基于概率神经网络算法的永磁同步直线电机局部退磁故障诊断研究

Partial Demagnetization Fault Diagnosis Research of Permanent Magnet Synchronous Motors Based on the PNN Algorithm

针对永磁同步直线电机(permanent magnet synchronous linear motor,PMSLM)的局部退磁故障问题,引入一种基于空间气隙磁密重构特征提取与概率神经网络(probabilistic neural network,PNN)算法相结合的局部退磁故障分类识别方法。采用等效磁化强度法分析永磁体在局部退磁情况下,PMSLM气隙磁密在不同空间位置的分布特性;利用有限元法定量计算PMSLM空间气隙中心线、气隙中心线上方、气隙中心线下方3个位置处的气隙磁密强度,将其融合为唯一识别退磁故障类型的特征量,并进行多种局部退磁故障类型下的仿真分析,建立了丰富的退磁故障样本库;建立神经网络和径向基网络,并对PNN网格结构进行优化,利用PNN分类算法实现局部退磁故障的精确分类识别,并进行分类器精度校验仿真实验。样机实验结果表明,所提方法能够准确辨识PMSLM局部退磁故障的组合类型,识别率高达到99.4%。

A partial demagnetization fault classification and identification method based on the spatial air gap flux density reconstruction and probabilistic neural network(PNN) algorithm was introduced to solve the partial demagnetization faults of permanent magnet synchronous linear motors(PMSLM). An equivalent magnetization method was used to analyze the distribution characteristics of PMSLM air gap flux density in different spatial locations under the condition of local demagnetization of permanent magnets. Finite element method was used to quantitatively calculate the air gap flux density on, above and below the air gap centerline, and they were fused into a special fault feature quantity to uniquely identify the types of demagnetization, and the simulation analysis under various types of demagnetization faults were carried out to build rich demagnetization sample database. Finally, the neural network and the radial basis network were established to make the grid structure optimized, and the PNN algorithm was used to realize the accurate classification and identification of local demagnetization faults, and the accuracy of the classifier was simulated. Prototype experiment results show that the proposed approach can classify PMSLM local demagnetization fault types, and the recognition rate is of 99.4%.