A Hybrid Physical Damage Neural Network for Wear Prediction of Self-Lubricating Bearings

Abstract Self-lubricating bearings are widely used in aerospace, marine, and other fields due to their excellent performance. Accurate wear prediction for self-lubricating bearings is crucial for ensuring reliability and safety. However, achieving both physical interpretability and high accuracy in predictive models remains a challenge, as these bearings typically operate under varying load conditions and in high-noise environments. In this article, a hybrid physical damage neural network is proposed for wear prediction. First, a “physics neuron operator” based on the Archard wear model is designed and embedded into the network to directly compute wear depth. Second, a cumulative damage law is introduced into this operator to quantify the degradation path of the bearing during operation. Finally, the evolution law of wear stages is encoded as a physical constraint in the loss function to compel the network's learning process to follow the true degradation mechanism. To validate the model, a dedicated test platform was built, and a full life cycle degradation dataset for self-lubricating bearings was collected. Experimental results show that the proposed model significantly outperforms existing methods in prediction accuracy. Furthermore, this article provides an in-depth analysis of the model's physical interpretability, revealing its internal working mechanism and significantly enhancing its credibility and generalization ability.