Welding Quality Prediction Method Based on Hidden Markov Model

Abstract

Aiming at the problem that the welding quality cannot be accurately controlled in the welding process, in this paper a welding quality prediction method based on Hidden Markov Model (HMM) is proposed. Based on the welding process parameters collected in history, this method selects the corresponding data segments of various defect states to train the Hidden Markov Model, and obtains the corresponding state models of various welding defects. Finally, by calculating the matching degree between the real-time working parameters and each state model, the welding quality and welding defects are predicted with high accuracy, so as to control the welding quality and provide reference for the adjustment of welding parameters. The performance of the proposal is verified by the experimental platform of TIG welding additive manufacturing.