Cantilever Beam Natural Frequency Prediction Using Artificial Neural Networks

Available methods for calculating frequency in cantilever beams have much complexity. In this study we present a new method for calculating natural frequencies in cantilever beams. For this purpose, we use the finite element method (FEM), dynamic analysis and artificial neural network (ANN) techniques to calculate the natural frequency. Finite element software was used to analyze 100 samples of cantilever beams, and the results were used as training and testing data sets in artificial neural networks. For the ANN. the multilayer feed-forward network and back-propagation algorithms were used. We made use of different transfer functions and built 45 different networks in order to find the best network performance. Mean squared error (MSE) was used to evaluate the network performance. Finally, the natural frequencies which were predicted by the ANN techniques were compared to the natural frequencies calculated from theoretical formulation, as well as to those obtained from FEM methods. The results obtained show that the error was quite small.