Artificial neural network to predict the weld status in laser welding of copper to aluminum

Laser welding of copper to aluminum is challenging due to the formation of complex intermetallic phases. More Al (~18.5 at. %) can be dissolved in Cu, in contrast to Cu (~2.5 at. %) in Al. Therefore, welding from copper side, large melting of Al can be achieved. However optimum Cu and Al must be melted for a strong joint. Finding the right amount is difficult and time consuming by tradition analysis technique like inspection by weld cross-sections. Considering the speed of the welding process and complexity of analysis involving with metallography cross-sections, alternative rapid method to qualify the welds are necessary. The acoustic emission during laser welding can give proportional information of the Al, Cu melted. With such an approach the weld status can be obtained in real time. In this paper the acoustic welding signal using an airborne sensor in the audible range of 20 Hz to 20 kHz, is correlated to the weld strength and material mixing (Al, Cu melt). Finally, the weld status is predicted by an artificial neural network based on the acquired signal.