多变量统计分析方法在制冷故障检测与诊断中的应用研究

摘要

研究了多变量统计分析方法在制冷装置故障检测和诊断中的应用.对ASHRAE资助下的一组实验数据进行预处理,对其进行故障检测和故障诊断.对于故障检测,可利用平方预测误差(Q统计量)等统计控制变量来判断系统是否在正常运行状态.而对于故障诊断,第一次尝试采用各变量对于平方预测误差的负荷结合各变量的变化率来得到其对于平方预测误差异变的贡献率.从而快速利用变量的变化方向和程度判断故障类型.从结果分析,这种方法可以在众多变量中过滤掉不显著的变化,迅速找到故障主因.利用多变量统计分析方法可以实现对制冷装置的运行状态进行实时监控和诊断.%According to the characteristics of gradual faults in refrigeration systems, a fault detection model based on Multivariate Statistical Analysis was presented. An experimental study was introduced, which was conducted on a 90-ton centrifugal chiller to produce a database that will be used in the development and evaluation of Fault Detection and Diagnostic (FDD) methods applied to chillers. The presented FDD model was validated and evaluated by the experiments. The results show that The PCA-FDD model can meet the detection performance requirement. Moreover, the model was applied to determine the type of faults for the first time, which achieved satisfactory results.