Abstract: Aiming at the problem of performance degradation and life shortening caused by long-term use of mechanical systems, a life prediction method is designed by using the particle filter algorithm, which can provide theoretical basis for timely maintenance of mechanical equipment and prolong the life of equipment. Firstly, a multi-model method is used to model the operation process of mechanical equipment, which overcomes the shortcoming of the traditional single model which is difficult to describe the life cycle of mechanical equipment. Secondly, by using the particle filter algorithm and system model switching matrix, the remaining service life of the system can be predicted. Finally, in order to improve the prediction accuracy, a compensation algorithm for predicting deviation is designed to achieve unbiased prediction. Taking the crack growth of high-speed train axle steel as an example, the correctness and effectiveness of the proposed method are verified by simulations.

Key words: Multi-model, particle filter, remaining useful life prediction, prediction deviation compensation

摘要： 针对机械系统由于长期使用所造成的性能下降，寿命缩短的问题，运用粒子滤波算法设计了一种寿命预测方法，可为机械设备进行适时维护，提高运行寿命，提供适当的理论依据。首先，采用多模型方法对机械设备运行过程进行建模，克服了传统单一模型难以描述其生命变化周期的缺陷；其次，借助于粒子滤波算法和系统模型切换矩阵，对系统运行状态进行预测，进而预测系统的运行寿命；最后，为了提高预测精度，设计了一种预测偏差的补偿算法，以期达到无偏预测的目的，并以高速列车车轴钢的裂纹生长为例，对文中方法的正确性和有效性进行了仿真验证。

关键词: 多模型, 粒子滤波, 寿命预测, 偏差校正