AI-Based Nondestructive Quality Evaluation of Friction Stir Welded Joints in Aluminum Parts for Electric Vehicles

In the manufacturing of electric vehicles, aluminum is widely used to enhance lightweight properties and heat dissipation in vehicle parts. However, aluminum is prone to issues such as internal pore formation and reduced strength compared to the base material after welding. Therefore, it is crucial to select an appropriate welding process based on the vehicle part's purpose, thickness, geometry and quality requirements. Friction Stir Welding (FSW) has become increasingly popular due to its effectiveness in reducing defects such as porosity, cracks, and thermal distortion. Various Non-Destructive Testing (NDT) methods, such as ultrasonic testing, radiographic testing, infrared thermography, and eddy current testing, have been utilized to evaluate the quality of FSW joints. However, research on real-time quality assessment using artificial intelligence (AI) in practical field applications remains limited. This study applies a high-resolution and high-reliability immersion-based A-scan automatic ultrasonic testing technique to detect micro-defects such as lack of penetration, tunneling, and kissing bonds in FSW joints and to acquire consistent data. Furthermore, using the acquired A-scan signal images, a real-time AI-based quality assessment program was developed to assess the acceptability of the FSW joint on-site.