Online Intelligent Identification of Resistance Spot Welding Quality Based on HMM for Stainless Steel Body of Metro Vehicle

Abstract

A new online monitoring device of resistance spot welding quality parameter for stainless steel body of metro vehicle was introduced. For multiple pulse welding process of resistance spot welding of thick stainless steel, there are multiple pulse waveforms on the electrode displacement curve. Therefore, quality estimation of resistance spot welding can't be done using the traditional maximum electrode displacement method. In the article, the intelligent identification method based on Hidden Markov Model (HMM) was suggested for quality estimation of resistance spot welding. The electrode displacement curves under different welding current were classified into three groups, which are qualified welds, unqualified welds and splash welds. The splash welds can be monitored using the sudden drop of electrode displacement curve easily. Two HMMs were trained using Baum-Welch algorithm for qualified welds and unqualified welds. Then, the welding samples to be verified were identified using Forward-Backward algorithm. It was resulted that the 100% of welds were identified correctly using HMM method.