Analysis of Moving Loads on Beams Using Surrogate Models Inspired from Artificial Neural Network

The problem of moving sequential loads over beams is a classic example of vehicle-bridge interaction. Many analytical, numerical, and experiments have been conducted in order to study the dynamic behavior of bridges. However, the computational cost involved in traditional methods is very high. The primary novelty of this paper lies in the creation of a non-dimensional mode-superposition to anticipate the bridge's peak dynamic responses, coupled with the application of ANN-based Multi Input Multi-Output metamodeling. The bridge is idealized as a Euler–Bernoulli beam of uniform cross-section subjected to sequential moving loads commonly known as High Speed Load model-B (HSLM-B). The pearson’s correlation coefficient indicates that the most influential parameter affecting the dynamic response of beams under moving loads is the interspatial distance between loads, i.e., ε followed by speed parameter η. Additionally, the comparison of the best-fitted surrogate models has been conducted to evaluate their robustness and efficiency. The results demonstrate an accuracy level of less than 10 percent, highlighting the high precision and reliability of the surrogate models.