Independent Review of the Mechanistic Empirical Pavement Design Guide and Software (2006)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

6. Provide better guidance on the use of grids, fabrics, and stress-absorbing membrane interlayers (SAMIs) for mitigating reflection cracking. 7. Improve the quality of pavement evaluation information for rigid pavements and the advice on design of rehabilitation using PCC. 9 CONCLUDING DISCUSSION The Guide in its present form is not suitable for immediate use by design engineers. It is a substan- tial piece of research that, with the further work being undertaken, could be developed into a pow- erful tool for design. It embraces innovations and combines for the first time in a single computer pro- gram several important aspects of design previously used independently. Inevitably, because of the vari- able levels of knowledge and development in the enormous number of aspects to pavement design, the soundness of the underlying engineering princi- ples varies considerably. The field calibration took a considerable effort by the 1-37A contractors, but it is inevitable that huge variability should have been experienced with use of the kind of data generated by LTPP. Therefore, the reliability of the designs is questionable. In addition to carrying out further calibration/validation work using more reliable performance data, the vast expe- rience of designs used in the past should be accom- modated. Any new method must provide solutions that agree with good performance in the past in com- parable circumstances. The advantage of an analyti- cally based method is that it can meet new conditions (in terms of materials, climate, and types of con- struction) in addition to standard situations. The importance of rehabilitation for the future is not adequately recognized in the Guide. Many improvements could be made based on experience elsewhere; these have been set out in Brown et al. (3). The various recommendations set out in this re- port, and supported by more detailed discussion in the companion reports (1–3), have been put forward by the review team with a view to improving the quality and reliability of the Guide as successive versions are developed and it moves toward accep- tance by the pavement engineering community. The recommendations are based on extensive experience in all aspects of pavement design, including an in- ternational perspective. The software can be a user-friendly package for engineers, but all the reported difficulties with its use should be corrected and the recent improvements should be incorporated. Before a usable Guide can be introduced to design practice, a number of strands of current and recent research need to be accom- modated to provide improvements. In addition, a professionally planned beta testing program is rec- ommended for the software. The final product must work well and reliably; otherwise, it will be very dif- ficult for confidence to be gained in its use by design engineers. The software will need to be accompa- nied by a well-written, professionally presented user guide that provides sufficient background and guid- ance, but is also far shorter than the comprehensive 1-37A research reports. PART II Detailed responses of the Project 1-40D team to the essential recommendations made by the Project 1-40D technical panel and the three independent re- view teams follow. (Note: N/P = not planned for Versions 0.900 or 1.0). 15 Panel Recommendations Area-# Recommendation 1-40D Team Response Version Eliminate calculation of IRI.P-1 N/P (continued) Disagree. AASHTO pavement design has consid- ered the highway user through serviceability (or smoothness) for 40+ years, and this vital criterion should continue to be included. The IRI models include indirectly the long-term profile effects not included in the flexible or rigid M-E models, such as profile changes as function of age, climate, and subgrade properties.

16 Panel Recommendations (Continued) Area-# Recommendation 1-40D Team Response Version Include endurance limit in flexible and rigid cracking models. P-2 0.900 On the rigid side, IRI works well. The standard error is good and IRI is a direct function of the M-E main distress types predicted. On the flexible side, the existing models have rela- tively high errors and are missing or insensitive to key distresses. New, improved IRI predictions models for new HMA, HMA over HMA, and HMA over PCC are under development. These new models are being de- veloped using a much larger and more recent LTPP data set that includes the following: • New HMA—GPS 1 & 2, SPS 1, 5, & 8 (>1200 time series data points). • HMA/HMA—GPS 6, SPS 5 (>700 time series data points). • HMA/PCC—GPS 7, SPS 6 (>300 time series data points). The new IRI models include key M-E-predicted dis- tresses, including rut depth, top-down and bottom-up fatigue cracking, and transverse cracking. Their effect on IRI is more significant than the existing models. Also included are pavement foundation (subgrade) properties known to affect the potential for foundation pavement movement (swelling and heaving), along with climatic factors that drive foundation movement. The improved models have considerably reduced model prediction errors. When considering the stan- dard error of the IRI models, it must be realized that the overall error (scatter of data points around the pre- dicted versus measured IRI one-to-one plot) includes IRI measurement error, section replication (or pure) error, and section input error as well as the true model error (lack of fit). The true model error is believed to be significantly less than the total error. Also, a much more robust data set is now available : >2000 obser- vations, including older pavements and better repre- sentation of climate, subgrade, and traffic conditions experienced in the United States. The Project 1-37A panel directed inclusion of IRI in the MEPDG. IRI should be retained; the Version 0.900 recalibration will substantially improve the error in the IRI calculations. Agree. No applicable results were available to the 1-37A team during the development of the MEPDG. However, endurance limit can be easily implemented in both flexible and rigid designs. The MEPDG software will be modified to include the endurance limit by keeping a tally of the loads that exceeded the limit and outputting that quantity or

