LoboVault Home

Sensitivity of MEPDG Using Advanced Statistical Analyses

LoboVault

Please use this identifier to cite or link to this item: http://hdl.handle.net/1928/12063

Sensitivity of MEPDG Using Advanced Statistical Analyses

Show full item record

Title: Sensitivity of MEPDG Using Advanced Statistical Analyses
Author: Sumee, Nasrin
Advisor(s): Tarefder, Rafiqul A
Committee Member(s): Storlie, Curtis B
Stormont, John C
Department: University of New Mexico. Dept. of Civil Engineering
Subject: MEPDG Sensitivity
Degree Level: Masters
Abstract: Recently, pavement design has changed from old method based on empirical relation to a new method called Mechanistic Empirical Pavement Design Guide (MEPDG). It is essential to perform a detailed sensitivity analysis of MEPDG outputs to input variables. In particular, MEPDG inputs need to be classified based on their influence on MEPDG outputs for New Mexico pavement conditions. In this study, sensitivity analyses are performed to identify a list of input variables that have significant impacts on the MEPDG outputs considering New Mexico pavement conditions. Sensitivity analyses are performed in two steps. In the first step, a preliminary sensitivity analysis is carried out by varying one input variable at a time while keeping the other inputs constant. The purpose of the preliminary sensitivity analysis is to prepare a short-list of significant input variables out of more than hundreds of variables in MEPDG. In the second step, sensitivity analyses are performed using advanced statistical approaches that consider interactions among the input variables. Parametric procedures such as tests for nonrandomness in scatterplots, linear and nonlinear regression analyses, and nonparametric procedures such as multivariate adaptive regression spline, gradient boosting machine are employed to identify and rank the significant input variables. Results show that predicted pavement performances are sensitive to traffic input variables such as Annual Average Daily Truck Traffic (AADTT) and percent of trucks in design lane. Both asphalt surface layer and total rutting are shown to be the most severe cases among all distresses for New Mexico pavements. Both AC and total rutting are highly sensitive to AADTT, percent of trucks in design lane, and bottom AC layer thickness. Outputs such as terminal IRI, longitudinal cracking, and alligator cracking are highly sensitive to bottom AC layer thickness. MEPDG outputs are also sensitive to HMA mix properties such as thickness, percent air void, binder content and PG grade. Longitudinal and transverse cracking are sensitive to base course material type, modulus and thickness. Depth of water table did not affect the MEPDG outputs at all. Transverse cracking and total rutting are sensitive to subgrade modulus, material properties, and gravimetric water content. MEPDG predicted outputs are found to be moderately sensitive to percent of trucks in design direction, traffic growth factor, and base thickness. Operational speed, depth of ground water table, and design lane width have very little to no effect to MEPDG predicted distresses. Finally, a list of significant variables is made for New Mexico pavement conditions. The list of significant inputs can be useful to pavement engineers to optimize pavement designs and analyze performances as well as for local calibration of MEPDG.
Graduation Date: December 2010
URI: http://hdl.handle.net/1928/12063


Files in this item

Files Size Format View
SENSITIVITY OF ... STATISTICAL ANALYSES1.pdf 3.268Mb PDF View/Open

This item appears in the following Collection(s)

Show full item record

UNM Libraries

Search LoboVault


Browse

My Account