Approximate entropy GRNN forecast for aluminum alloy pulsed MIG welding stability

The model and sample data that able to reflect the stability of the different welding process parameters were obtained through analyzing welding voltage signals in aluminum alloy pulsed MIG welding by approximate entropy. On this basis, a method that predicts the approximate entropy of the voltage signals by generalized regression neural network (GRNN) was proposed. The structure and algorithm of the GRNN prediction model were introduced and the prediction experiments on ample data were done. The results show that the average error of the predictive value is 9.08%, the accuracy rate of it is 90.92%, and the results meet the forecast accuracy of aluminum alloy pulsed MIG welding process stability.