Grinding burns generated during the machining of bearing rings severely degrade their service performance and lifespan. Therefore, accurate detection and identification of grinding burns are crucial for quality assessment and safety assurance of bearing rings. Traditional burn evaluation methods suffer from limitations such as being destructive, low efficiency, and poor real-time capability. Although current mainstream non-destructive testing techniques like the multi-parameter magnetic non-destructive testing method (3MA) offer advantages such as non-invasiveness, stability, and high efficiency, the unclear transmission mechanism from grinding burns to magnetic responses hinders their quantitative detection accuracy. This study takes G95Cr18 bearing steel as the research object and investigates the micro- to meso-scale mechanisms influencing its magnetic properties due to grinding burns. By integrating intelligent recognition methods based on deep learning, the limitations of conventional detection techniques can be effectively overcome, significantly improving the accuracy and efficiency of grinding burn identification. Furthermore, the multi-dimensional characterization and data analysis methods mentioned can be further integrated with deep learning technology. This integration not only provides a feasible technical pathway for quantitative grinding burn detection based on electromagnetic non-destructive testing but also addresses specific challenges in current deep learning-based burn identification methods. It promotes the advancement towards intelligent detection and classification, offering valuable insights for innovation in quantitative evaluation technology for grinding burns.

Bearing rings, Grinding burns, Multi-parameter magnetic non-destructive testing, Deep learning