Use of PDA Dynamic Test Results in Determining the Bearing Capacity of Single Pile in Marine Structures by Artificial Neural Networks

In this study, using the results of 100 PDA dynamic loading tests obtained from different projects and using three types of artificial neural networks (ANN), the loading capacity of a single pile is evaluated. Initially, the Perstron multilayer neural network was used as one of the most widely used neural networks. In the following, a combination of neural-fuzzy networks is used from the nephrophysical network, and at the end, the neural network is used as a function of the radial basis of the successful network in nonlinear problems. Unlike conventional behavioral models, neural network-based models do not explain how input parameters affect output. In this research, by performing sensitivity analysis on the optimal structure of the models introduced in each stage, an attempt has been made to examine this ambiguity to some extent. Also, introducing the relationships governing a neural network model can give engineers more confidence in using them to facilitate analysis and design.