2007 SEAUPG CONFERENCE-SAN ANTONIO, TEXAS

Transcription

1 of HMA Mixtures to Prevent Fatigue Cracking in Flexible Pavements Fatigue Cracking NCHRP 9-38 Brian Prowell Ray Brown Photo courtesy of FHWA Flexible (Asphalt) Pavement Fatigue Top Down Fatigue Definition of for PCC (Huang) Bottom Up Fatigue Surface Surface Base -2 2 Surface Base Base Sub-base base Sub-base base Subgrade Concept of for HMA (Monismith and Mclean) Long Life Pavement (Nunn)

2 Objectives Confirm existence of Effect of Material Properties on Shortcut method to determine Suggested changes to design guide to include Dr. J. Epps ERES H. VonQuintus Team NCHRP Panel NCAT Dr. E. R. Brown Dr. B. Prowell Dr. D. Timm Dr. S. Maghsoodloo AI M. Anderson Dr. S. Carpenter UNH Dr. Daniel Defining the What is the? HMA Fatigue A level of strain below which there is minimal fatigue damage over an essentially infinite number of loading cycles, which would not lead to failure; failure being bottom-up fatigue cracking Axles loads resulting in pavement strains less than the endurance limit do not cause damage! Idealized Expected Target micro Strain Target micro Strain,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, Number of Cycles to 5% Stiffness Number of Cycles to 5% Stiffness

3 Practical Definition of the Nunn defined long-life life pavement as those that last 4 years without structural strengthening Practical Definition of the Max. passenger cars per hour (65 mph) = 235 One truck replaces 2.5 cars in rolling terrain = 94 trucks/hour; 22,56 trucks/day; 329,376, trucks in 4 years 25% trucks =,52 trucks/day or 48,29,2 trucks in 4 year Class 9 vehicle 4 load repetitions (steer axle would be lighter) = 592,876,8 repetitions for 25% trucks Typically, not all trucks are loaded. Washington DOT study suggests.2 ESALs per Class 9 vehicle or 395,25,2 ESALs for % trucks in 4 years Practical Definition of the 5 million load repetitions is approximate maximum in 4 years Assumes shift factor of supported by SHRP is that laboratory strain that provides for 5 million cycles to failure Test Plan Beam Fatigue Testing Beam Fatigue AASHTO T32 Micro-strain levels: Two replicates at each strain level Test to highest strain level where both replicate survive 5 million cycles

4 NMAS 9. AC% Test Plan Granite Elastomer Optimum B B B, U B, U Opt. +.7 B, U B, U Modified testing to a maximum of 5 million cycles. Testing to be conducted at progressively lower levels until two replicates last 5 million cycles. Additional tests at intermediate strain level to better define endurance limit. B = beam fatigue, U = uniaxial tension Mix Design 9. mm NMAS used at 23 NCAT Test Track Fine-graded granite/limestone blend Optimum AC = 4.5% PG and PG binders PG 64/67-22 Binder Test Data Test Value, kpa Failure Temp. G*/sinδ Orig. at G*/sinδ RTFO at G*(sinδ) ) PAV at Data Analysis Test Results Beam Fatigue Stiffness vs Nf Goal: Accurately Exponential Model estimate fatigue Logarithmic Model life near endurance Power Model limit. Weibull Function Ratio of Dissipated Energy Uniaxial Tension Exponential Model Recommended by AASHTO T32 PG Sample 5 at 7 ms Exponential Life Comparisons Tested to Failure PG at Optimum Flexural Stiffness, MPa y = 397.6e -5E-8x R 2 =.4992 y = e -2E-8x R 2 =.4776 y = 389.5e -E-8x R 2 =.7259 Measured nf,,,,,,,,, Line of Equality 5,, 5,, 5,, 2,, 3,, 25,, 4,, 35,, 5,, 45,,,,,,,,,,, Loading Cycles 5% Initial Expon. (4.E+6) Expon. (.E+7) Expon. (To Failure) Predicted nf (exponential model)

