4 5
DRIVE Axle Brake Simulation with Experimental Dynamometer Torque Output 12000
10000
8000
brake system model response to step up and step down 6000 experimental DYNO Torque Response to step input
4000
Brake Torque Magnitude (ft-lb) Magnitude Torque Brake 2000
0 -1 0 1 2 3 4 5 time (s)
Figure 2.12 Brake dynamometer simulation and experimental data: response to a step input in command pressure. Top panel: pressures, simulated command (treadle), simulated chamber, experimental chamber, with simulated speed output. Bottom panel: simulated torque output, experimental dynamometer torque output.
42 Figure 2.13 shows the response to several cycles of a simulated step-up-then- release application using the same dynamometer model. As the brake pressure is momentarily released and falls below Ppo, the torque versus speed relationship is re- applied as per equation (2.6). As the hub speed decreases, the increasing torque magnitude can be seen rising in the bottom panel of Figure 2.13, as a consequence of the change in Tb versus Pc coefficients with respect to application speed.
DRIVE Axle Brake Simulation: Treadle & Chamber Pressure
step up/down command simulated treadle (or command) 100 simulated brake chamber dyno speed (ft/s)
80
60 simulated dyno speed (ft/s)
40
pressure (psi) and speed (ft/s) 20
0 -1 0 1 2 3 4 5
DRIVE Axle Brake Simulation with Experimental Dynamometer Torque Output 12000
10000
8000
6000
4000
Brake Torque Magnitude (ft-lb) 2000
0 -1 0 1 2 3 4 5 tim e (s) Figure 2.13 Brake dynamometer simulation. Top panel: pressures, simulated command (treadle), simulated chamber, simulated dynamometer speed (ft/s). Bottom panel: simulated brake torque output.
43 Tables 2.4 through 2.7 contain the model coefficients described in equations (2.4) through (2.6).
Cij i = 0 i = 1 i = 2 j = 1 0.000331 -0.0259 0.4262 j = 2 -0.0433 2.4617 53.66 j = 3 0.6582 -31.738 -201.3
Table 2.4 Model Fit Coefficients for the Steer Axle Brake Torque Model
Cij i = 0 i = 1 i = 2 j = 1 -0.000620 0.0455 -0.7757 j = 2 0.0311 -2.9823 207.33 j = 3 -0.3425 38.555 -1286.1
Table 2.5 Model Fit Coefficients for the Drive Axle Brake Torque Model
Cij i = 0 i = 1 i = 2 j = 1 -0.00023 0.0191 -0.7024 j = 2 0.0186 -2.431 204.02 j = 3 0.1065 7.205 -1087.7
Table 2.6 Model Fit Coefficients for the Trailer Axle Brake Torque Model
Cij i = 0 i = 1 i = 2 j = 1 -0.00099 0.0814 -1.7314 j = 2 0.0986 -8.4106 324.4 j = 3 -0.6748 69.392 -2533.7
Table 2.7 Model Fit Coefficients for the ADB Brake Torque Model
44 2.9 Conclusions
In this chapter, the development of a new empirical brake torque model was presented. The model is intended for simulating heavy truck brake torque as a function of brake chamber pressure and application (hub) speed, during a single hard stop. The model can easily be applied to brake simulations or integrated with heavy vehicle dynamic simulations. Having characteristics of models developed for similar purposes in the past, this model uses empirical equations that describe the torque versus pressure relationship; the coefficients of those equations adjust with respect to application speed. For many modeling applications, the empirical formulas may be easier to apply than other methods, such as lookup tables.
The empirical brake torque model developed was derived using contemporary dynamometer data from representative examples of s-cam drum brakes currently available for steer, drive, and trailer applications. Modern production air disc brakes are also characterized and show robust behavior with respect to speed, versus drum brakes of similar capacity.
Also, this chapter addressed the option of correcting the brake torque output as a function of the vertical load (or work history) of the brake during a single stop, concluding that further investigation was needed.
Finally, it was shown that models such as this, that are designed and developed to model heavy vehicle brakes during single stop simulations (as opposed to long braking- on-hill simulations), do not need to be adjusted for brake fade.
