2appendices_NN

Sensitivity Analysis (Test Model over Range of Input Parameters)

The final step in model verification and validation was sensitivity analysis. Sensitivity analysis is the determination of how variations in input parameters quantitatively affect predicted distress and IRI. The goals of a sensitivity analysis are as follows:

• If the PCC design distress prediction models can simulate observed trends in the development and progression of distress.
• What factors contribute to variability in the predicted distress (and may require additional research to strengthen the knowledge base).
• What model parameters (or parts of the model itself) are significant or insignificant?
• If and which (group of) input parameters interact with each other.

To run an effective sensitivity analysis it must be ensured that the entire range of each significant input parameter is examined individually along with their interaction with other input variables (e.g., interaction between PCC flexural strength and elastic modulus). The general approach adopted in this study was to repeatedly run each new PCC design distress prediction model for different combinations of input parameters within the expected inference space for PCC rehabilitated pavements (e.g., unbonded JPCP overlay thickness ranges from 6 to 15 in) and then evaluating the predicted distress and IRI for reasonableness. The steps used in sensitivity analysis are as follows:

1. Determine key input parameters.
2. Assign typical ranges for the key input parameters selected.
3. Create virtual PCC rehabilitated pavement test sections based on the parameters selected and their ranges.
4. Run the new PCC design distress prediction models to predict distress and IRI for each virtual PCC rehabilitated test section.
5. Assess the reasonableness of the predicted distress and the relative importance and sensitivity of each input parameter.

Reasonableness was assessed by determining the percentage of the total amount of variation in predicted distress that each input parameter accounts for (when all others input parameters are kept at some constant level) and observing the trends in predicted distress with increasing age or traffic applications.

4.0 Preliminary Verification of All Algorithms and Models

Four scenarios representing different rehabilitation alternatives were used to preliminarily determine the reasonableness of all algorithms and models used for new PCC design and adopted for PCC rehabilitation design. The algorithms and distress models evaluated were as follows:

• Algorithms.
  • JPCP joint load transfer efficiency (LTE).
  • JPCP differential energy (DE).
  • JPCP fatigue damage (top-down and bottom up).

6

Acknowledgment of Sponsorship

This work was sponsored by the American Association of State Highway and Transportation Officials (AASHTO) in cooperation with the Federal Highway Administration and was conducted in the National Cooperative Highway Research Program which is administered by the Transportation Research Board of the National Research Council.

Disclaimer

This is the final draft as submitted by the research agency. The opinions and conclusions expressed or implied in this report are those of the research agency. They are not necessarily those of the Transportation Research Board, the National Research Council, the Federal Highway Administration, AASHTO, or the individual States participating in the National Cooperative Highway Research program.

Acknowledgements

The research team for NCHRP Project 1-37A: Development of the 2002 Guide for the Design of New and Rehabilitated Pavement Structures consisted of Applied Research Associates, Inc., ERES Consultants Division (ARA-ERES) as the prime contractor with Arizona State University (ASU) as the primary subcontractor.

Research into the subject area covered in this Appendix was conducted at ARA-ERES. The author of this Appendix is Mr. Leslie Titus-Glover. Dr. Darter provided technical and managerial coordination of the Rehabilitation with PCC design group, monitored progress, set schedules and deadlines, and provided periodic technical review of research results as they became available. Mr. Jag Mallela provided assistance with technical and managerial coordination.

Foreword

This appendix presents a detailed summary of the verification and validation (and when required re-calibration) of distress prediction models originally developed for new PCC (jointed plain concrete pavement [JPCP] and continuously reinforced concrete pavement [CRCP]) design to determine their suitability for rehabilitation of existing pavements (flexible, composite, and rigid) with PCC design. PCC rehabilitation of existing pavement design alternatives considered in the Design Guide and described in this appendix are as follows:

• JPCP restoration.
• Unbonded JPCP or CRCP over existing PCC.
• Bonded PCC over existing JPCP or CRCP.
• JPCP or CRCP overlay over existing flexible pavement.

The information contained in this appendix serves as a supporting reference to PART 3, Chapter 7 of the Design Guide.

