PREDICTION OF CREEP AND SHRINKAGE BEHAVIOR FOR DESIGN FROM SHORT TERM TESTS
B. L. Meyers University of Iowa Iowa City, Iowa D. E. Branson University of Iowa Iowa City, Iowa C. G. Schumann Chicago Bridge and Iron Co. Plainfield, Illinois
Presents simple empirical equations for predicting long-time creep and shrinkage properties of concrete. Prediction accuracy that previously required almost 4 months of testing can now be achieved with only 28 days of creep and shrinkage data. General constants are presented for use when the 28-day experimental program is not feasible.
The importance of having adequate mated from general relationships. time-dependent concrete properties and The prediction equations developed accurate prediction methods was de- in References 2, 3 and 4, and included monstrated by Branson and Kripanaray- in the analysis of prestressed concrete anan1 1) who showed that loss of pre- structures reported in Reference 1, form stress and camber in non-composite and the basis of this work. It will be shown composite structures could be predicted that using the methods described in this twice as accurately when experimen- paper, prediction accuracy that previ- tally determined material parameters ously required almost 4 months of test- were used, as compared to predictions ing can be obtained with only 28 days made using material properties esti- of creep and shrinkage test data.
PCI Journal/May-June 1972 29 800
760
600 Branson / -O 500 ^^" /t Ross 400
300
200
100
80 160 240 320 400 480 560 t days
Fig. 1. Predicted creep comparing Ross equation with Branson Eq. (1)
Previous proposals data collection. Such equations have been proposed by Thomas(a ), Mc- for creep prediction Henry(6), Saliger(7), Shank(8), and Creep prediction methods that might Troxell, et al(9). be useful to the design engineer can be A number of simpler hyperbolic divided into two general categories: equations, which do approach a finite (1) the creep-time relationship is ex- limit, have also been suggested. Those pressed in the form of an equation, and used most often are the equations of usually requires that one or more em- Ross (10) pirical constants be determined experi- c — a mentally, and (2) creep is expressed by + bt °- a standard curve which can be modi- and Lorman(") fied by a number of factors to allow for mt various mix and storage conditions. The c— n+t o- latter prediction method does not re- quire experimental data but is usually where c = creep, t = time, o = stress, less accurate than using an empirical and a, b, m and n are experimentally equation based on actual measure- determined empirical constants. ments. Methods using standard creep curves In Category 1, about a dozen expo- can be represented by those suggested nential or hyperbolic equations have by Jones, et al( 12 ), and Wagner(13). been suggested. Most exponential equa- Jones used a standard curve, valid for tions, which have the practical disad- specific mix and storage parameters, vantage of not approaching a finite which can be corrected for other condi- limit, are - of doubtful value to the de- tions by using a set of correction fac- sign engineer because they are unwield- tors. Wagner s method differs only in ly and/or require extended periods of that standard values of ultimate specific
3Q 30
25
20
a 15 L) 0 O P 10
5
0 0 10 15 20 25 3C
Actual Time Under Load - Weeks Fig. 2. Accuracy of predicting 1-year creep from short-time tests
creep are supplied in lieu of the stan- dard creep curve. (E85)t = f t' t^ (Esh)u (2) The literature is rather sparse in the where area of shrinkage prediction, although a number of complex methods have been Ct = creep coefficient, defined as suggested( 14,15 ). However, since the ratio of creep strain to ini- methods developed in this paper have tial strain, at any time t their basis in the creep prediction meth- (e3n,)t = shrinkage strain at any time ods already discussed, further analysis t of available shrinkage methods will be Cu = ultimate shrinkage coeffi- omitted. cient In an attempt to increase the accur- (esh),, = ultimate shrinkage strain acy of creep and shrinkage prediction, c, d, e, f = empirical constants Meyers, et al( 6 , combined the meth- It is interesting to note that in addition ods of Ross and Jones. Although some to developing the well known creep improvements were made, it was noted equation, Ross suggested a shrinkage that this method, as well as others, pre- equation similar to Eq. (2) in 1937(10). dicted ultimate creep fairly well, but Comparisons with measured data did not adequately represent the defor- show that the form of the creep predic- mation behavior of the material early in tion in Eq. (1) is more representative of its life. the full range of creep behavior than This difficulty was overcome by the form originally suggested by Ross. Branson, et al 2,3'4), who proposed the Such a comparison is made in Fig. 1. following standard prediction equations Because the Branson equation is more to representative of the full range of creep Ct u (1) d +to C behavior, it can be used in an accurate
PCI Journal/May-June 1972 31 Moist cured and steam cured concretes 100 • O, • Eq. ( 3 ) rO q ■_..O 4) 80
U • o 60 a ♦ •
40 • Nor. Wt. Sand_Lt.Wt.l Al1_Zt.t. I,Moist o ( 12,1) (20,2)-O-(19,3) (2,3) ••(12,21) (2, 1) (17, 4) U 20 III,Moist C (18,7) (20,2 •(17,4) Tp.I,Steam a (18,2) ♦(17,6) T .III Steam 7(20,3) (21 V 17. 7.) (2 0 160 320 480 640 800 Time after loading in days Fig. 3. Creep coefficient as a percent of ultimate vs. time, comparing Eq. (3) with test data. Loading ages are 7 days for moist cured and 2 to 3 days for steam cured concretes. (In each set of parentheses, the numbers refer to the source of the data and the no. of specimens, respectively.)
