Weld Geometry Monitoring for Metal Inert Gas Welding Process with Galvanized Steel Plates Using Bayesian Network

Abstract

We present a novel method to monitor the weld geometry for metal inert gas (MIG) welding process with galvanized steel plates using Bayesian network (BN), and propose an effective method of extracting the weld reinforcement and width online. The laser vision sensor is mounted after the welding torch and used to profile the weld. With the extracted weld geometry and the adopted process parameters, a back propagation neural network (BPNN) is constructed offline and used to predict the weld reinforcement and width corresponding to the current parameter settings. A BN from welding experience and tests is presented to implement the decision making of welding current/voltage when the error between the predictive geometry and the actual one occurs. This study can deal with the negative welding tendency to adapt to welding randomness and indicates a valuable application prospect in the welding field.