Neural Network Modeling and Parameter Prediction for Arc Length of Pulsed Gas Metal Arc Welding

Aiming at the five core pulse parameters of pulsed gas metal arc welding, such as the base current, pulse current, pulse time, pulse frequency and current rising speed, the arc length changes during welding process under different parameter combinations are collected by using the high-speed photography system, and based on 60 experimental results, a BP neural network prediction model for arc length is established. The correlation coefficient R²=0.91, and the prediction error range is -7.065 2%,7.301 0%. The BP neural network prediction model can well reflect the variation trend of arc length affected by the pulse parameters in single factor prediction.