prediction method based on only 28- It can be seen from Fig. 2 that for most day data. It can also be shown that Eq. available methods, in order to predict (2) accurately represents the full range creep to within an error coefficient of of shrinkage vs. time behavior(2.3.4). 10 percent, 20 weeks of data is re- It is significant that accurate predic- quired. tion can be obtained with 28-day data in light of information presented by Ne- Prediction equations ville and Meyers( 17). The accuracy of any method can be evaluated in terms after Branson, of an error coefficient M(17). et alr=.2"" To solve Eq. (1) and (2) for C,, and M= VCi— Cti)2/n (€,h)t, three unknowns must be evalu- CCI ated in each case (d, c and C. are un- where Ct = creep after one year pre- known for Eq. (1), and for Eq., (2), e, f dicted from measured and (E8h)„ are unknown). Of the three creep after t weeks under unknowns, two from each equation are load empirical constants while C,, and Cd = actual creep after i years (e8n,),, are material properties. under load In order to determine the empirical n = number of specimens or constants d and c, creep data from Ref- experimental sets for erences 12, 16 and 18-21 were normal- which creep was observed ized with respect to C,, and plotted in at time t Fig. 3. In most cases, three data points
32 a. Moist cured concrete 100ҟ •
0
0 (4a) 60
• 4 N 40 w Nor. Wt. Sand-Lt.Wt All-Lt.Wt. I,Moist 0(12,1) (20, 1) x(123) (2,3) •(12,21) (19, 1) 4J 20 (20,2) •III,Moist D(20, 1) (18, 1) • (20,2) WN (22,3) 0 0ҟ160ҟ320 480ҟ640ҟ800 Time after initial shrinkage considered in days
y100 b. Steam cured concrete r. N •.p 1 + 80 0 o • F. (4b) a D d 60
X40 W Nor.Wt. All-Lt.Wt. 20 Type I,Steam 0 (20, 1) • (20, 2) N Type III,Steam o (20, 1) (21, 8) ■ (20, 2) (21,42) W 0 0ҟ16oҟ320ҟ480ҟ640ҟ800 Time after initial shrinkage considered in days
Fig. 4. Shrinkage as a percent of ultimate vs. time. Curve a, based on Eq. (4a), is for moist cured concrete, initial shrinkage considered for 7 days; Curve b, based on Eq. (4b), is for steam cured concrete, initial shrinkage considered for 2 to 3 days. (For plotted data, the numbers in parentheses refer to data source and no. of specimens, respectively. Three data points for a specific time refer to upper and lower limits and an average value. Only one data point indicates too narrow a range to show.)
