Adaptive matching of electrical parameters in gas metal arc welding based on audio sensing

Abstract

In Gas Metal Arc Welding (GMAW) of non-standard products, the welding electrical parameters (welding current and arc voltage) require frequent adjustment and matching according to varying workpiece conditions. This parameter matching process is cumbersome, time-consuming, and significantly influenced by the operator. To address this, an adaptive welding parameter matching method is proposed. Three classes of typical welding audio signals under different parameter-matching states were collected. Features of loudness, pitch, and timbre were extracted frame by frame and combined into a feature matrix. Principal component analysis was then performed on the feature matrix for supervised classification. Based on this, a closed-loop system was established. It utilizes the Mahalanobis distance between the short-term welding audio feature vector and the centroid of the satisfactory-matching class as a feedback signal. A PID controller then adjusts the welding current, driving the short-term audios feature vectors toward the centroid of the satisfactory parameter-matching class until they stabilize within a preset threshold range, thereby achieving adaptive matching with the set arc voltage. Experimental results demonstrate that the system achieves satisfactory parameter matching within seconds, exhibiting a fast response. This demonstrates the preliminary feasibility of audio based adaptive parameter matching for GMAW under controlled conditions.