摘要针对烧结钕铁硼在服役期间产生疲劳破坏过程的突发性和不确定性，提出了一种粒子滤波与微磁监测方法相融合的疲劳破坏预测方法。通过将烧结钕铁硼的力磁耦合模型引入粒子滤波理论框架中，设计了基于相对磁导率的粒子滤波状态方程，并采用构建的微磁监测系统来获取烧结钕铁硼疲劳破坏过程中不断变化的磁感应强度信号，建立了基于磁感应强度的粒子滤波观测方程。进一步地，将状态方程和观测方程联立，实现了基于粒子滤波方法的烧结钕铁硼疲劳破坏预测过程。试验结果表明:粒子滤波方法对于烧结钕铁硼疲劳破坏演变过程的预测具有良好的可靠性和准确性，其中在疲劳载荷为120MPa时粒子滤波预测结果的RMSE（Root Mean Square Error，均方根误差）仅为76.69，是支持向量机预测结果的1/4。本研究为脆性磁体的疲劳损伤预测提供了新的思路。
Abstract：A novel method of predicting sintered NdFeB fatigue damage prediction based on particle filter and micro-magnetic monitoring against the urgency and uncertainty during the sintered NdFeB service is put forward. First, in order to accurately reflect the dynamic change relationship between system state variables and output variables, the force-magnetic coupling model of sintered NdFeB is introduced into the particle filtering theoretical framework. Furthermore, the particle filtering evolution equation based on relative permeability is designed. According to the changeable magnetic induction intensity signal during the fatigue damage of sintered NdFeB obtained from the micro-magnetic monitoring system, the particle filtering observation equation based on magnetic induction intensity is constructed. It means that the system status variable was characterized by input variables and output variables. Then, the sintered NdFeB fatigue damage prediction process based on particle filtering algorithm is realized by associating the evolution equation and the observation equation. By monitoring the magnetic induction intensity signal of the stress concentrated region, the characteristic relationship between stress and magnetic induction intensity in sintered NdFeB can be established. Therefore, it can dynamically detect and monitor the fatigue damage of NdFeB by observing the changes of magnetic induction intensity signals. Finally, in order to verify the feasibility and effectiveness of the fatigue failure process of sintered NdFeB by using particle filter method, fatigue experimental study on sintered NdFeB magnets were performed. The experimental results show that the estimation precision and accuracy rating of predicting the sintered NdFeB evolution of fatigue damage is high. Compared with the particle filter prediction and support vector machine, when the fatigue load is 120MPa, the RMSE (Root Mean Square Error) of the particle filter prediction result
is only 76.69. And at this time the result of particle filter prediction is 1/4 of support vector machine. This research provides a new idea for predicting the fatigue failure of brittle sintered magnets.
2022, Vol. 43 
