Calibration Approach

The conventional approach to calibrating an ME method, which has been used since calibration of the Shell Method and Asphalt Institute Method in the 1970s and early 1980s, through calibration of the Mechanistic-Empirical Pavement Design Guide (ARA, 2004) consists of the following:

Identify short sections of pavement

oPreferably most of the pavements have some failure on them, otherwise the time to failure would be uncertain because it hasn’t occurred yet.

oThe sections need to have a construction time history.

Collect the materials properties on the test sections.

Backcast the traffic and materials properties to the time of construction.

Simulate the performance using measured materials properties using Miner’s Law, which has the following issues:

oThe response engine calculating critical stresses, strains, and deformations is unverified.

oThe damage evolution and predicted state of damage on the section is also unverifiable because use of Miner’s Law forces the shape of the damage evolution curve.

oOnly the end state of distress is used for calibration.

Find calibration coefficients for the calculated damage-to-distress transfer function to minimize the errors between observed and measured distress.

Use the variability around the minimized error transfer function for reliability.

The conventional approach has several limitations:

•It requires expensive and time-consuming sampling and testing of materials properties for each section, resulting in a small number of sections being available for calibration.

•It ignores the fact that a design-bid-build (low-bid) designer does not know the performance-related properties of the materials the contractor will bring to the job; this results in a blurred understanding of the sources of variability and their consideration in the design reliability approach.

The new calibration approach developed by the UCPRC to calibrate CalME aims to improve calibration and the reliability approach used in ME design by doing the following:

Use all the good quality distress performance data and as-built data in the Caltrans PMS databases collected since 1978 and quality checked over the last 10 years; this provides orders of magnitude more performance data for calibration, with the data organized by project.

Use median properties to match median performance, and use the variability of observed median performance to determine between-project variability, after using CalME to account for the effects of climate, pavement cross section, and traffic.

oThe weighted average performance of a set of mixes from the UCPRC databases was used to represent the time periods present in the cracking and rutting performance data since 1978.

Backcalculate within-project variability by matching the shape of observed performance time history.

It was assumed that calibration using the very large amounts of data available in the PMS performance data and representative mix data for that time period would provide a more comprehensive calibration than just using detailed sampling and testing of materials from a few projects. Once the calibration is completed, comparisons with sections with detailed sampling and testing can provide additional validation.

The new approach also explicitly separates within-project and between-project variability in the calibration, and in the design method. This allows for use of appropriately different between-project reliability factors for PRS and non-PRS projects. The need for explicit consideration of between-project reliability and the inherent problem of calibrating using measured materials properties in a design-bid-build approach is expressed in the following excerpt from the MEPDG report:

From MEPDG report Section 3.3.2 OVERVIEW OF FLEXIBLE PAVEMENT DESIGN PROCESS

3.3.2.1 Design Inputs Trial Design Inputs and Site Conditions

A major difficulty in obtaining adequate design inputs is that the desired project specific information is not generally available at the design stage and must often be estimated several years in advance of construction. The actual materials used in a project may not 3.3.4 even be known until a few weeks before construction begins. The designer should obtain as much data as possible on in-situ material properties, traffic, and other inputs for use in design to obtain a realistic design. The designers should also conduct a sensitivity analysis to identify key factors that affect pavement performance. Based on sensitivity analysis results, provisions could be made in the contract documents for stringent control of the quality of key material properties (e.g., asphalt concrete stiffness), or the design could be modified to make the pavement performance less sensitive to the input in question. (ARA, 2004)

The inclusion of between-project reliability in CalME v3.0 overcomes the need for the sensitivity analysis in the project design process called for in the MEPDG report to assess the range of potential materials that might be delivered to the project—which depends on who wins the design-bid-build contract.

ARA Inc., Guide for Mechanistic-Empirical Design of New and Rehabilitated Pavement Structures, ERES Consultants Division, ARA Inc. 2004. Prepared by the National Cooperative Highway Research Program, Transportation Research Board, National Research Council.