基于集成学习的转子部件脱落故障诊断方法  

Research on Fault Diagnosis Technology of Rotor Part Falling-off Based on Spark

作　　者：周晓[1] 马圣杰 ZHOU Xiao;MA Shengjie(School of Mechanical and Electronic Engineering,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学机电工程学院,湖北武汉430070

出　　处：《数字制造科学》2022年第1期16-22,共7页

摘　　要：针对常规信号处理方法在多传感器数据综合利用中的不足,通过时频域和全息谱等技术进行故障特征提取,构造用于故障诊断的转子部件脱落故障诊断数据集。针对单模型在转子脱落故障诊断中识别准确率不高的问题,提出了一种集成学习模型,并为了处理海量监控数据,基于Spark框架进行并行化设计。实验证明该方法是一种可行的转子脱落故障诊断方法,相对于单模型能提高故障诊断的准确度,且能通过并行化实现良好的加速效果。To cope with the shortcomings of conventional signal processing methods in the comprehensive utilization of multi-sensor data,fault feature extraction is enabled using time-frequency domain and holographic spectrum.Accordingly,a set of fault diagnosis data for indicating the falling off of the rotor component has been constructed.An integrated learning model which can accurately recognize the rotor shedding has been brought up.To process massive monitoring data,this model has been designed in parallel based on the Spark framework.Experimental results verify the feasibility of the proposed fault diagnosis method for rotor shedding based on an integrated learning model.In comparison with the fault diagnosis based on single model,it can improve the accuracy of the fault diagnosis,thus achieving good accelerating property through parallelization.

关 键 词：集成学习 转子脱落故障 全息谱 Spark计算框架 

分 类 号：TH17[机械工程—机械制造及自动化] TP181[自动化与计算机技术—控制理论与控制工程]