APPROXIMATE EIGENVALUE OF PLATE BY ARTIFICIAL NEURAL NETWORKS

The general goal of this paper is to determine natural frequency of a plate by artificial neural network with various supporting conditions. One of the most famous training of neural network is back propagation algorithm. This algorithm is a systematic method for training multi-layer artificial neural network. Back propagation algorithm is based on gradient descant which means that it moves downward on the error declination and regulates the weights for the minimum error. In this research، the real frequency is calculated using ANSYS program and is defined as a goal function for neural network so that all outputs of the network can be compared to this function and the error can be calculated. Then using a set of inputs including dimensions or specifications of plate، a neural network is made. After the determination of algorithm and quantities of the network، the phases of training and testing of the results are carried out and the output of the network is created. It is concluded that the results show the performance of the neural network and that the time of frequency calculation is considerably reduced.