Real-time temperature monitoring of weld interface using a digital twin approach

Abstract

A new friction processing technique is developed for lap welding of Al–Cu sheets using inter diffusion of copper and aluminium at their interface. The metals selectively melt at the interface at a temperature near their eutectic point. This paper proposes a digital twin approach for real-time temperature monitoring at the joint interface using the machine’s real-time current data. The real time temperature data is used to predict exact instance of interface melting and to control the resulting weld microstructure. The digital twin model is calibrated using a finite element model which is in turn calibrated using experiments. Moving average of machine current and temperature history are used to predict real time interface temperature using a linear regression based recursive machine learning model with high precision. The model predictions have an R2 value of 99.5%. The digital twin approach resulted in significant increase in joint strength and fracture energy.