Using multiple neural networks to estimate the screening effect of surface waves by in-filled trenches

Trenching is an economical and effective method to reduce surface vibrations and isolate structures from shaking. Previous reports on vibration screening concentrated on either experimental work or analytical study. Due to the construction of more complex structures in the last two decades, presenting more complicated boundary conditions, a variety of numerical methods have been used. Complexity of formulation, the large number of parameters involved, and the difficulty and time required to analyze an effective vibration screening makes the direct numerical approach impractical. The purpose of this paper is to explore the use of an artificial neural network to estimate the effectiveness of a vibration screening trench. Three artificial neural networks, BPN, GRNN, and RBF, are used to evaluate the performance of a chosen physical model. The results show that all three models can be used to evaluate effectiveness of screening trenches with varying accuracy, with GRNN having the highest accuracy. There is much stronger agreement with data of numerically calculated results for neural networks than for empirical multi-variate regression methods.