Critical Responses of Flexible Pavements Under Superheavy Loads and Data-Driven Surrogate Model

Superheavy Load (SHL), a specially manufactured vehicle for transporting superheavy cargo, such as wind turbines, has weight, size, and loading configurations, differs from general truck traffic, e.g., the 13 vehicle classes defined by Federal Highway Administration (FHWA). To characterize non-generic configurations of SHLs, an efficient and structured analysis method is needed to predict unexpected damages that can occur in flexible pavement subjected to SHL loading. In this paper, existing methodologies including the superposition method and the nucleus segment approach are introduced to characterize the loading range and magnitude of each SHL. To identify potential damages by SHLs on flexible pavement, a set of experimental matrices considering pavement properties and loadings from the nucleus segment of each SHL is established-i.e., a total of 3456 cases of flexible pavement analysis, varying in thickness and modulus of elasticity of each layer, and in types of loading, are performed using a Layered Elastic Theory (LET)-based analysis program, MnLayer. As a result of the mechanistic investigation, critical pavement responses under SHLs and FHWA class 9 truck (reference vehicle) are determined. Furthermore, damage ratios using transfer functions available in Mechanistic-Empirical Pavement Design Guide (MEPDG): A Manual of Practice are calculated for each SHL compared to the reference vehicle, FHWA class 9 truck. Finally, an advanced statistical prediction model, Generalized Additive Model (GAM), is constructed as a data-driven surrogate model to provide relatively high performance in predicting target responses of flexible pavements from varying explanatory variables (e.g., flexible pavement properties and loading conditions of SHLs).

Automated summary

This paper introduces an efficient and structured analysis method for predicting damages that can occur in flexible pavement subjected to superheavy load (SHL) loading. Existing methodologies including the superposition method and the nucleus segment approach are used to characterize the loading range and magnitude of each SHL. A set of experimental matrices considering pavement properties and loadings from the nucleus segment of each SHL are performed to identify potential damages by SHLs on flexible pavement. Critical pavement responses under SHLs and a reference vehicle (FHWA class 9 truck) are determined and damage ratios are calculated for each SHL compared to the reference vehicle. An advanced statistical prediction model, Generalized Additive Model (GAM), is constructed as a data-driven surrogate model for predicting target responses of flexible pavements.