PCI Journal/May-June 1972 33
Table 1. 28-day extrapolation of creep
Specimen Ci28 C.. 0365 r3 5 C730 C730 V365 y730 designation Experimental Predicted Experimental Predicted Experimental Predicted C;65 C,3 4 0.97 2.28 1.82 1.77 1.86 1.91 0.973 1.027 6 1.15 2.70 2.06 2.09 2.17 2.26 1.015 1.041 8 0.92 2.16 2.03 1.67 2.14 1.81 0.823 0.845 12 0.82 1.93 1.66 1.50 1.71 1.62 0.904 0.947 16 0.82 1.93 1.55 1.50 1.69 1.62 0.968 0.959 20 0.64 1.50 1.30 1.16 1.44 1.26 0.892 0.875 24 0.73 1.72 1.37 1.33 1.52 1.44 0.971 0.947 71 1.37 3.22 2.46 2.50 2.73 2.70 1.016 0.989 72 1.25 2.94 2.36 2.28 2.61 2.46 0.966 0.943 73 1.20 2.82 2.75 2.18 2.31 2.36 0.793 0.793 V 74 1.28 3.01 2.46 2.33 2.62 2.52 0.947 0.962 6N6 1.90 4.47 3.45 3.46 3.72 3.75 1.003 1.008 6N28 1.52 3.58 3.01 2.78 3.32 2.99 0.924 0.901 6S2 1.10 2.59 2.21 2.01 2.53. 2.17 0.910 0.818 6S7 1.04 2.45 2.20 1.90 2.52 2.06 0.864 0.817 6S28 0.95 2.24 2.20 1.74 2.51 1.88 0.791 0.749 10N6 1.04 2.45 1.79 1.90 1.94 2.06 1.061 1.062 10N28 0.75 1.76 1.59 1.36 1.74 1.48 0.855 0.851 10S2 0.65 1.53 1.30 1.18 1.45 1.28 0.908 0.883 1057 0.72 1.69 1.34 1.31 1.50 1.42 0.978 0.947 10528 0.66 1.55 1.43 1.20 1.62 1.30 0.839 0.802 8N6 1.73 4.07 3.02 3.16 3.19 3.42 1.046 1.072 8N28 1.88 4.43 3.40 3.36 3.70 3.72 0.988 1.005 8S7 1.41 3.32 2.45 2.58 2.74 2.78 1.053 1.015 8S28 1.35 3.18 2.59 2.48 2.95 2.67 0.958 0.905 6M5 1.51 3.56 2.78 2.76 3.01 2.98 0.993 0.990 6M28 1.10 2.59 2.48 2.00 2.67 2.16 0.806 0.809 6R7 0.74 1.74 1.70 1.35 1.93 1.46 0.853 0.756 6R28 0.60 1.41 1.54 1.09 1.78 1.18 0.708 0.663 1OM5 0.93 2.18 1.84 1.69 1.97 1.83 0.918 0.929 10M28 0.92 2.16 1.93 1.67 2.12 1.81 0.865 0.854 1082 0.68 1.60 1.34 1.24 1.49 1.34 0.925 0.899 10R7 0.66 1.55 1.33 1.20 1.46 1.30 0.902 0.890 10828 0.63 1.48 1.40 1.15 1.56 1.24 0.821 0.795 8M5 1.57 3.70 2.96 2.87 3.19 3.10 0.970 0.972 8M28 1.73 4.07 3.00 3.15 3.23 3.42 1.05 1.059 8R2 1.09 2.56 2.10 1.98 2.34 2.14 0.943 0.914 8R7 1.13 2.66 2.32 2.06 2.55 2.23 0.888 0.875 8R28 1.08 2.54 2.34 1.97 2.64 2.13 0.842 0.807 6R2 0.90 2.12 1.80 1.64 2.00 1.78 0.911 0.890 0.6 XCu C28 _. C^a C 365 36s = 10 13650.6 Cu = 280.6 +280.6 0.025 10
are shown for a particular specimen t (moist cured) (4a) category and time. They represent the (e8n)t = 35 + t upper and lower limits and average values of these data. Only one data (steam cured) (4b) (E'h)t 55- t (Esn)u point is shown for a specific time when the spread between upper and lower Using the basic Eqs. (3), (4a) and values is small. Eq. (3) was derived by (4b), general prediction equations can fitting a curve to the average values of be supplied to the designer by specify- the data. ing average values C. and (€8h). This to.s was done in Reference 1 where it was shown that loss of prestress and camber Ct = 10 + t0.6 Cc (3) could be predicted to within ± 30 per- Eq. (3) can be used for both moist and cent of actual results using average steam cured concrete. values of C,u and (e8h)u. Keeton(28) Similarly Eqs. (4a) and (4b) were and Pauw( 23 ) have also used Eqs. (3), developed from shrinkage data plotted (4a) and (4b) to predict structural re- in Fig. 4. sponse with an adequate degree of ac-
34 Table 2. Determination of error coefficient
