Prediction of Fundamental Frequency and Critical Buckling Load of a Cracked Cantilever Beam Using Artificial Neural Networks

Alireza Nezamabadi¹ Alireza Pakrooyan²

1) Assistant Professor, Department of Mechanical Engineering, Islamic Azad University of Arak branch,Iran
2) PhD student, Department of Mechanical Engineering, Islamic Azad University of Arak branch,Iran

IOI Code : XAFE-CKKKD
Publication : 1st national conference in mechanical engineering(mechconf.com)
Abstract :
    In this study, a back propagation artificial neural network model with a hidden layer and the Levenberg–Marquardt algorithm and the MSE network performance function was used to predict fundamental frequency and critical buckling load in a cracked cantilever beam. The experimental data required for this neural network are provided from laboratory results. First, the data of all existing tests are divided randomly into three categories: training data, test data, and validation data. Then, the network with a various number of hidden layer neurons and for different transfer functions for all of these data were trained separately, and a model with the lowest error in predicting fundamental frequency and critical buckling load was selected as the optimal neural network model. A comparison was also made between the neural network results with the finite element results. In the end, we examined the effect of various positions of crack on the fundamental frequency and critical buckling load as the neural network innovation which were not available in past papers. A comparison between the experimental results and the neural network showed that the trained network can predict the fundamental frequency and critical buckling load values in the cracked cantilever beam well.

Keywords : fundamental frequency critical buckling load artificial neural networks(ann) levenberg marquardt algorithm network performance function fem