17 Panel Recommendations (Continued) Area-# Recommendation 1-40D Team Response Version Re HMA PD, turn off per- manent deformation model? Incorporate repeated load testing parameters? Don’t use LTPP data. Set up so that all calibration is done on a local rather than na- tional basis. P-3 P-4 N/P 1.0 0.900 and 1.0 percentage. Inclusion of an endurance limit has no practical effect on the flexible design results because the analysis is still the same; Projects 1-40D and 9-38 will provide additional information on this topic. However, this capability will be included with an upper limit of 100 microstrain. For rigid design, a stress-to-strength ratio will be in- cluded to implement an endurance limit; the percentage of damage caused by ratios below 0.5 will be reported. Disagree. What is the downside of considering only fatigue cracking? We believe that it would be a major step backward for several reasons. First, the recali- brated model will be improved over the existing model and future models will be improved over this one; over time and use, a very robust permanent deformation model will evolve. Next, such an approach will tend to overestimate the thickness and can’t be used for war- ranty or PRS applications. Third, the effect of poor mixtures on premature failures cannot be known. Elim- inating permanent deformation would be tantamount to deleting an aspect of pavement design that has the most effect on highway safety. Last, there would be no way to derive equivalent flexible and rigid pavement designs using a criterion such as limiting strain only. Agree in principle. The use of repeated load test parameters to supplement the E* modulus will be evaluated in Version 1.0. Disagree. LTPP is the largest single source of high- quality, reliable data collected anywhere in the world. Contrary to popular belief, a well-constructed, statisti- cally balanced national calibration exercise should be able to account for all local issues, provided the data are well represented in the database. Robust national cali- bration is needed as a foundation for local calibration. However, agencies can use local data to further refine nationally calibrated models. Not every agency will or needs to conduct a local calibration. Therefore, local cal- ibration is an option, not a requirement. Further, it is un- clear how the validity of local calibration can be judged without comparison with national calibration statistics. The calibration data set being used in NCHRP Project 1-40D includes other data than LTPP, including MnRoad, WesTrack, the AASHO Road Test, NCAT test track, and other field pavement experiments. Local calibration data provides a means to reduce bias (over or under prediction) and the standard error of estimate. At least one-half of the errors in the current models are input, distress measurement, and pure (continued)

18 Panel Recommendations (Continued) Area-# Recommendation 1-40D Team Response Version Incorporate new E* model. Include FWD back-calcula- tion capability in the rehab design. Re change in design philos- ophy, calculation of perfor- mance prediction versus use of limiting strains? P-5 P-6 P-7 1.0 1.0 or later (replication) errors. Further, LTPP has formal and standardized methods of collecting distress data, which may or may not be the case for locally collected data. So, it is unclear how the validity of local calibra- tion can be judged without comparison with national calibration statistics. However, we will test this hypothesis in Version 0.900 by taking out state data sets for comparison. Robust national calibration as a foundation for local calibration is needed. For Arizona, use of the national calibration as a starting point gave excellent results. Not every state will conduct a local calibration. In Version 1.0, statistical jackknifing techniques to test national calibration could be applied. LTPP data continue to be improved, cleaned up, and updated (Version 0.7 was calibrated with data collected through 2000, whereas Version 0.900 will use perfor- mance data collected through 2004). The Project 1-40D contractors have also reviewed and improved the data. Inclusion of modified calibration factors in Version 1.0 will permit the effective use of local calibration data. Agree. This will be done in Version 1.0. Agree. However, the Project 1-37A panel directed that no specific back-calculation computational method be used. If resources are available, future work can incor- porate a general ARA, Inc., interface program that will be able to run several popularly used back-calculation programs. This program will be made available to AASHTOWare, pending further discussions. This program includes several available back-calculation methods and data-processing capabilities. Disagree. The incremental damage and performance prediction approach was directed by the Project 1-37A panel over a 6-year period. As stated above, going back to a limiting strains concept would make it all but impossible to develop equitable flexible and rigid designs and would greatly limit the capability and usefulness of the design method. In practice, any per- formance type can be “turned off” by the user, by changing the output instructions to allow display of only selected performance data. The MEPDG provides a framework for going forward in “bite size” portions. Version 0.900 will be much more robust in areas where there are interactions of different levels of input data. What are the benefits of having capabilities that are not available in the present AASHTO guide? The MEPDG allows comparison of equivalent HMA and PCC designs and provides a tool for LCCA. The