2

APPENDIX NN

CALIBRATION OF REHABILITATION OF EXISTING PAVEMENTS WITH PCC

1.0 Introduction

The objective of NCHRP Project 1-37A was to develop a guide for the design of new and rehabilitated pavement structures based on mechanistic-empirical (M-E) principles. An M-E based Design Guide will provide the highway community with a state-of-the-practice tool for the design of pavement structures representing a major paradigm shift in current empirical design procedures. The PCC rehabilitation design procedure outlined in this Guide is very similar to that of new PCC pavement design described in Part 3—DESIGN ANALYSIS, Chapter 4 of this Guide. Key aspects of design such as the pavements material, traffic, and climate characterization are essentially the same. Because of the mechanistic based principles used, the models and algorithms used to compute pavement responses such as stress, strain, deflections, damage, and distress are applicable to both new PCC design and PCC rehabilitation design.

Mechanistic analysis make it possible to extend the use of distress prediction models developed for new PCC design to PCC rehabilitation design. However, the PCC design models used in PCC rehabilitation design must be verified and validated to ensure their suitability for predicting distress within the inference space of PCC rehabilitated pavements.

This appendix presents a detailed summary of the verification and validation (and when required re-calibration) of distress prediction models developed for new PCC (jointed plain concrete pavement [JPCP] and continuously reinforced concrete pavement [CRCP]) design to determine if they are suitable for use for PCC rehabilitation design of existing pavements (flexible, composite, and rigid). PCC rehabilitation of existing pavement design alternatives considered in the 2002 design procedure were:

• JPCP restoration (i.e., concrete pavement repair [CPR] including diamond grinding) for existing JPCP only.
• Unbonded JPCP over existing PCC (JPCP, jointed reinforced concrete pavement [JRCP], and CRCP).
• Bonded PCC over existing JPCP or CRCP.
• JPCP or CRCP overlay over existing flexible pavement.
• Reconstruction with JPCP or CRCP.

Verification, validation, and recalibration was done by applying both analytical and statistical techniques. Where the results of verification and calibration showed that the new PCC design distress prediction models were inadequate the PCC distress models were recalibrated to determine the best set of model coefficients that could be used to predicted distress with accuracy while minimizing error. The distress models presented in this appendix are for PCC rehabilitation design only. A description of HMAC rehabilitation design distress models are described in Appendix MM. Also, a comprehensive description of the M-E based design models

3

for new PCC pavement design have been presented and discussed in several chapters of the 2002 Design Guide and in several appendices. These chapters and appendices describe in-depth how the models were developed, how the models were applied, and how input parameters such as pavement material and design properties, traffic, and climate were characterized.

2.0 Overview of Model Verification and Validation

The goal of verification and validation is to determine whether distress models for new PCC design when applied to PCC rehabilitation design would predict distress accurately with minimal bias and reasonable error. When properly done, verifying and validating a model should lead to improving the model's credibility with users and improve upon the reliability of the entire PCC rehabilitation design procedure. Verification and validation for this study is defined as follows:

• Verification is comparing the conceptual model and algorithms with actual field observed trends and estimates or trends and estimates determined from mechanistic analysis (1).
• Validation is ensuring that a model within its domain of applicability possesses a satisfactory range of accuracy consistent with the intended application of the model (2).

3.0 Framework for Model Verification and Validation

Model verification and validation consisted of the following steps:

1. Conduct a preliminary verification of all algorithms and models.
2. Comparison of model distress predictions to observed distress (measured from in-service pavements).
3. Perform a comprehensive sensitivity analysis.

Preliminary Verification of All Algorithms and Models

Preliminary verification of algorithms and models consisted of determining whether all the algorithms and pavement response models used in estimating stress, deflection, load transfer, crack width, and distress for new PCC design models work reasonably well within the inference space for PCC rehabilitation design. As an example, can the algorithm developed to estimate stress and deflections in a new JPCP over an asphalt treated base be used to determine the same stresses and deflections for a JPCP overlay over an existing HMAC? Note that AC-treated materials and HMAC could have very different gradations, resilient modulus, binder type and content, voids, and so on.

