Deep Transfer Learning-Enabled Hardness Classification of Bearing Rings Using Pulsed Eddy Current Testing

Hardness is a crucial mechanical property index for materials, and with the increasing demand for product quality, the efficient, cost-effective, and reliable testing of hardness has become essential. However, the complex microstructure transformation during heat treatment has made it challenging to establish a non-contact, rapid, and affordable method for hardness testing. In this article, a novel hardness classification method for bearing rings was proposed. First, signals with rich information are obtained by pulsed eddy current testing, which are then used to achieve end-to-end hardness classification through a deep transfer learning model. Next, a layer-wise fine-tuning scheme is employed to optimize the model’s performance based on the available data. Finally, the loss function of the model is optimized to compact the data in the feature space. The experimental results demonstrate that our proposed method achieves strong robustness, making it a promising candidate for practical applications.