Failure analysis and life extension of mechanical products under multiple service conditions

Abstract

Abstract During the service of engineering machinery products, the interaction of mechanical load, service environment and other conditions leads to the diversity and uncertainty of failure modes. Mechanism analysis is difficult to accurately calculate the dynamic changes of performance. Most existing machine learning methods only focus on the remaining useful life, without considering the internal mechanism and connection of multiple failure modes. In this paper, a resistance degradation model is established based on the data of multiple service cycles. According to the service data under different working conditions, a dynamic resistance degradation model based on classification and regression tree (CART) and a product service life enhancement strategy based on working condition sequence optimization are proposed. Firstly, the characteristics of different working conditions are analysed. Secondly, the data of multiple sensors are fused, and CART is used to fit the degradation model. Finally, a product service life optimization problem is designed to increase the life. Experiments show that the proposed method can fit the resistance degradation process, and can significantly improve the life of mechanical products.