Research on neural network method for online temperature measurement of FSW

The friction stir welding process is used for the fuel tank of heavy launch vehicles. Due to the increase in the size of the fuel tank and the large temperature difference between the long and thick welding seams, it is difficult to achieve the stability of the welding process, resulting in a significant decrease in the strength and performance of the welds. This paper studies the temperature signals in the core region of friction stir welding process, establishes a real-time temperature detection system for friction stir welding, and further studies and constructs a BP neural network model to predict welding temperature, laying a foundation for temperature control and improving welding stability in friction stir welding. The results show that the established real-time temperature detection system for friction stir welding can meet the requirements of engineering applications while effectively detecting the welding temperature; The BP neural network model established based on the temperature measurement system can predict the welding temperature using the spindle rotation speed, feed speed, and pressing amount.