变工况下基于加权理论的气阀泄漏诊断方法研究

往复式压缩机是石油化工等行业中不可或缺的关键设备，而压缩机结构复杂、故障率高，其中气阀故障是其主要的故障形式之一。同时，负荷调节工况与气阀故障工况的相互耦合使得示功图变化规律更加复杂，增加了故障诊断的难度。为此，探究了变负荷及气阀故障工况下示功图几何特征的变化规律，提出一种针对变负荷与气阀故障耦合工况下的气阀故障诊断方法。该方法利用反向传播（BP）神经网络进行特征分类，首先分别依据示功图几何特征（包括面积、形心和形心主惯性矩等）以及灰度矩阵统计特征得到压缩机的负荷，再进一步结合故障特征判断气阀的故障类型。为提高诊断结果的准确度，将加权证据融合理论应用于故障分类过程，最终获得精准的气阀故障评估结果。基于实验台数据，对不同泄漏率的气阀故障进行实验验证，负荷预测的准确率为97.5%，气阀泄漏故障识别的准确率为96.1%。

A valve leakage diagnosis method based on weighted theory under variable working conditions

A reciprocating compressor is a critical component in the petrochemical and other industries. The compressor has a complex structure and a high failure rate, with valve failure being one of the main failure types. Furthermore, the coupling of load regulation conditions and valve fault conditions makes the change rule of the indicator diagram more complicated, and increases the difficulty of fault diagnosis. In this work, the change rule of the indicator diagram geometry feature under variable load and valve fault conditions was explored. An air valve fault diagnosis method was then proposed under coupling conditions of variable load and valve fault conditions. The method uses a back-propagation (BP) neural network to classify features. The compressor load is first obtained according to the geometric characteristics of the indicator diagram (including area, centroid and centroid moment of inertia) and the statistical characteristics of the gray matrix, and then the fault type of the air valve is judged based on the fault characteristics. In order to improve the accuracy of the diagnosis results, the weighted evidence fusion theory was applied to the fault classification process, and accurate evaluation results of valve faults were finally obtained. The air valve faults with different leakage rates have been experimentally verified based on the test-bed data. The accuracy of load forecasting is 97.5%, and the air valve leakage fault identification accuracy is 96.1%.