基于改进主元分析DDPCA的滚动轴承过渡模态早期故障检测方法

Incipient Fault Detection Method of Rolling Bearing Based on DDPCA

目的 提出一种深度差分主元分析方法用于滚动轴承早期故障检测,解决滚动轴承在运行过程中长期处于变转速等多模态工况,故障特征难以提取和划分的问题。方法 结合差分算法和深度分解原理的分段PCA故障检测方法,使用差分方法对原始数据进行处理,通过K-means聚类方法将具有相似变量特征的过渡模态数据划分成为相同过渡子模态;结合深度分解理论对每个过渡子模态建立故障检测模型,并通过机械故障综合模拟实验台收集的数据验证模型准确性。结果 随着分解阶数的增加,对过渡模态早期故障检测效果逐渐提升,对滚动轴承过渡子模态的划分越来越清晰,误报的情况也随着分解阶数的增加而逐渐减少;滚动轴承持续减速状态下外圈故障一阶分解检测的漏检率为17.2%,二阶分解检测的漏检率为8.6%,三阶分解检测的漏检率为6.6%。结论 笔者所提方法对过渡子模态进行多层分解,可以准确提取过渡子模态中的故障特征并建立分段检测模型,提高了过渡模态的滚动轴承早期故障检测的准确性。

This paper proposed a deep difference principal component analysis(DDPCA)method for early fault detection of rolling bearings in extracting and dividing fault features of multimodal conditions.The parameters were adjusted according to the demand,which lead to the rolling bearing being in variable speed and multi-modal working conditions.In view of the dynamic process characteristics of rolling bearings in transition mode,the fault characteristics were difficult to extract and classify,the failure detection of rolling bearings in incipient stage cannot be carried out using the uniform detection model This method uses differential method to process original data,classifies the transition mode data with similar variable characteristics into the same transition submodes by K-means clustering method,establishes fault detection model for each transition submode in combination with depth decomposition theory.The layer rate of the outer ring fault detection was 17.2%,8.6%and 6.6%.The multi-layer decomposition of the transition submodes extracts the fault characteristics of the transition submodes accurately,the model is established to improve accuracy of the incipient failure detection of the rolling bearing in the transition mode.