基于KPCA和SVM的滚珠丝杠副润滑失效故障诊断

Fault Diagnosis of the Lubrication of Ball Screws Based on KPCA and SVM

为了能够更加充分地表征振动信号在时域、频域、时频域等多特征参数与滚珠丝杠副润滑失效状态的线性关系,提高滚珠丝杠副润滑失效故障诊断的准确率,提出了基于KPCA和SVM的滚珠丝杠副润滑失效故障诊断模型。提取了滚珠丝杠副在3种润滑状态和5种转速下的振动信号特征,构建了润滑特征混合域特征集,并按照累计贡献率大于90%的标准对特征集进行核主元分析,将筛选出的能够反映滚珠丝杠副润滑失效故障特性的主元作为主要特征量,最后将筛选出的主要特征量输入支持向量机(SVM)内进行训练测试,并在同一转速的条件下作对比试验。结果显示:利用核主元分析法进行主元提取能够有效降低原始数据的维度,并提高模型诊断的准确率;当电机转速为100 rpm时,所建立的基于KPCA和SVM的滚珠丝杠副润滑失效故障诊断模型的诊断准确率高于其他转速下的诊断准确率,准确率为93.33%;这说明该方法在对滚珠丝杠副润滑状态进行诊断时,通过控制电机转速可有效提高诊断的准确率。

To build the linear relationship between the lubrication failure state and the vibration signals in time domain,frequency domain,time frequency domain et al,and then improve the accuracy of the lubrication failure diagnosis,this paper proposed a new diagnosis model of the lubrication failure for ball screws based on KPCA and SVM.We obtained the vibration signal characteristics of the ball screw pair under 3 lubrication states and 5 speeds,and the lubrication feature mixed domain feature set is constructed,We analyzed the extracted parameters by the KPCA method according the criterion of the cumulative contribution rate higher than 90%,based on which the component that can reflects the lubrication failure characteristics of ball screws is selected as the main characteristic quantity.Finally,the obtained characteristic quantity is trained and tested in SVM,and a series of contrast test is made at the same speed.The results show that:KPCA can effectively reduce the dimension of original data and improve the accuracy of model diagnosis;When the motor speed is 100 rpm,the diagnostic accuracy of the model based on KPCA and SVM is 93.33%higher than that of other speeds;This shows that the method can effectively improve the accuracy of diagnosis by controlling the motor speed when diagnosing the lubrication state of ball screws.