Application of artificial neural network models to predict ultrasonic welding parameters for joining copper to aluminium sheet

Abstract

Ultrasonic welding is a method of combining solid-state components that does not require the use of heat or flux and instead relies on vibration to hold the components together. When connecting copper wires to an equipment alternator in electrical and automotive applications, a solid-state joining method is utilized as the preferred connection method. This situation calls for the utilization of ultrasonic welding of metal. In businesses that use ultrasonic metal welding processes, issues with weld quality and strength appear to be the most frequently encountered difficulties. These issues are frequently the result of inadequate parameter selection, which may be traced back to the welding process. To produce ultrasonically welded joints with a thickness of 0.3 mm, the welding characteristics of copper to aluminum connections, such as welding pressure, weld duration, and vibration amplitude, are considered in this study. Compared to the size of this weld, alternative joint construction methods could potentially cause damage to the joint. To conduct the tests, a suitable experimental design was conceived and put into action by employing Taguchi's robust design process in a manner that was tailored to the specifics of the situation. With the help of a synthetic neural network, the answer was predicted, and its accuracy was checked using the mean square relative error, the Nash-Sutcliffe efficiency, and the mean absolute percentage error.