P-8 P-9 P-10 P-11 P-12 19 Panel Recommendations (Continued) Area-# Recommendation 1-40D Team Response Version Turn off HMA top-down cracking model. Unbound layer rutting. Re viability of the CRCP cracking model, incorporate Version 10 of the Zollinger model. Improve the HMA thermal cracking model. Incorporate placeholders for future capabilities. N/P 1.0 N/P 0.900 1.0 AASHTO guide overestimates the thickness needed for both HMA and PCC pavements. The capability for Monte Carlo simulations is needed; this was di- rected by the Project 1-37A panel, but could not be accomplished within the time and resources available to that project. It is believed that, overall, it will be far better to begin to implement the full MEPDG now, even with some known model deficiencies, than to revert to limiting strain or other criteria only. Clearly, large errors are inherent in this approach. Future research will enhance all of the existing models in the MEPDG, but this can only happen if they can be evaluated across the country at the same time. Disagree. This model will be recalibrated for Ver- sion 0.900 using a larger dataset. At present, no other viable model is available. We suggest this model remain in the MEPDG with the option to turn on a display of its results. Agree. Will further consider in Version 1.0. However, unbound layer modulus is a function of the material type. There are already two models for unbound materials in the software (same form, different shift fac- tors). Lytton also used a single model for a variety of materials. Dr. Sherif El-Badwy’s Ph.D. thesis at Arizona State University found that one model form was appro- priate for a variety of subgrade types in Arizona. The new ICM will improve estimates of rutting in unbound layers. The Version 0.900 recalibration with cleaned-up LTPP data will reduce prediction errors significantly. Disagree. Version 10 is not recommended for inclu- sion at this time by either Dr. Won or Dr. Zollinger for CRCP. It is recommended that future inclusion of various aspects of Version 10 technology be fully considered. Agree. The use of additional sections and data does im- prove the Level 3 predictions, but with some inconsis- tency in the results. However, the model may require revision; this work is in progress, including necessary laboratory testing. Dr. Bill Buttlar is under contract to work on this issue for NCHRP. ASU has prepared spec- imens and is testing at different binder contents to check the thermal cracking model with Level 1 inputs. Agree. The software can include an initial screen with a summary of impending or potential changes due to re- search projects in progress on such critical topics as en- durance limit, top-down cracking, and local calibration.

20 Recommendations of 1-40A(01) Flexible Design Team Area-# Recommendation 1-40D Team Response Version Check consistency of guide and software. Fix all software bugs identi- fied during the public evalua- tion and independent review. Moisture-modulus reduction equation. Incorporate new G* and E* prediction models developed at ASU. Include an option to use the Hirsch model for E*. Resolve time of loading and frequency issues. Evaluate default modulus at OMC values in Table 2.2.51. Remove top-down cracking model. De-emphasize IRI predic- tions in evaluation of flexi- ble designs. Incorporate improved “robust” models for rutting and fatigue cracking. F-1 F-2 F-3 F-4 F-5 F-6 F-7 F-8 F-9 F-10 0.900 and 1.0 0.900 and 1.0 0.900 1.0 N/P 1.0 0.900 and 1.0 N/P 1.0 N/P Agree. Version 0.900 will be accompanied by a sup- plement to the design guide that will describe changes in the software and how they affect the guide itself. Agree. Has been or will be done in Versions 0.8 and 0.900. Agree. Accomplished in Version 0.8. Agree. Will be done in Version 1.0. Disagree. Comparisons show poorer results than with Witczak E* and G* equations (see the results based on 5000 data points in Javed Bari’s Ph.D. dissertation). The present form of the Hirsch model requires different inputs and is based on a smaller dataset. Agree. Will study this issue and incorporate changes in Version 1.0 if possible. This will include the addition of an optional capability for inputting layer moduli. Frequency is defined as 1/t, not as 1/ω. This is the def- inition used in engineering, physics, and chemistry, and is the appropriate form for this application. Agree. This will be done and the software updated as necessary. Two sets of default values are provided at present, one at OMC and the other at equilibrium. Disagree. This model will be recalibrated for Ver- sion 0.900 using a larger dataset. At present, no other viable model is available. We suggest this model re- main in the MEPDG with the option to turn on a dis- play of its results. We believe it is far better to imple- ment a model with some deficiencies than to ignore the distress altogether. Disagree. A new, up-to-date, and greatly expanded data set from LTPP has been obtained and new IRI models developed that show lower standard error and account for all the MEPDG-predicted distress types for flexible pavements. Disagree. See responses to items P-2, P-3, and F-11. Calculated strain provides a connection with repeated load deformation. The new version of the ICM in Version 0.900, which will incorporate the Thorntwaite moisture index and provide a better indication of base and subgrade mois- ture, will improve estimates of rutting in unbound layers. Recalibration with cleaned-up LTPP data will reduce prediction errors significantly. Absence of

