Computational estimation of lateral pile displacement in layered sand using experimental data

In this research, 150 experimental tests were conducted on a series of derived single piles embedded in sand layer. The sand samples were reconstructed with the pluviation method in the two relative densities of 30% (loose state) and 75% (dense state). The pile lateral displacement values were recorded subjected to the combination of the lateral and vertical static loads until failure occurred. In addition, for the estimation of lateral displacement of piles, different Multi-Layer Perceptron Artificial Neural Network models (MLP-ANN) were created and tested. The experimental results showed that when the driving energy increased from 2 J to 4 J, the piles lateral displacement values increased by about 20%. Besides, the measured lateral displacement values of shorter piles were significantly lower compared to the longer piles at the same loading conditions. In addition, the results of computational efforts indicated that the optimum MLP-ANN model presented satisfactory results in the terms of mean square error and regression indices. Moreover, sensitivity analysis showed that pile length and driving energy were the most influential factors on the estimation of lateral displacement recordings.