Estimation of FWD Parameters for Evaluation of the Quality of Portland Cement Concrete Pavement

Abstract

This study aims to accurately predict two important parameters for evaluating the quality of Portland Cement Concrete (PCC) pavement: the modulus of subgrade reaction (Z1) and the elastic modulus of concrete slab (Z2). To achieve this, advanced Machine Learning (ML) models were used, including ANN-TLBO, ANN-BBO, and ANN-GA. These hybrid models combine Artificial Neural Network (ANN) with optimization techniques such as Teaching Learning-Based Optimization (TLBO), Biogeography-Based Optimization (BBO), and Genetic Algorithm (GA). The dataset used for modeling consists of 510 Falling Weight Deflectometer (FWD) tests from National Highway 18 in Quang Ninh Province, Vietnam. Standard statistical measures were used to validate and compare the performance of the models. Results showed that all three models performed well, with ANN-TLBO achieving the best results for predicting Z1 and Z2. Thus, the ANN-TLBO model can be used for accurate prediction of these important parameters for evaluating PCC pavement quality.