Time-domain sparse Bayesian learning with PSO algorithm for CA mortar void detection of slab track system

A newly developed time-domain sparse Bayesian learning methodology with the particle swarm optimization (PSO) algorithm was proposed for the void identification in the cement-emulsified asphalt (CA) mortar of the slab track system for the first time. The CA mortar void identification process involves an iterative expectation maximization technique in the sparse Bayesian learning framework to calculate the damage parameters and hyperparameters for the CA mortar stiffness distributions utilizing measured time-domain vibration data. The PSO algorithm was incorporated to minimize the objective function for obtaining the most probable values of damage parameters. Comprehensive numerical case studies were firstly carried out to validate the feasibility of the proposed methodology, and then the effects of accelerometer placement on the CA void detection results were investigated. In order to further experimentally demonstrate the applicability of the proposed methodology, impact hammer tests were conducted on the scaled slab track models. Encouraging void identification results indicate that the CA mortar void location and severity can be successfully identified with very high accuracy by utilizing the presented methodology, and the associated posterior uncertainties were calculated by utilizing the posterior covariance matrix of the damage parameters, which were kept at an acceptable level.