Automatic defect identification using thermal image analysis for online weld quality monitoring

Abstract

Online weld-monitoring systems are being developed to reduce the cost and delays in detecting defects and rectifying welding parameters. It is known that variations (a) in arc positioning, (b) in heat input and (c) due to the presence of contaminants distinctly manifest as differences in the spatial and temporal surface temperature distributions. In this paper, it is demonstrated that (i) offset positioned thermal imaging of online TIG welding is a feasible non-destructive monitoring technique for detecting porosities in the AA2219 welding, (ii) spatio-temporal temperature distributions close to and in the vicinity of the weld pool can provide statistically distinct features in defect-free and defective weld regions, and (iii) thermal image-based assessment compares very favorably with post-weld radiography assessment for significant defect occurrence. Given the high frame rates and temperature resolution of currently available infrared cameras, it is believed that infrared thermography can be a practical weld-monitoring option capable of providing reliable assessment comparable to more elaborate off-line assessment.