基于决策树算法的多联机气液分离器插反故障诊断

Opposite-insertion Fault Diagnosis for Gas-liquid Separator in Variable Refrigerant Flow (VRF) System based on Decision Tree Algorithm

本文将决策树算法应用于多联机气分插反故障诊断中,搭建了多联机实验平台采集数据,根据专家知识及数据变化模型验证选取了建模的特征变量,采用决策树C5.0算法构建气分插反故障诊断模型,进一步对由模型分类规则生成的最优变量即过冷器的EEV(电子膨胀阀)进行深入分析和验证。结果表明:将决策树算法应用于多联机气分插反故障诊断的方法,准确率为96%,此诊断方法能满足多联机故障诊断实际运用的需要,并可及时处理多联机发生气分插反故障时,系统过热度降低,从而保证多联机系统的制冷效果和能效比,可通过增大过冷器EEV开度调节。

This paper presents a data-mining-based method with a decision tree algorithm to diagnose gas-liquid separator opposite-insertion faults for a variable refrigerant flow （VRF） system. First, the VRF experimental platform was established for data collection. Then, expert knowledge and verification methods for the data variation model were used to select appropriate model variables. The C5.0 decision tree algorithm was employed to develop a fault diagnosis model. Finally, the effect of the electronic expansion valve （EEV） of the subcooler was analyzed and validated; this effect was selected as the best variable on the basis of classification rules generated by the model. The results show that the fault diagnosis method based on the decision tree exhibits desirable effectiveness for diagnosing gas-liquid separator opposite-insertion faults, with which the fault diagnosis accuracy is up to 96%. Moreover, the proposed technique can meet the requirements for online application of fault diagnosis for VRF systems. This method incorporating a decision tree algorithm to diagnose gas-liquid separator opposite-insertion faults for a VRF system exhibits very high accuracy and reliability; therefore, the method can meet the actual demands of fault diagnosis for VRF systems. Because the occurrence of gas-liquid separator opposite-insertion faults corresponds to a reduction in the degree of superheating, increasing the opening of subcooler EEV can ensure the cooling effect and a good energy efficiency ratio.