针对噪声环境下难以提取齿轮故障特征、诊断准确率低的问题，提出了一种将基于综合评价指标的变分模态分解(variational mode decomposition，VMD)参数寻优、核主成分分析(kernel principal component analysis，KPCA)特征融合和BP网络相结合的齿轮故障诊断方法。首先，为了有效评价VMD分解后的各固有模态函数(intrinsic mode function，IMF)分量及避免变分模态分解需要人为设定相关参数的问题，设计了一种基于包络熵与峭度的综合评价指标，用于建立VMD参数寻优中的适应度函数及筛选最优IMF分量;其次，按最优参数进行VMD分解后对最优IMF分量提取多域特征集，再利用KPCA模型对其进行特征的融合;最后，通过BP网络模型进行故障诊断。实验表明，与其他传统方法相比，在相同实验条件下该方法提高了齿轮故障的识别率，准确率高达98％，证明了该方法的有效性。
Fault diagnosis method of gear based on parameter optimization VMD-KPCA and BP network
In view of the problems of difficulty in extracting gear fault features and low diagnosis accu-racy under noisy environment,a gear fault diagnosis method was proposed,which combined VMD pa-rameter optimization based on comprehensive evaluation indicators,KPCA feature fusion and BP net-work.Firstly,in order to effectively evaluate the IMF components after VMD decomposition and avoid the problem of manually setting relevant parameters for VMD,a comprehensive evaluation index based on envelope entropy and kurtosis was designed to establish a fitness function for VMD parameter opti-mization and screen the optimal IMF components.Secondly,after performing VMD decomposition ac-cording to the optimal parameters,a multi domain feature set was extracted from the optimal IMF com-ponent,and then the KPCA model was used to fuse its features.Finally,fault diagnosis was performed using the BP network model.The experiment shows that under the same experimental conditions,com-pared with other traditional methods,this method improves the recognition rate of gear faults,with an accuracy of up to 98％,proving the effectiveness of this method.
万方数据
维普
