基于数据驱动的风洞风机故障诊断研究

Fault Diagnosis of Wind Tunnel Fans Based on Data-Driven

中文关键词:  风机故障  特征提取  故障诊断  随机森林

英文关键词:fan failure  feature extraction  fault diagnosis  RF

中文摘要:

      风洞风机是风洞动力系统的核心部件，是提供风洞试验条件必须的动力源。目前在国内外低速风洞设备中，主要使用大功率电机驱动风扇，其故障集中发生在电机和风扇机构上，一旦发生故障会造成大量的停机时间，严重影响试验工作。基于风机运行采集的大量数据，将故障预测与健康管理（PHM）技术应用于风洞风机故障诊断研究，重点探讨风机故障诊断系统构建的关键步骤，以期达到提前诊断风机故障的目的。数据分析使用主成分分析（PCA）法提取风机多维故障特征，选用随机森林（RF）算法诊断风机故障状态，并探讨了两种适合算法的硬件搭建方法。通过试验验证，RF算法具有很好的故障分类能力，平均分类准确率达94%、召回率95%、查准率94%，可精确区分风洞风机故障状态。

英文摘要:

      The wind tunnel fan is the core component of the wind tunnel power system and a necessary power source for providing wind tunnel test conditions.At present,in the low-speed wind tunnel equipment at home and abroad,the fans are mainly driven by high-power motors,and their failures are concentrated in the motors and fan mechanisms,which will cause a lot of downtime once the failures occur and seriously affect the test work.Based on the large amount of data collected from fan operation,the prognostics and health management (PHM) technology to wind tunnel fan fault diagnosis research,focusing on the key steps of the fan fault diagnosis system construction,in order to achieve the purpose of early diagnosis of the fan faults in advance.In data analysis,principal component analysis(PCA) is used to extract multidimensional fault characteristics of fans,and the random forest(RF) algorithm is selected to diagnose the fault status of fans,and two hardware construction methods suitable for the algorithms are discussed.Through experimental verification,the RF algorithm has good fault classification capability,with an average classification accuracy of 94%,a recall rate of 95%,and a precision rate of 94%,which can accurately distinguish the fault status of wind tunnel fans.