45 2.10 Chapter 2 References
1. T.M. Post, P.S. Fancher, and J.E. Bernard, “Torque Characteristics of Commercial Vehicle Brakes,” SAE 750120. 2. A.L. Dunn, G.J. Heydinger, G. Rizzoni, and D.A. Guenther, “New Model for Simulating the Dynamics of Pneumatic Heavy Truck Brakes with Integrated Anti- Lock Control,” SAE 2003-01-1325. 3. D. Yanakiev, J. Eyre, and I. Kanellakopoulos, “Longitudinal Control of Heavy Duty Vehicles: Experiment Evaluation,” California PATH Research Report, UCB-ITS- PRR-98-15. 4. M.A. Flick, “An Overview of Heavy Vehicle Brake System Test Methods,” SAE 96225. 5. C. Hatipoğlu, T. Acarman, and Ü. Özgüner, “Pneumatic Pressure Control: Blending Simulations to Implementation,” Proceedings of ESDA 2002 Conference, 2002. 6. TruckSim version 5.0 User Manual (Ann Arbor, MI: Mechanical Simulation Corporation, 2003). 7. E. Milich, “An Evaluation of VDM Road and VDANL Vehicle Dynamics Software for Modeling Tractor-Trailer Dynamics” (master’s thesis, The Ohio State University, 1999). 8. A.L. Dunn, “The Effects of Cornering Force Variation on Articulated Vehicle Predictions,” NHTSA / VRTC internal report, September 2000. 9. W.R. Garrott, P.A. Grygier, J.P. Chrstos, G.J. Heydinger, M.K. Salaani, J.G. Howe, and D.A. Guenther, “Methodology for Validating the National Advanced Driving Simulator's Vehicle Dynamics (NADSdyna),” NHTSA report, 2001. 10. “Mechanics of Heavy Truck Systems,” short course material, University of Michigan Transportation Research Institute (UMTRI), July 2000.
46 CHAPTER 3
DEVELOPMENT OF AN ANALYTICAL MODEL FOR SIMULATING THE DYNAMICS OF PNEUMATIC HEAVY TRUCK BRAKES WITH INTEGRATED ANTI-LOCK CONTROL
3.1 Abstract
This chapter introduces a new nonlinear model for simulating the dynamics of pneumatic-over-mechanical commercial vehicle braking systems. The model employs an effective systems approach to accurately reproduce forcing functions experienced at the hubs of heavy commercial vehicles under braking. The model, which includes an on-off type ABS controller, was developed to accurately simulate the steer, drive, and trailer axle drum (or disc) brakes on modern heavy commercial vehicles. This model includes parameters for the pneumatic brake control and operating systems, a 4s/4m (four sensor, four modulator) ABS controller for the tractor, and a 2s/2m ABS controller for the trailer. The dynamics of the pneumatic control (treadle system) are also modeled. Finally, simulation results are compared to experimental data for a variety of conditions.
47 3.2 Motivation
The goal for this research and development was to create a model via a systems approach that can accurately simulate the brake system dynamics during ABS-assisted braking. The component level model allows detailed study of the influence from individual component parameters on system performance.
Although vehicle modeling experts have developed algorithms intended to simulate braking systems (with varying degrees of detail), many of the published models lack certain elements, such as brake torque hysteresis, which have a significant influence on the character and magnitude of forcing functions experienced by the vehicle chassis due to braking forces. And, although previous models have provided various depths of complexity, the description of their components in technical forums is often vague due to the necessary protection of a manufacturer’s competitive edge. In addition, perennially increasing computer power allows researchers the ability to add more sophistication to new dynamic models used to simulate the forcing functions at vehicle hubs (which ultimately control the vehicle dynamics). Therefore, this “new generation” brake model contributes significantly to the levels of detail and accuracy available in pneumatically actuated commercial vehicle brake models accessible in the public domain.
The brake models discussed herein were developed to be used as part of larger nonlinear multi-body vehicle dynamics simulations. Specifically, the brake system models are used as system components; these models run in the Simulink® environment, in parallel with TruckSim™ version 5.0 heavy truck dynamics simulation software. When the models are running in parallel, the TruckSim™ software effectively models the dynamic behavior of non-brake related hardware, such as suspension deflection, aerodynamics, hitch deflection parameters, and the very complex force and moment dynamics from the tires.