Predicted, Experimental, Predicted, Experimental, 365 days 365 days (Cr - C1) (C1 - G)2 730 days 730 days (C1- C,) (Ct - C1)2 Ct C; 365 365 Ct C, 730 730 1.77 1.82 0.05 0.0025 1.91 1.86 0.05 0.0025 2.09 2.06 0.03 0.0009 2.26 2.17 0.09 0.0081 1.67 2.03 0.36 0.1296 1.81 2.14 0.33 0.1089 1.50 1.66 0.16 0.0256 1.62 1.71 0.09 0.0081 1.50 1.55 0.05 0.0025 1.62 1.69 0.07 0.0049 1.16 1.30 0.14 0.0196 1.26 1.44 0.18 0.0324 1.33 1.37 0.04 0.0016 1.44 1.52 0.08 0.0064 2.50 2.46 0.04 0.0016 2.70 2.73 0.03 0.0009 2.28 2.36 0.08 0.0064 2.46 2.61 0.15 0.0225 2.18 2.73 0.57 0.3249 2.36 2.31 0.05 0.0025 2.33 2.46 0.13 0.0169 2.52 2.62 0.10 0.0100 3.46 3.45 0.01 0.0001 3.75 3.72 0.03 0.0090 2.78 3.01 0.23 0.0529 2.99 3.32 0.33 0.1089 2.01 2.21 0.20 0.0400 2.17 2.53 0.36 0.1296 1.90 2.20 0.30 0.0900 2.06 2.52 0.46 0.2116 1.74 2.20 0.46 0.2116 1.88 2.51 0.63 0.3969 1.90 1.79 0.11 0.0121 2.06 1.94 0.12 0.0144 1.36 1.59 0.23 0.0529 1.48 1.74 0.26 0.0676 1.18 1.30 0.12 0.0144 1.28 1.45 0.17 0.0289 1.31 1.34 0.03 0.0009 1.42 1.50 0.08 0.0064 1.20 1.43 0.23 0.0529 1.30 1.62 0.32 0.1024 3.16 3.02 0.14 0.0196 3.42 3.19 0.23 0.0529 3.36 3.40 0.04 0.0016 3.72 3.70 0.02 0.0004 2.58 2.45 0.13 0.0169 2.78 2.74 0.04 0.0016 2.48 2.59 0.11 0.0121 2.67 2.95 0.28 0.0784 2.76 2.78 0.02 0.0004 2.98 3.01 0.03 0.0009 2.00 2.48 0.48 0.2304 2.16 2.67 0.51 0.2601 1.35 1.70 0.35 0.1225 1.46 1.93 0.47 0.2209 1.09 1.54 0.45 0.2025 1.18 1.78 0.60 0.3600 1.69 1.84 0.15 0.0225 1.83 1.97 0.14 0.0196 1.67 1.93 0.26 0.0676 1.81 2.12 0.31 0.0961 1.24 1.34 0.10 0.0100 1.34 1.49 0.15 0.0225 1.20 1.33 0.13 0.0169 1.30 1.46 0.16 0.0256 1.15 1.40 0.25 0.0625 1.24 1.56 0.32 0.1024 2.87 2.96 0.09 0.0081 3.10 3.19 0.09 0.0081 3.15 3.00 0.15 0.0225 3.42 3.23 0.19 0.0361 1.98 2.10 0.12 0.0144 2.14 2.34 0.20 0.0400 2.06 2.32 0.26 0.0676 2.23 2.55 0.32 0.1024 1.47 2.34 0.37 0.1369 2.13 2.64 0.51 0.2601 1.64 1.80 0.16 0.0256 1.78 2.00 0.22 0.0484 84.66 2.1195 91.17 3.0894
(C1 - C )2 (C1 - C1)2 - 3.0894 = 0.0772 1 -_ 2.1195 = 0.05299 n 40 nҟ40
(C` (C` - C;)2 =V5.299 x 10- 2 = 0.23 = V7.72 X 102=0.278 n n C;/n = 84.66/40 = 2.12 C;/n = 91.17/40 = 2.28 M M 0.23 x 100 =10.85% (365 day analysis) = 0.278 >_!- = 12.20% (730 day analysis) 2.12 2.28 curacy. It is not difficult to see that su- sumed to accurately represent the perior results could be obtained if creep-time relationship, it can be seen methods were available to more accur- that only one point on an experimental ately predict the material parameters creep-time curve is required to solve C. and (e$h)^,.. the equation for Cu; i.e., if Ct at any time is known then Eq. (3) becomes Creep prediction from to.6 Cu = Cc - u.sl (5) 28-day data [10 + t If the general form of Eq. (3) is as- and C. can be evaluated, thereby giv-
PCI Journal/May-June 1972ҟ 35 Table 3. Details of concrete mixes and mixing procedure
Ingredients Idealite Haydite Haydite Haydite for 1 cu. yd. by Hydraulic Press Brick by Buildex by Carter-Waters Cement (Type I) 705 lb. 705 lb. 611 lb. 658 lb. Coarse 820 lb. 20 ft.3 = 825 lb. 22.5 ft.3 = 977 lb. 23.5 ft. ҟ= 1318 lb. aggregate 60%-3/4 to 5/16 in. 40%-5/16 in. to No. 8 3/4 in. to No. 4 3/4 in. to No. 4 3/16 to 1/8 in. Sand 1395 lb. 1150 lb. 1020 1 b. 816 lb. Water 292 lb. 350 lb. 350 lb. 415 lb. Admixtures Darex-7/8 oz./sack — — WRDA-50 oz.
