摘要
往复泵作为石油石化行业重要的输送设备,通过振动监测手段来保证系统的安全稳定运行具有重要的现实意义。如何对往复泵的非平稳和非线性信号提取特征并进行准确识别是诊断中的关键问题。针对往复泵故障特征的提取,提出了一种利用ICEEMDAN-GRNN神经网络相结合的诊断方法。首先利用ICEEMDAN对采集的原始信号进行分解得到若干个IMF分量,然后计算IMF分量的奇异谱熵并构造特征向量,再将特征向量输入到GRNN神经网络进行训练和模式识别。研究表明:该方法可以有效提取往复泵的故障特征并进行准确的模式识别。
The reciprocating pump plays an important role in the petrochemical industry procedure,which is crucial in ensuring the systematic safety and stability through vibration monitoring method.How to extract non-stationary and nonlinear signals features of reciprocating pump and accurate recognition is the key problem in diagnosis.Aiming at the extraction of reciprocating mechanical fault features,ICEEMDAN-GRNN neural network algorithm are used to recognize the valve failure modes.The IMF components are obtained by decomposing the collected signals with the ICEEMDAN algorithm in calculating the singular spectral entropy and constructing the feature vector.The feature vector is input into the GRNN neural network for training and mode recognition.The study shows that the signal processing method can effectively extract the feature values of reciprocating compressor's valve faults,while the GRNN neural network algorithm could be used in following pattern recognition process with higher accuracy.
作者
别锋锋
都腾飞
庞明军
谷晟
BIE Feng-feng;DU Teng-fei;PANG Ming-jun;GU Sheng(School of Mechanical Engineering,Changzhou University,Jiangsu Changzhou213016,China)
出处
《机械设计与制造》
北大核心
2021年第3期127-131,共5页
Machinery Design & Manufacture
基金
国家自然科学基金(51376026)。
作者简介
别锋锋,(1979-),男,湖北仙桃人,博士后,副教授,主要研究方向:机械设备状态监测与故障诊断,振动信号处理与分析。
