基于多源信号融合的往复式压缩机气阀健康评估  

Reciprocating compressor valve health assessment based on multi-source data fusion

作　　者：尧阳烽 余永华[1,2] 王康 聂方 胡嘉[1] 徐德峰 YAO Yangfeng;YU Yonghua;WANG Kang;NIE Fang;HU Jia;XU Defeng(School of Naval Architecture,Ocean and Energy Power Engineering,Wuhan University of Technology,Wuhan 430063,China;Key Laboratory of Marine Power Engineering and Technology of Ministry of Transport,Wuhan 430063,China;China Aerodynamics Research and Development Center,Mianyang 621000,China)

机构地区：[1]武汉理工大学船海与能源动力工程学院,湖北武汉430063 [2]船舶动力工程技术交通运输行业重点实验室,湖北武汉430063 [3]中国空气动力研究与发展中心,四川绵阳621000

出　　处：《机电工程》2024年第11期2003-2011,共9页Journal of Mechanical & Electrical Engineering

基　　金：先进船舶发动机技术全国重点实验室开放基金资助项目(LAB-2023-02)。

摘　　要：压缩机气阀测点处信号是多个激励源的综合响应,因此,仅依靠单一信号难以准确评估压缩机的健康状态。针对这一问题,提出了一种基于多源信号融合的往复式压缩机气阀健康状态评估方法。首先,以某型四级高压往复式压缩机为研究对象,通过故障模拟实验,获取了其进排气阀不同健康状态的热工参数和声发射信号;然后,提取了不同信号源的时域特征和频域特征及热工参数,以气阀健康状态下各特征参数的均值作为健康基准,计算了气阀不同健康状态样本与健康基准的马氏距离(MD),基于多源信号融合理论,将不同信号计算所得的马氏距离相融合,进行了样本重构;最后,基于决策树构建了压缩机气阀健康状态评估模型,评估了气阀的健康状态。研究结果表明:各单一信号源的评估准确率分别为70.6%、87.2%和85.2%,而基于多源信号融合重构后,样本不同健康状态的区分度显著提高。基于决策树构建的气阀健康状态评估模型可以有效识别气阀的健康状态,识别的准确率可达100%,具有良好的健康状态评估效果。Aiming at the problem that the signal at the gas valve measuring point is the comprehensive response of multiple excitation sources,it is difficult to accurately evaluate the health state of the reciprocating compressor gas valve by only relying on a single signal.Therefore,a reciprocating compressor gas valve health evaluation method based on multi-source signal fusion was proposed.Firstly,taking a type of four-stage high-pressure reciprocating compressor as the research object,the thermal parameters and acoustic emission signals in different health states of the inlet gas valve and exhaust gas valve were obtained through fault simulation test.Then,the time-domain characteristics,frequeny-domain characteristics and thermal parameters of different signal sources were extracted,and the mean value of each characteristic parameter under the gas valve health state was used as the health benchmark to calculate the Mahalanobis distance(MD)between the gas valve samples with different health states and the health baseline.Based on the multi-source signal fusion theory,the Mahalanobis distance obtained from the multi-source signal was fused to reconstruct the sample.Finally,the gas valve health state evaluation model was constructed based on the decision tree to evaluate the health state of the gas valve.The research results show that the evaluation accuracy of each single signal source is respectively 70.6%,87.2%and 85.2%,while the differentiation of different health states of samples reconstructed based on multi-source signal fusion is significantly improved.The gas valve health status evaluation model built based on decision tree can effectively identify the health status of the gas valve with an accuracy of 100%,and has better evaluation effect.

关 键 词：活塞式压缩机 气阀特征参数提取 马氏距离 健康状态评估方法 决策树 数据采集系统 故障模拟实验 样本重构 

分 类 号：TH457[机械工程—机械制造及自动化]