Incorporation of Reliability into Mechanistic-Empirical Pavement Design

Abstract

Pavement thickness design traditionally has been based on empiricism. However, mechanistic-empirical (M-E) design procedures are becoming more prevalent, and there is a current effort by AASHTO to establish a nationwide M-E standard design practice. Concurrently, an M-E design procedure for flexible pavements tailored to conditions within Minnesota has been developed and is being implemented. Regardless of the design procedure type, inherent variability associated with the design input parameters will produce variable pavement performance predictions. Consequently, for a complete design procedure, the input variability must be addressed. To account for input variability, reliability analysis was incorporated into the M-E design procedure for Minnesota. Monte Carlo simulation was chosen for reliability analysis and was incorporated into the computer pavement design tool, ROADENT. A sensitivity analysis was conducted by using ROADENT in conjunction with data collected from the Minnesota Road Research Project and the literature. The analysis demonstrated the interactions between the input parameters and showed that traffic weight variability exerts the largest influence on predicted performance variability. The sensitivity analysis also established a minimum number of Monte Carlo cycles for design (5,000) and characterized the predicted pavement performance distribution by an extreme value Type I function. Finally, design comparisons made between ROADENT, the 1993 AASHTO pavement design guide, and the existing Minnesota design methods showed that ROADENT produced comparable designs for rutting performance but was somewhat conservative for fatigue cracking.

Variability in Pavement Design

“… The coefficient of variation appears, on average, to range from 20 to 40%, which is in line with the results found by Noureldin et al (1994) and Timm et al (2000), see Table 1. …”

Reliability in pavement design

“…The analysis has been carried out for both the base and the binder course layers. The coefficient of variation appears, on average, to range from 10 to 30%, which is in line with the results found by Noureldin et al (1994) and Timm et al (2000), see section 2.8. …”

“…Therefore, the traffic weight is the single most important input parameter (Timm et al 2000, Shekran et al 2002, Cunagin and Kent 1998.  Timm et al (2000) found that the output distributions of both fatigue and deformation performance, obtained from Monte Carlo simulation by using the input distributions, may well be approximated by an extreme-value Type I distribution (sometimes referred to as Gumbel distribution).…”

“…Simple simulation techniques are also often used to assess the pavement performance (Lemer and Moavenzadeh 1971, Timm et al 2000, Hong and Somo 2001. All of these techniques basically attempt to evaluate the multidimensional probability integral that defines the structural failure.…”

“…The use of the Monte Carlo Simulation (MSC) technique to evaluate the variation in predicted performance based on the variability of a set of input parameters for a specified pavement design section is recommended by many authors (Graves and Mahboub, 2007;Darter et al 2005;Zhang and Piepmeyer, 2005;Timm et al, 2000). The use of this technique in the context of reliability analysis in pavement design has been discussed in chapter 3.…”