Surrogate based prediction models for elastically supported beams under the action of opposite moving loads

Abstract

A hybrid model that incorporates an artificial neural network (ANN) and non-dimensional numerical analysis has been developed to predict the displacement response of an elastically supported (ES) beam under the action of sequential moving loads in opposite directions. The dataset acquired from non-dimensional numerical modeling was utilized for training multi-layered feed-forward ANN, resulting in the development of surrogate models. The input parameters include the beam and load parameters, whereas the beam’s displacement is the output parameter. The robustness and efficiency of the hybrid models have been highlighted using statistical metrics, which show that training and validating data produce promising predictions. Further, the relationship between input and output parameters was represented using Pearson’s correlation. Results indicated that the effectiveness of the ES beam depends on the critical combination of stiffness parameter κ and interspatial distance parameter ϵ under particular speed parameter R. To bypass the computational cost, a user-friendly interface has been developed to assess the displacement response of elastically-supported beams under the action of one-way and two-way moving loads. Further, the proposed methodology has wide application in analyzing and designing short to medium-span bridges in critical locations such as canal/road crossings under the action of opposite moving loads using elastic support conditions.