基于支持向量数据-贝叶斯模型的冷水机组故障检测与诊断研究

Investigation on Fault Detection and Diagnosis of Chiller Based on Support Vector Data Description and Diagnostic Bayesian Network Model

本文采用了结合支持向量数据描述方法和诊断贝叶斯网络的故障检测与诊断模型,以模拟退火法优化支持向量描述方法的参数训练故障检测模型,以故障数据获取诊断贝叶斯网络结构与各节点的条件概率表构建故障诊断模型。针对螺杆式冷水机组制冷剂充注量不足、蒸发器水流量下降、冷凝器水流量下降及制冷剂水供水温度传感器偏差、冷却水出水温度传感器偏差故障,提出了支持向量数据-贝叶斯模型对5种故障检测与诊断能力,分析影响模型检测与诊断能力的因素。结果表明:SVDD-DBN模型对故障诊断效果受无故障预测模型准确度影响,且在置信度为99.7%时表现较好;对症状越独立的故障,模型诊断效果越好,制冷剂充注量不足、冷却水出水温度传感器偏差故障诊断率可达100%。

A fault detection and diagnosis model based on support vector data description method and diagnosis Bayesian network is adopted.The parameters of support vector description method are optimized by simulated annealing method to train the fault detection model.The structure of diagnosis Bayesian network and conditional probability table of each node are obtained from the fault data to construct the fault diagnosis model.For insufficient refrigerant charge fault,water flow of evaporator decreases fault,condenser water flow decreases fault,refrigerant water supply temperature sensor deviation fault,cooling water outlet temperature sensor deviation fault of screw chiller,the detection and diagnosis ability of the proposed support vector data description and diagnostic Bayesian network(SVDD-DBN)model is researched and the factors that affect the ability are analyzed.The results show that the proposed SVDD-DBN model has a good performance when the confidence is 99.7%;the more independent the symptom is,the better the effect of model diagnosis is.The diagnosis rates of insufficient refrigerant charge fault and cooling water outlet temperature sensor deviation reach 100%.