Nugget and corona bond size measurement through active thermography and transfer learning model

Resistance Spot Welding (RSW) is considered a preferred technique for joining metal parts in various industries, mainly for its efficiency and cost-effectiveness. The mechanical properties of spot welds are pivotal in ensuring structural integrity and overall assembly performance. In this work, the quality attributes of resistance spot welding, such as nugget and corona bond sizes, are assessed by analyzing the thermal behavior of the joint using a physical information neural network (PINN). The investigation starts by examining the phase and amplitude gradient of the thermal signal. A convolutional neural network (CNN) is employed to estimate the size of the nuggets and corona bonds. The CNN architecture is based on the Inception V3 architecture, a state-of-the-art neural network that excels in image recognition tasks. This study presents a suggestion to adopt a new methodology for automatic RSW quality control based on thermal signal analysis.