21 Recommendations of 1-40A(01) Flexible Design Team (Continued) Area-# Recommendation 1-40D Team Response Version In fatigue, provide a generic form of the equation. Include an endurance limit. Provide PD models for each distinct type of material in a flexible pavement structure. In Level 2, include monthly inputs for all layers of the pavement section. Focus on local rather than national calibration. 50% minus 200 material criterion too high? F-11 F-12 F-13 F-14 F-15 1.0 0.900 and 1.0 1.0? 0.900/1.0 N/P measured trench data puts a lower limit on the error, but trenching will be done in 9-30A. This is a theoreti- cally sound model. Lytton’s model, looked at origi- nally, gave much poorer results. Agree. Will consider for Version 1.0. N cannot be accurately calculated without considering E*. K3 can already be set to zero if desired, but this would neces- sitate a recalibration. This form may not give the expected results in the program. A relation between K1 and K2 will add complexity to the calculation. A strain limit will be added. Agree. Will be considered in Version 0.900 recalibra- tion. However, unbound layer modulus is a function of the material type. There are already two models for unbound materials in the software (same form, differ- ent shift factors). Lytton also used a single model after investigating 40 different soils. Sherif El-Badwy’s Ph.D. dissertation research found that one model form was appropriate for a variety of subgrade types in Arizona. Although the basic model is the same, it is currently calibrated for fine- or coarse-grained materials. Will further consider in Version 1.0. Disagree, but can test in Version 1.0 if so directed. This is already done for unbound materials, but if it is implemented for HMA, the effects of aging will be eliminated. HMA modulus depends on frequency and temperature. Model already predicts modulus on a monthly basis. Will not permit interaction between mix and structural designs, nor will it account for aging over time. Would also affect calibration, proba- bly to a high degree of error. Disagree, but the results of Project 1-40B will be adopted in Version 1.0. Will also test in Version 0.900 by taking out state data sets for comparison. Need ro- bust national calibration as a foundation for local cali- bration. For Arizona, use of the national calibration as a starting point gave excellent results. Not every state will conduct a local calibration. In Version 1.0 could apply statistical jackknifing techniques to test national calibration. LTPP data continues to be improved and cleaned up; 1-40 contractors have also reviewed and improved on the data. Inclusion of modified calibra- tion factors in Version 1.0 will permit the effective use of local calibration data. Disagree with using 35%. Will change the write-up in the design guide supplement to provide justifica- tion for value of 50% and discuss how it will be applied. (continued)

