Artificial Intelligence (AI) has emerged as a transformative force in the field of Non-Destructive Testing (NDT) since 2016, with its adoption accelerating across industries after 2019 due to global disruptions such as the COVID-19 pandemic, halted travel and aerospace operations, labor shortages, material cost inflation, and geopolitical conflicts. These challenges have driven an urgent need for smarter, more automated, and labor-efficient inspection solutions. In the NDT domain, Magnetic Particle Inspection (MPI), or Magnetic Particle Testing (MT), remains a widely used technique for detecting surface and near-surface defects in ferromagnetic materials. It is particularly prevalent in the steel, automotive, and aerospace industries—especially where large volumes of safety-critical parts require inspection. Traditionally, MPI relies heavily on manual labor for processes including part loading, magnetization, magnetic powder application, visual inspection, and demagnetization. These labor-intensive procedures not only incur high operational costs but also demand significant training and skilled manpower.
This paper presents the development and implementation of an AI-powered, vision-based automated inspection system tailored for detecting surface defects in automotive cast components, specifically Constant Velocity Joints (CVJs). The system integrates advanced computer vision with magnetic particle inspection equipment, leveraging the capabilities of modern CCD cameras and deep learning algorithms. Key technologies of the system include a One-Stage detection architecture, an Anchor-Free model design, and an End-to-End processing pipeline, enabling high-speed, real-time defect recognition with robust generalization performance. This integrated solution achieves reliable and standardized quality control, significantly reducing reliance on manual inspection and enhancing throughput. Beyond CV Joint (CVJ) inspection, the proposed system offers strong scalability and adaptability, making it suitable for a broad range of cast and forged ferromagnetic components. Its application demonstrates the growing viability of AI-driven automation in NDT processes and its potential to reshape quality assurance in industrial manufacturing environments.