Steel grades classification and heating defects detection of bearing steel bars via pulse eddy current testing

Abstract Nondestructive examination of bearing steel raw material is a crucial step in ensuring the production of bearing rings. Traditional nondestructive examination methods struggle with real-time sorting, leading to potential production risks. This study proposes a pulse eddy current testing (PECT) approach combined multidimensional features and classification algorithm, to realize the recognition of steel grades and heating defects of bearing steel bars. Firstly, the two-probes differential PECT system was established. Subsequently, pulse signals were collected, and 27 features in the time domain, and frequency domain were extracted and analyzed. Finally, two classification algorithms, BP neural networking and Rindom Forest, are used to classify bearing steel. Experimental results demonstrate the approach proposed can sort bearing steel, die steel, and high-speed steel bars. Furthermore, it is capable of distinguishing between various types (brand, heat number) of bearing steel bars and identifying their overheated or burnt conditions.