Prediction of Weld Strength in Power Ultrasonic Spot Welding Process Using Artificial Neural Network (ANN) and Back Propagation method

Abstract

In this presented work, an Artificial Neural Network (ANN) connected with the backpropagation method was employed to predict the strength of joining materials that were carried out by using an ultrasonic spot welding process. The models created in this study were investigated, and their process parameters were analyzed. These parameters were classified and set as input variables like applying pressure, time of duration weld and trigger of vibrating amplitude. In contrast, the weld strength of joining dissimilar materials (Al-Cu) is set as output parameters. The identification from the process parameters is obtained using several experiments and finite element analyses based on prediction. The results of actual and numerical are accurate and reliable; however, their complexity has a significant effect due to being sensitivity to the condition variation of welding processes. Therefore, an efficient technique like an artificial neural network coupled with the backpropagation method is required to use the experiments as input data in the simulation of the ultrasonic welding process, finding the adequacy of the modeling process in the prediction of weld strength and to confirm the performance of using mathematical methods. The results of the selecting non-linear models show a noticeable potency when using ANN with a backpropagation method in providing high accuracy compared with other results obtained by conventional models.