Mixing procedure: 1. Proportion and batch sand and aggregate. 2. Add approximately one-half of required water. 3. Mix for approximately two minutes. 4. Proportion and batch cement. 5. Add admixtures along with remaining water. 6. Mix for approximately three minutes or until homogeneous mixture is obtained. ing a continuous equation for creep as are calculated in Table 2. It can be a function of time. seen, from Fig. 2, that to obtain an The accuracy of the method can be error coefficient of 10 percent Neville evaluated from Table 1. Shown are l- and Meyers( 7 ) indicate that the tests and 2-year creep coefficients predicted should be carried out for about 20 from ` measured 28-day creep coeffi- weeks. Using the prediction method de- cients and experimental 1- and 2-year veloped herein similar accuracy can be creep coefficients (experimental data obtained with only 28 days (4 weeks) from References 12, 16 and 18-21). The of data collection; if greater than 28- ultimate creep coefficient Cu was esti- day results are obtained, increased ac- mated by substituting Ct at 28 days curacy can be expected. into Eq. (5). The data show that 53 per It is obvious from the above develop- cent of the calculated values are within ment that C„ can be estimated if C t at 10 percent, and 83 percent of the cal- any time is known. The creep coeffi- culated values are within 20 percent of cient Ct at 28 days is recommended the one-year observed values. Similar here for two reasons: figures for 2-year data are 50 percent 1. Strength and elastic properties are of the calculated values within 10 per evaluated based on 28-day tests; cent, and 80 percent within 20 per it was deemed desirable to main- cent of the observed values. In both tain this standard time interval. cases over 90 percent of the calculated 2. The accuracy obtained using less values are within 30 percent of the ob- than 28-day data was considered served values. It should be noted that a unsatisfactory. 30 percent variation in material prop- erties represents a significantly lower variation in comparing calculated struc- Experimental tural deformations and actual structural verification of 28-day deformations. creep prediction An additional measure of the accur- acy of the method is indicated by the method error coefficient M. The average error The 28-day creep prediction method coefficients for I-year and 2-year pre- and the general form of the creep-time diction for the 40 sets of data analyzed relationship suggested in Eq. (3) were
36 Table 4. Concrete properties
Property Idealite Haydite by H.P.B. by Bldx. byC-W -1 1-3 IS H-1 B-4 CW-4 f-7 daysҟ psi 6,700 6,150 5,600 5,150 3,650 3,450 f -14 daysҟ Psi 8,250 - 5,800 5,900 4,500 4,750 f-28 daysҟ psi 9,350 8,750 6,100 -- - Unit weight (wet)ҟpcf 124 125 - 113 105 115 Unit weight (dry)ҟpcf 123 124 122 113 103 113 Measured entrained airҟ% 4 6 --- - Slump,ҟ in. 2 2 h - 2^ 2 1 h E-7 days, psi X 10-6 secant @ 0.5 f - 3.20 3.04 2.93 2.45 2.66 initial tangent - 3.33 3.10 3.05 2.84 2.84 33 w 1 3.68 3.55 3.32 3.84 2.21 2.44 E6-14 days, psi x 10 -6 secant @ 0.5 - - - 3.06 2.51 2.88 initial tangent - - - 3.28 2.84 3.10 33 w3 fl 4.08 - 3.38 3.00 2.51 2.70 Ee 28 days, psi x 10 -6 secant @ 0.5 f - 3.28 - - - - initial tangent - 3.38 - - - - 33 w3 f1 4.35 4.23 3.47 - - - Relative humidity,ҟ(range) 20-50 25-50 21-50 7-48 10-48 10-48 percentҟ(avg.) 39 40 40: 28 32 32 Temperature, deg. Fҟ(range) 79-84 80-84 78-85 75-87 77-87 77-87 (avg.) 83 82 82 82 83 83