5 Conclusions Regarding Exponential Model Good match to measured failure if sample tested to failure Tends to under-predict fatigue life when extrapolating test results Not recommended for predicting Nf near endurance limit Flexural Stiffness, MPa Logarithmic Model y = Ln(x) R 2 =.974 PG Sample 5 at 7 ms y = Ln(x) R 2 =.9763 y = -6.7Ln(x) R 2 =.98 5,,,, 5,, 3,, 25,, 2,, 4,, 35,, 5,, 45,, Loading Cycles 5% Initial Log. (4.E+6) Log. (.E+7) Log. (To Failure) Flexural Stiffness, MPa million cycles Logarithmic Model 5,, PG Sample 5 at 7 ms y = -6.7Ln(x) R 2 =.98,, 3 million cycles 5,, 2,, 25,, 3,, y = Ln(x) + 43 R 2 = ,, 4,, 45,, 5,, Conclusions Regarding Logarithmic/Power Model Extrapolation from a low number of cycles ( million or less) can result in significant overestimation of Nf When testing to 5 million cycles, important to match slope at high number of cycles May be best method of estimating fatigue life at strain levels equal to or less than the endurance limit Loading Cycles 5% Initial Log. (To Failure) Log. (E+6) Weibull Function S ( t) = exp( λ n S(t) ) = probability of survival until time t λ = scale parameter (intercept) γ = shape parameter (slope) Solved by linear regression of: ln( ln( SR )) = ln( λ) + γ ln( n) SR = stiffness/initial stiffness Applied to HMA by Tsai et al 22 γ ) Ln(-Ln(Stiffness Ratio)) Weibull Function y =.88x R 2 =.976 PG at Optimum Sample 5 at 7 ms y =.7x R 2 =.998 Weibull Function based on testing to failure lies on same line as function from million cycles y =.78x R 2 = Ln (Cycles) Linear (4E+6) Linear (.E+7) Linear (To Failure)

6 ln(-ln(sr)) Weibull Function y = 2.468x R 2 =.995 y =.438x R 2 = ln Cycles Conclusions Regarding Weibull Function Appears data can be extrapolated from a lower number of cycles Prediction of Nf generally good. Believed to underestimate Nf at strain levels less than the endurance limit Flattening of slope near endurance limit may be indicative of endurance limit. Single stage Weibull Function does not account for change in slope (resulting in lower estimate of Nf described above) Sample 2 2 ms Linear (Sample 2 2 ms) Sample 3 ms Linear (Sample 3 ms) PG at Optimum PG at Optimum Indications of the based on Stiffness micro-strain R 2 =.997 Sample 23 Sample 4 Sample 3 E+ E+2 E+4 Cycles to Failure (5% Stiffness) Measured Logarithmic Model 3-Stage Weibull Function Lower Confidence Upper Confidence Lower Prediction Upper Prediction Power (Measured) PG at Optimum PG at Optimum PG at Optimum + PG at Optimum + Sample Sample 5 micro-strain R 2 =.926 Sample 3 micro-strain y = 44.x R 2 =.8775 E+ E+2 E+4 E+6 Cycles to Failure (5% Stiffness) Measured Logarithmic Model Weibull function Lower 95% CI Upper 95% CI Lower 95% PI Upper 95% PI Power (Measured) E+8 E+ E+2 E+4 E+6 E+8 Cycles to Failure (5% Stiffness) Measured Logarithmic Weibull Power (Measured)

7 Prediction Intervals for Endurance limit for PG micro-strain = 5 million cycles 3 micro-strain million cycles Endurance limit for PG micro-strain = 5 million cycles 3 micro-strain = million cycles Visually approximately 225 microstrain,, but results more variable Slight indication of higher endurance limit for optimum plus binder content Ratio of Dissipated Energy Ratio of Dissipated Energy W RDEC = n W W n n+ x x where, RDEC = ratio of dissipated energy W n = total dissipated energy at cycle n W n+x = total dissipated energy at cycle n+x X = the number of cycles between the two data points Shen and Carpenter 25 ΔDE/DE 4.E-3 3.5E-3 3.E-3 2.5E-3 2.E-3.5E-3.E-3 5.E-4 Plateau Value (PV) Region I Region II Region III.E+,, No. of Cycles Ghuzlan and Carpenter 2 Steps to Determining PV Example of Dissipated Energy Data Predicted DE Sample 2 at 2 ms Fit power model to stiffness vs loading cycle data (for tests terminated prior to failure) Early cycles often need to be neglected Determine number of cycles to 5% stiffness Fit power model to dissipated energy vs loading cycle data Early cycles often need to be neglected Calculate RDEC at number of cycles = 5% initial stiffness Predicted DE, kpa ,,,, R^2 =.9 y =.667x ,, 2,, Legend Number Denotes Starting Cycle 25,, Measured Loading Cycles

