Adaptive Visual Sensing Deviation Detection and Real-Time Tracking Control of Swing-Arc Narrow-Gap Weld Based on Variation Coefficient Recognition.

Weld tracking aims to provide real-time compensation for weld deviations caused by groove assembly inaccuracies and thermal shrinkage during Gas Metal Arc Welding (GMAW). To achieve precise tracking control in swing-arc narrow-gap GMAW based on passive visual sensing, a Self-Adaptive Coefficient-of-Variation Recognition (SCVR) algorithm is proposed for adaptively detecting weld deviation by filtering out welding interference. SCVR adaptively constructs a data window by discriminating the original variation coefficient to acquire the raw data distribution of the groove centerline. It then designs an in situ bandpass data filter to locally search the data segment with the minimal coefficient of variation for adaptive bandwidth determination. By applying the filter to the raw data, the disturbed data are removed, in situ retaining the data with the globally minimized variation coefficient. Finally, SCVR recognizes the real groove center from the filtered data, accurately detecting a weld deviation by comparing this center to the torch position. Additionally, an SCVR-based real-time tracking control system incorporating a PLC-based actuator with a PI controller for optimal stability is developed to correct the torch position in real time, achieving a high tracking precision of -0.161~+0.126 mm. Experimental results demonstrate the robust adaptability and effectiveness of the SCVR-based weld detection and tracking control system.