22 Recommendations of 1-40A(01) Flexible Design Team (Continued) Area-# Recommendation 1-40D Team Response Version Present defaults for AASHTO classes only. CSM design is not state of the art. Use NLA information. No consideration of cold- mixed asphalt-treated granu- lar materials. Uncouple the component programs: EICM, JULEA, and FEM. Provide selected response outputs such as stress, strain, and displacement. F-16 F-17 F-18 F-19 F-20 N/P 1.0 1.0 N/P 1.0 Disagree. Better to have both because the two classifi- cations do not match. However, the team will look again at how closely the Unified and AASHTO classes match. Agree. This major effort will be considered in Version 1.0. There was a paucity of data for cement-treated bases in the initial calibration dataset. Agree. Will consider in Version 1.0. There is a lack of a specific model form for cold-mixed materials, so would use same model as for asphalt concrete. It will be included in the pull-down menu as another layer type. Disagree. Stand-alone versions of all these programs exist now that are readily available to the technical community. Agree in principle. Can be done in Version 1.0, but at a reduced number of calculation points. Otherwise, the running time of the program will be unacceptably increased. An alternate but still undesirable solution would be to run JULEA as a stand-alone program. Recommendations of 1-40A(02) Rigid Design Team Area-# Recommendation 1-40D Team Response Version Overall design process, input requirements: Inputs that have little effect on de- signs or those that cannot be determined when designs are being made should be shown and “grayed out” so they cannot be changed without justification and significant effort by the design engineer. Overall design process, traffic inputs: Simplify the traffic inputs. They are too complex and incorporating design inputs into the design procedure is time consum- ing. Consider incorporation of traffic input program TRAFLOAD for inputting traffic data. Software validation and verification: All software should be beta tested before releasing it to design agencies. R-1 R-2 R-3 0.900 and 1.0 0.900 and 1.0 1.0? Agree. 1-40D has already grayed out or made some inputs more difficult to change. These include perma- nent curl/warp and others related to shrinkage. Others that meet the suggested criteria will also be identified and either grayed out or placed elsewhere. Alterna- tively, default values may be tailored to the require- ments of individual states in future versions. Agree. Some changes have already been made. The software has been modified to accept outside files con- taining all required input traffic data. Version 1.0 work will provide the capability to use the TRAFLOAD product for this purpose. However, this will not sub- stantially simplify the design process because all of the same data are required by TRAFLOAD. Most traffic inputs are also required for the current 1993 AASHTO Guide in order to compute ESALs properly. Agree in principle. The 1-40D team will alpha test new versions of the software, but there are no funds available to conduct beta testing. This could be done as part of Version 1.0 if funds are provided, or it may be a task best left to AASHTOWare.

23 Recommendations of 1-40A(02) Rigid Design Team (Continued) Area-# Recommendation 1-40D Team Response Version Design manual editing: Manuals accompanying the MEPDG software must be carefully edited so that state- ments in the manuals are consistent with the software. Develop regional/state calibrations for JPCP: Prominently display ranges of values and regions of the U.S. used when developing the calibration coefficients. JPCP curling/warping and permanent warping: Give special attention to curling and warping during and after hardening. Improve the estimation of the Effective Permanent Curl/Warp value to account for additional factors known to affect this parameter. JPCP, Evaluation of climatic effects: The methods by which climatic conditions are used in the design pro- cedure and the relationship between curling and warp- ing must be carefully docu- mented. Among the factors to consider: 1. More precisely explain the stresses and condi- tions that cause top-down and bottom-up cracking and the means to separate these distresses in the de- sign process. 2. Discuss how creep was taken into account in early-stage warping and curling process. R-4 R-5 R-6 R-7 1.0? 1.0 1.0 or later 0.900 and/or 1.0 Agree. Text changes necessitated by corrections or en- hancements of the software in Version 0.900 will be in- cluded in the supplement to the design guide planned at the end of Version 0.900. A full review of the consistency of text and software is desirable, but will require funds above what is now planned for Stages 0.900 and 1.0. Agree. The 1-40D team strongly supports the concept of regional or state calibration of the MEPDG. Such recommendations were made in the 1-37A documen- tation. Such an effort is under way in several states at this time, and Project 1-40B is developing a manual to guide such activities. The display of regional values used to develop regional calibration coefficients is a good idea and will be done in Version 1.0 (in coordi- nation with Project 1-40B). Agree. In Project 1-37A, significant effort was spent in trying to accomplish this recommendation. The 1-40D team will make a major effort to develop a better procedure to estimate this input during the design stage. Insufficient data as to construction conditions for the LTPP and other calibration sections limit the ability to completely fulfill this recommendation. However, this recommendation can be partially addressed in Version 1.0 or later by renewing efforts to relate the permanent curl/warp parameter to preva- lent construction, materials, and climatic conditions as available data permit. Agree. 1. A more detailed account will be included in the supplement provided with Version 0.900. 2. Creep is indirectly considered in the permanent curl/warp value because it is based on long-term performance of JPCP and CRCP which gives the slabs the opportunity to settle into the base course. (continued)