8 Micro- Strain PG at Optimum Replicate Replicate 2 Nf PV Nf PV E E E E E+7 5.3E E+7 4.7E-9 2.E+7 6.4E-9 9.9E+7 5.4E- 3.6E+2 9.3R-5.9E+3 6.4E-6 Critical PV for long life = 8.57E-9 Plateau Value.... E-5 E-6 E-7 E-8 E-9 E- E- E-2 E-3 E-4 E-5 PV vs. Nf y =.2937x -.5 R 2 =.999 y =.287x NCAT NCHRP 9-38 Cycles to 5% Initial Stiffness Power (NCAT NCHRP 9-38) E+ E+2 E+4 Shen and Carpenter Power (Shen and Carpenter) Round Robin Test Matrix Mini Round Robin for Beam Fatigue Lab/Mix PG at Optimum PG at Optimum Plus PG at Optimum NCAT X X X Asphalt Institute X X X University of Illinois X X X VA Transportation Research Council X SEM Materials X University of California X Code Precision Estimates Average of all Labs Std. Dev. Between Cell Averages (Sx) Repeatability Standard Deviation (Sr) Reproducibility Standard Deviation (SR) Between Lab Standard Deviation of Lab Means(SL) Within- Lab Coefficient of Variation, % Between- Lab Coefficient of Variation, % PG at 8 ms 8,629 4,36 3,396 5,67 3, PG at 4 ms 37, ,965 69,47 372,52 33, PG at 8 ms 9,97 2,88 6,395 6, PG at 4 ms 55,84 34,35 364, ,572 29, Predicting Test three samples each at 4 and 8 micro-strain Transform data (log-log) log) Fit regression Calculate strain level corresponding to 95% one-sided lower prediction limit for 5 million cycles Run three beams at predicted strain level to 2 million cycles Extrapolate with Weibull Function

9 Summary of Predicted s Binder Predicted 95% One-Sided Lower Prediction Limit PG PG PG PG Optimum PG Using for Pavement Design M-E E Methods Design procedures that use equivalent axles and equivalent temperatures (DAMA) Design procedures that use equivalent temperatures but axle load distribution (PerRoad) Design procedures that calculate and use pavement temperatures at specific depths (MEPDG) PerRoad Deflections calculated based on average temperature for up to five seasons Equation to adjust HMA stiffness Uses single definition of endurance limit Uses field transfer function for loads exceeding endurance limit Probabilistic calculations (Monte Carlo Simulation) 23 Test Track Temperature Distribution Histogram 7 2% Frequency % 8% 6% 4% 2% Frequency Cumulative % % Air Temperature, F

10 What is the Effect of Predicted on Pavement Thickness? PG at Optimum AC% Predicted =82 micro- strain - Test Track perpetual thickness = 5 in. 95% One-Sided Lower Prediction Limit = 3 micro-strain Test Track perpetual thickness = 8 in. Test Track Traffic = million ESALs in 2 years Conclusions 5 million load repetitions is a practical maximum for 4 years of traffic Thus, considering a shift factor of, 5 million cycles in the lab approximates the maximum number of load repetitions The single-stage stage Weibull function offers a conservative approach to extrapolate fatigue stiffness data Conclusions (Continued) The Logarithmic model fits samples tested below the endurance limit if some early cycles are ignored The PV identified by Shen and Carpenter indicates long life, but not necessarily the endurance limit Determining the PV requires a double extrapolation Endurance limit can be extrapolated from testing to 2 million cycles Conclusions (Continued) There is an endurance limit An endurance limit of approximately 5 micro-strain was indicated for the PG mix; 225 micro-strain for the PG mix Optimum plus asphalt content may increase the endurance limit slightly Questions? New Contact Information: Brian Prowell Advanced Material Services, LLC 975 Mall Blvd., Suite 22 Auburn, AL 3683 (334) Brian.AMSLLC@CharterInternet.com