Monitoring system and anomaly detection in dissimilar friction stir lap welding based on MEMS sensor, power consumption, and temperature

Abstract This study presents a multi-sensor monitoring strategy for dissimilar friction stir lap welding (FSW), combining tri-axial MEMS-based acceleration measurements with motor power consumption and thermocouple data. The approach aims to detect and characterize process instabilities and defect formation, including porosity, tunnel voids, and excessive flash, without requiring destructive testing. Experimental campaigns were conducted on AA 2024-T3 and AA 7075-T6 aluminum alloys using six welding conditions, intentionally generating both sound and defective joints. The results reveal distinct power and acceleration trends across the plunging, dwell, and welding phases, which are directly linked to the nature of the defects. The analysis revealed that stable welds were associated with motor power near 1200 ± 8 W, RMS acceleration below 0.10 m/s 2 , and vibration frequencies around 150–220 Hz. In contrast, defective joints exhibited power deviations up to ± 250 W, RMS acceleration as high as 0.24 m/s 2 , and dominant frequencies shifting toward 300–400 Hz, often correlating with tool wear and porosity formation. These results demonstrate the effectiveness of the proposed method in enabling early detection of welding anomalies through sensor-based signal analysis, contributing to process optimization and enhanced weld quality. This approach has significant implications for industrial applications, offering a scalable and reliable framework for sustainable manufacturing.