24 Recommendations of 1-40A(02) Rigid Design Team (Continued) Area-# Recommendation 1-40D Team Response Version 3. More thoroughly explain curl stresses in slabs. 4. Separate and display the warping, curling, and load-associated stresses as an optional output for the designer. CRCP, Validation and Streamlining of Input Data: Justify and validate the inputs required in the design pro- cedure. Procedure requires substantially more inputs than current design proce- dures. Not all inputs are fully justified or validated. Make efforts to classify inputs assigning levels of varying importance or sig- nificance to each, so that the design process can be simplified. CRCP, Thickness Design Consistency: In some in- stances, thickness require- ments obtained using the MEPDG are not consistent with current practice. This issue needs to be resolved or reasons for the differences adequately expressed. R-8 R-9 0.900 and 1.0 0.900 and 1.0 3. A more thorough explanation will be included in the supplement. 4. Complete separation is not possible because, to make the process computationally efficient, load, thermal, and moisture stress computations are intertwined in the MEPDG software. Agree. This recommendation can be carried out in full. A fuller explanation of inputs and their effect and significance will be prepared as part of the supplement to the design procedure. Regarding the number of inputs for CRCP as compared with the current AASHTO procedure, the following results were found: AASHTO: 94 inputs MEPDG: 130 inputs This comparison does not include axle load distribu- tion for either the AASHTO guide or the MEPDG and monthly vehicle volume adjustments for the MEPDG. The increase in the number of inputs for the MEPDG is not dramatic considering the analytical power it affords the designer. Some inputs have already been “grayed out” or made more difficult to change. These include permanent curl/warp and shrinkage inputs. Others that meet the suggested criteria will also be identified and either grayed out or made difficult to change. Disagree. The conclusion that CRCP thickness is too great is based on one reference prepared by Roesler and Kohler at the U. of Illinois. The reviewers state that this reference concludes that “the thickness of CRCP is as much as 30 percent greater than slabs for JPCP.” Communication with Drs. Roesler and Kohler deter- mined that neither of them had made such a compari- son and that the statement attributed to them is incor- rect. They prepared an illustrative design of a CRCP for very heavy traffic in the Chicago area and obtained a thickness of 14 in., similar to what IDOT is building for very heavy traffic loads. We verified their CRCP results and then used the same inputs for a JPCP pave- ment with the result that a 15-in. JPCP slab would be needed. Many other direct comparisons of CRCP and JPCP have been made and CRCP is generally thinner or occasionally equal to JPCP. Other comparisons of CRCP thickness were made between the AASHTO guide and the MEPDG. The AASHTO method always gives a thicker CRCP

25 Recommendations of 1-40A(02) Rigid Design Team (Continued) Area-# Recommendation 1-40D Team Response Version CRCP, Zero-Stress Verification: The zero-stress temperature (permanent curl/warp) input has a sig- nificant effect. Guidelines must be provided to assist the designer in selecting the appropriate input for this variable for different condi- tions and locations. Factors that affect this input include concrete temperature during hydration, climatic condi- tions at time of placement, and quantity and type of cement. CRCP, Alternative Failure Mechanisms: Punchout mechanisms other than the one in the design procedure need to be considered, evaluated, and validated. R-10 R-11 1.0? 1.0 or later design (as it does with JPCP). Texas DOT sponsored a major sensitivity analysis of CRCP and found no issues with thickness. Missouri DOT compared JPCP and CRCP for a given project and determined that a 10-in. CRCP was approximately equivalent to a 12-in. JPCP. Thus, there appears to be no evi- dence that the MEPDG produces unreasonable thickness results. However, this issue does indicate a need to reduce the standard error of model prediction as much as possible (the punchout model has a relatively high standard error) and major efforts will be made in the 1-40D recalibration effort to improve the CRCP design method so that better predictions will be made and the standard error reduced. Agree. There are two inputs mentioned here: zero stress (temperature at set) and permanent curl/warp. The zero-stress temperature is critical to CRCP and both are critical to JPCP. Indeed, this recommendation is similar to R-6 above for JPCP. During Project 1-37A, major resources were spent in trying to accomplish this recommendation for JPCP, but similar resources were not available for CRCP. The 1-40D team will make a major effort to estimate this input specifically for CRCP. Our ability to do this is limited, however, by insufficient data on construction conditions for LTPP section and other calibration sections. Note: the MEPDG specifies that crack widths should be < 0.02 in, crack LTE > 95%, and crack spacing < 6 ft over the design life. If these criteria are met, punchouts will rarely be a major problem. The MEPDG includes the classic punchout mecha- nism identified many years ago as the major distress type in CRCP by several researchers based on field and theoretical observations. The mechanism calcu- lates the transverse tensile bending stress at the top of the CRC slab between the truck wheels (of an axle) placed between two narrowly spaced transverse cracks. The transverse cracks may (<95% LTE) or may not (>95% LTE) be deteriorated. The rectangular piece of PCC may or may not punch down into the base. More than 80% of the so-called “punchouts” for sections used to calibrate the punchout model did not punch down, but simply showed the longitudinal fatigue crack (top down). (continued)

