IR thermography as a non-destructive testing (NDT) tool has its distinct advantage — its non-intrusive and non-contact nature. This makes the IR based NDT especially attractive for the highly automated assembly lines. IR for weld quality inspection has been explored in the past, mostly limited to the offline post-processing manner in a laboratory environment. No online real-time RSW inspection using IR thermography has been reported. Typically for postprocessing inspection, a short-pulse heating via xenon flash lamp light (in a few milliseconds) is applied to the surface of a spot weld. However, applications in the auto industry have been unsuccessful, largely due to a critical drawback that cannot be implemented in the high-volume production line – the prerequisite of painting the weld surface to eliminate surface reflection and other environmental interference. This is due to the low signal-to-noise ratio resulting from the low/unknown surface emissivity and the very small temperature changes (typically on the order of 0.1°C) induced by the flash lamp method.
An integrated approach consisting of innovations in both data analysis algorithms and hardware apparatus that effectively solved the key technical barriers for IR NDT. The system can be used for both real-time (during welding) and post-processing inspections (after welds have been made). First, we developed a special IR thermal image processing method that utilizes the relative IR intensity change, so that the influence of surface reflection and environment interference can be reduced. Second, for the post-processing inspection, a special induction heater is used to replace the flash lamp, resulting in temperature changes on the order of 10°C. As a result, the signal-to-noise ratio increased by several orders of magnitudes with no surface painting needed, and the inspection results are more accurate and reliable. For real-time inspection, the heat from welding (with temperature exceeding 1000°C) was utilized. Third, “thermal signatures” were identified to uniquely correlate to different weld quality attributes through computational modeling of heat transfer and extensive testing of specially designed ranges of welding conditions. Novel IR image analysis algorithms that automatically and intelligently identify the “thermal signatures” from the IR images and positively determine the weld quality in less than a second were developed.