概率神经网络在往复压缩机多故障诊断技术研究中的应用

The Application of Probabilistic Neural Network in The Reciprocating Compressor Multiple Fault Diagnosis Technology Research

往复压缩机结构复杂,故障种类较多,导致诊断难度较大。针对往复压缩机多故障诊断难的问题,提出了一种基于概率神经网络的往复压缩机多故障诊断分类诊断方法,该方法提出5种无量纲因子作为往复压缩机多种故障特征数值,将其融合形成特征矩阵,输入概率神经网络进行往复压缩机多故障诊断分类。通过工程案例分析,该方法在往复压缩机多故障诊断分类方面的准确性有了显著提高并大大缩短了诊断分类时间,为往复压缩机多故障诊断分类提供了一种快捷有效的手段。

Because of the complex structure and a variety of fault of reciprocating compressors,it is difficult to diagnose reciprocating compressor failure. On account of multi-fault diagnosis of reciprocating compressors,it proposes a multi-fault diagnosis classification method which is based on probabilistic neural network. Five kinds dimensionless factor as reciprocating compressors feature values of multiple failures are proposed by the method,which fuse to form the feature matrix,which are the input of probabilistic neural network to carry out faulty diagnosis classification about reciprocating compressors. The case analysis shows that the accuracy of the method in a multi-fault diagnosis of reciprocating compressors has been significantly improved and the time of method has been greatly reduced,which provides a quick and effective means for reciprocating compressors multiple fault diagnostic classification.