26 Recommendations of 1-40A(02) Rigid Design Team (Continued) Area-# Recommendation 1-40D Team Response Version The 1-37A team took great pains in putting together the performance data for the original calibration; this included obtaining and reviewing year-by-year distress maps from each LTPP section considered and manually verifying the database. In this review the team observed that the predominant form of distress is the classical longitudinal crack between two transverse cracks, usually 3 to 5 ft from the edge of the slab. Occasionally this rectangular piece of PCC punched down into the base forming the classic punchout. However, Y-cracks and “ladder” cracks/punchout clusters were also observed, albeit not as commonly. The punchout count from the LTPP and other databases used in the calibration also included Y-cracks that had punched down (medium to high severity). Other mechanisms such as single deteriorated transverse cracks were not included in the count. The 1-40A reviewers identified one punchout mecha- nism they thought was not included in the MEPDG: “. . . some recent research results that indicate loss of load transfer is not the only condition for the develop- ment of punchouts. Under controlled test conditions it has been shown that punchouts can develop even though there is no measurable loss in load transfer across the cracks.” Agree. Transverse cracks do not need to deteriorate; permanent deformation or a major settlement of the foundation can lead to a longitudinal crack in the CRCP. However, the MEPDG does not assume that transverse cracks must deteriorate. Fatigue damage is accumulated at the top of the slab with every load ap- plication even if no crack deteriorates. This mecha- nism is already included in the MEPDG. In summary, this recommendation for consideration of other punchout mechanisms is reasonable and can be done in Version 1.0 or later if the specific mechanisms can be identified. The supplement will include a clearer discussion of the conditions where transverse cracks do not deteriorate and of other potential punchout mechanisms. A change of terminology from punchout to “short longitudinal fatigue crack” may also be help- ful since the former term has created confusion.

F-SW-1 F-SW-2 F-SW-3 F-SW-4 F-SW-5 F-SW-6 F-SW-7 27 Recommendations of 1-40A(03) Team on Flexible Design Area-# Recommendation 1-40D Team Response Version Omit the top-down cracking distress prediction compu- tations and await the results of ongoing research on this subject. Improve the application of soil mechanics principles to the characterization of soils and unbound materials, and, in particular, omit the use of CBR in design. Put the rut depth prediction procedures in abeyance and replace with good HMA mixture design combined with a permanent deforma- tion test and the introduction of allowable stress levels for the granular layers and subgrade. Provide in the software for input of traffic data from the program developed under NCHRP Project 1-39. Introduce the FWD back- analysis subroutine into the software to deal with non- linearity of the lower layers of the pavement. Improve the guidance to users of the Guide within a more succinct, user-friendly version of the document. In revising the Guide for use in the short term, a more balanced approach is needed in which the various ele- ments of design are approxi- mately of the same level of complexity. N/P N/P N/P N/P 1.0 1.0 or later N/P N/P Disagree. Will recalibrate for Version 0.900. Leave it in, but disabled with the option to turn it on in the analysis. No other viable model available. Much more data is now available. Agree in principle. Continued improvement in the treatment of soils and unbound materials is both nec- essary and desirable. However, the MEPDG uses modulus as the input, not CBR. Disagree. While a sound HMA mix design is always required, there is a serious downside to putting the rut depth prediction in abeyance. Such an approach will tend to overestimate the required layer thickness, and it is not compatible with warranty or PRS applications. Further, the Version 0.900 recalibration with cleaned- up and larger LTPP and other datasets will substan- tially improve the model correlation and further reduce its standard error. Disagree. The use of allowable stress levels for granular layers and subgrade is similar to the CBR method in which a great enough layer thickness is used to reduce the stress level below the allowable limit. While this methodology can be added to the MEPDG as a separate, optional design criteria if directed by the panel, it is not recommended. Agree. This capability is planned for incorporation in Version 1.0. Agree. However, the Project 1-37A panel directed that no specific back-calculation computational method be used. If resources are available, future work can incor- porate a general ARA interface program that is licensed for use to AASHTOWare; this program includes several available back-calculation methods. Agree in principle. However, no resources are cur- rently available to carry out this effort. [Note: A proj- ect to accomplish this recommendation is tentatively planned for FY 2007.] Agree in principle. The Project 1-37A panel directed a flexible “mix and match” approach in which the designer can choose the same or different levels of complexity for the various design elements. At the most basic level (Level 3), default values can be used exclusively, but combining Level 2 and 3 values or Level 1 and 2 values will reduce the error and increase the reliability of designs.