These complex vehicle dynamics and brake system simulations are combined to study the effects on stability of using ECBS (electronically controlled braking system) actuated, high-torque brakes, such as air disc brakes (ADB), on the prime mover (tractor) 48 while keeping traditional pneumatically controlled drum brake technology on the trailer. The results of the analytical studies on braking stability are discussed in detail in Chapter 4.
3.3 Background
A history of various published brake models for commercial vehicles is as follows. • During the 1970’s, Fancher, Post, et. al., from the University of Michigan’s UMTRI, published several technical papers that reported and discussed brake torque measurements [1] and brake models [2].
• During the 1980’s, the National Highway Traffic Safety Administration’s Vehicle Research and Test Center, led by Richard Radlinski, published many papers discussing measured braking performance [3] and brake system compatibility [4] for heavy trucks.
• A significant portion of the current literature on heavy vehicle brakes is summarized in the University of Michigan “Mechanics of Heavy Duty Truck Systems” short course [5].
• A model similar to the one presented herein was published in 1998, by Yanakiev et al., at UCLA [6].
• ABS models exist, but many are understandably simple or vague in the interest of protecting complex proprietary control algorithms.
• The default ABS model in TruckSim™ v. 5.0 models the ABS system by setting “on” and “off” thresholds for the vehicle brakes [7].
• Recent contributions have been made by Hatipoğlu, et. al., with regard to modeling pneumatic brake systems and ABS systems [8, 9, 10].
49 3.4 Model Overview in General
This new generation model includes the following significant features to simulate pneumatic brake system dynamics.
• First-order differential equations model the pneumatic system dynamics for the control (treadle) circuit and main brake actuation circuits
• Time delays for control (treadle) signals based on physical location of the associated modulator valve
• 4s/4m (four sensor, four modulator) integrated ABS control system for the tractor
• 2s/2m integrated ABS control for the semitrailer
• Simulated ABS controller calculation lag
• ABS control strategy based on longitudinal wheel slip level and tangential acceleration, tuned to match actual vehicle performance on wet and dry surfaces
• ABS system integration and control that commands each of six modulator valves in the 10-brake-position system to build (apply), hold, or dump (release) brake pressure
• Quadratic model of brake torque output as a function of application speed and chamber pressure (also discussed in [11]).
• Brake system hysteresis, as seen in the modern s-cam drum type brakes, common on Class-8 heavy commercial vehicles.
• The option of simulated ECBS (electronically controlled braking system) control
• The option of simulated torque output for disc brakes with various sizes of pneumatic brake chambers based on experimental brake dynamometer results
50
3.5 Dynamics of the Pneumatic Brake System
The forces and moments transmitted by an actual brake system, whether pneumatic or hydraulic, are capable of relatively high-frequency load inputs. The frequency and magnitude of these dynamic braking forces become all the more important when abrupt brake application and release, inherent with ABS control, is applied to the system.
3.5.1 Modeling the Dynamics of the Pneumatic Brake Chambers
Some inspiration for the mechanical systems models contained herein came from a paper by the UCLA School of Engineering, which addressed the use of tractor semitrailer brakes to assist speed and following distance control of trucks traveling in convoy [6]. In their work, Yanakiev et. al., modeled the first-order system response of the pneumatic pressure, converting said pressure into force along the chamber pushrod, then coupling that force with a linear transfer function which considered slack adjuster length, S-cam effective radius, brake drum diameter, and brake lining material coefficient of friction to produce a brake torque.
It is obvious from their and other vehicle system models, including TruckSim™ version 5.0, that the dynamic effects due to the pneumatics are highly important in accurately modeling heavy truck brake behaviors.
This model therefore uses the traditional first-order ordinary differential equation to model the dynamics of the pneumatic pressure within each brake pressure chamber, which in turn drives the brake torque output for each brake. As seen in equation (3.1), the rate of change of the chamber pressure is directly proportional to the difference between that chamber pressure and the command (treadle) pressure.
51 1 PPP =−() ctcτ (3.1)
and
ττ=≤> filling when PP c po and P c 0 ττ=>> rising when PP c po and P c 0 ττ=<