Preliminary verification of algorithms and models was done as follows:

1. Identify different PCC rehabilitation design scenarios that represent selected PCC rehabilitation alternatives (e.g., JPCP restoration, unbonded JPCP over existing PCC, bonded PCC over existing CRCP, and so on).
2. Develop input parameters for each scenario (i.e., each PCC rehabilitation alternative).

4

Comparison of Model Distress Predictions to Observed Distress

3. Use the inputs for each scenario to estimate key pavement responses such as stress, deflection, joint load transfer efficiency (LTE), damage, and so on and predict distress (transverse joint faulting and transverse cracking for JPCP and punchouts for CRCP).

4. Examine all outputs for reasonableness (values of estimates, trends, etc).

It can be concluded that the new PCC design models are suitable for PCC rehabilitation design if the outcome of steps 1 through 4 is determined to be reasonable. Where the outcome was determined not to be reasonable it was necessary to modify the new PCC design algorithms and distress models to improve upon their applicability to PCC rehabilitation design or modify calibration constants, or both.

Comparison of model predicted distress and actual field measurements of distress is the next important step in model verification and validation as this shows the ability of the new PCC design models to accurately predict distress for PCC rehabilitation design. Comparisons were made by applying the following steps:

1. Assemble a database of in-service PCC rehabilitated test pavements. The database must contain the test pavements design, material, construction, climate, traffic, and measured pavement distress and IRI.
2. For each test pavement, run the relevant new PCC design distress models using the input data assembled in step 1.
3. Compare the new PCC design models predicted distress to measured distress (from the in-service pavement test sections) and evaluate the predictive capacity and accuracy of the distress models.

The predictive capacity and accuracy of the distress models were evaluated using statistical analysis as follows:

1. Determine the correlation between predicted and measured distress (R²).
2. Determine the residual error between predicted and measured distress (mean square error, [MSE]).
3. Use paired t-tests to determine whether there is a significant difference on average between measured and predicted distress for the in-service pavements analyzed.

The statistical analysis was conducted using data obtained from in-service rehabilitated pavements located in North America. Note that comparisons of model predicted distress and measured distress was done only after the new PCC design algorithms and models had been determined to be reasonable based on the preliminary analysis described in the preceding sections.

5

Sensitivity Analysis (Test Model over Range of Input Parameters)

The final step in model verification and validation was sensitivity analysis. Sensitivity analysis is the determination of how variations in input parameters quantitatively affect predicted distress and IRI. The goals of a sensitivity analysis are as follows:

• If the PCC design distress prediction models can simulate observed trends in the development and progression of distress.
• What factors contribute to variability in the predicted distress (and may require additional research to strengthen the knowledge base).
• What model parameters (or parts of the model itself) are significant or insignificant?
• If and which (group of) input parameters interact with each other.

To run an effective sensitivity analysis it must be ensured that the entire range of each significant input parameter is examined individually along with their interaction with other input variables (e.g., interaction between PCC flexural strength and elastic modulus). The general approach adopted in this study was to repeatedly run each new PCC design distress prediction model for different combinations of input parameters within the expected inference space for PCC rehabilitated pavements (e.g., unbonded JPCP overlay thickness ranges from 6 to 15 in) and then evaluating the predicted distress and IRI for reasonableness. The steps used in sensitivity analysis are as follows:

1. Determine key input parameters.
2. Assign typical ranges for the key input parameters selected.
3. Create virtual PCC rehabilitated pavement test sections based on the parameters selected and their ranges.
4. Run the new PCC design distress prediction models to predict distress and IRI for each virtual PCC rehabilitated test section.
5. Assess the reasonableness of the predicted distress and the relative importance and sensitivity of each input parameter.

Reasonableness was assessed by determining the percentage of the total amount of variation in predicted distress that each input parameter accounts for (when all others input parameters are kept at some constant level) and observing the trends in predicted distress with increasing age or traffic applications.

4.0 Preliminary Verification of All Algorithms and Models

Four scenarios representing different rehabilitation alternatives were used to preliminarily determine the reasonableness of all algorithms and models used for new PCC design and adopted for PCC rehabilitation design. The algorithms and distress models evaluated were as follows:

• Algorithms.
  • JPCP joint load transfer efficiency (LTE).
  • JPCP differential energy (DE).
  • JPCP fatigue damage (top-down and bottom up).

6