28 Recommendations of 1-40A(03) Team on Rigid Design Area-# Recommendation 1-40D Team Response Version JPCP and CRCP New Designs: Include longitudinal and corner cracking as dis- tress mechanisms and provide improved advice with regard to raveling. JPCP and CRCP New Designs: Review the perma- nent warp assumptions when a granular base is used. JPCP and CRCP Rehabilitated Designs: Improve procedures for structural evaluation of concrete pavements. Include longitudinal cracking as one of the distresses considered in pavement evaluation. JPCP and CRCP Rehabilitated Designs: Correct error noted in Equa- tion 3.2.28, which forms part of the computation of load transfer efficiency (LTE) between slabs and across cracks. Review determination of dowel LTE contribution, where there may be an error in the equations leading to very high LTE values under all conditions. Check the advice on computa- tion of base LTE contributions (Tables 3.4.8 and 3.7.20). R-SW-1 R-SW-2 R-SW-3 R-SW-4 1.0 or later 0.900 0.900 0.900 Agree. The 1-37A research team was directed to not include corner cracking. The current MEPDG frame- work can easily incorporate a corner cracking model; this distress is significant in non-doweled JPCP but does not occur in properly doweled JPCP. Longitudinal cracking is a potentially significant type of distress that could be added to the JPCP analysis. This distress appears to occur in dryer climates in the western United States. Longitudinal cracking initiat- ing at the top of the slab is directly calculated in the CRCP design procedure; the critical transverse stress location is identified and fatigue damage is accumu- lated at this point. It is certainly possible to apply a similar algorithm to JPCP. Additional resources would be required to accomplish this. Agree. This excellent suggestion will be carried out during the Version 0.900 recalibration. The effect of base course type (including granular base) on the permanent curl/warp parameter for JPCP and CRCP will be directly considered during the recalibration effort. It is possible that the parameter will be reduced when a softer base course is used. Agree. The relevant design guide text will be modified in the supplement to include longitudinal cracking of JPCP in the pavement evaluation process. Agree to all. Equation is being reviewed and will be corrected as needed. Equations are being reviewed and will be corrected as needed. A check of base LTE contribution will be made to en- sure reasonableness.

R-SW-5 R-SW-6 29 Recommendations of 1-40A(03) Team on Rigid Design (Continued) Area-# Recommendation 1-40D Team Response Version JPCP and CRCP, Check and correct the detail concerning base erodibility, upward curl, and overburden on subgrade in relation to the computations for faulting: In Part 2, base erodibility is described in terms of Class rather than Factor. A unit is required for upward curl. The unit for overburden on subgrade should be appropri- ate for pressure (psi) rather than force (lb). JPCP and CRCP, Short- comings in the calibration procedures for rehabilitation models: Calibration work for JPCP restoration should be carried out using data from sites covering a wider range of conditions. Further data are needed on the performance of CRCP overlays to existing PCC pavements for heavily traf- ficked sites. Prediction of transverse cracking and joint faulting in JPCP overlays does not appear to match the avail- able evidence particularly well, and the issue should be investigated further, both by seeking additional evidence and by carrying out a wider range of comparative designs. This may eventually identify areas where the model could be improved. 0.900 0.900 and 1.0 Agree to all. These items are being checked and corrected as needed. A new base erosion model has been developed that considers the loss of support under CRCP more realis- tically. The definition of base erosion will be made clearer for JPCP. Agree in principle, but the available data are what they are. We will search for additional data for restoration proj- ects. However, all relevant data from LTPP and other sources was used for this purpose in Project 1-37A. Project 1-40B identified some additional CRCP overlay performance data from Illinois from both experimental sites and one in-service project. The recalibration of new JPCP will hopefully provide improved modeling that may show better prediction for JPCP overlays.