Beam yielding load identification by neural networks

Abstract

The paper presents the application of Artificial Neural Networks for the identification of the load causing a partial yielding in the cross-section of a simple supported beam. The identification of the load was based on a change of the dynamic parameters (eigenfrequencies) of the partially yielding structure. On this basis and using neural networks a tool for the location and evaluation of the load causing the deformation was built. The optimum network architecture, learning algorithm, number of epochs, and the minimum number of eigenfrequencies have been found. In order to come to the final conclusions, a wide variety of network architectures (from simple networks with four neurons in one hidden layer to complex networks consisting of two or three simple networks), learning algorithms and different numbers of learning epochs have been tested.