主元分析用于多联式空调系统传感器故障检测和诊断

Sensor Fault Detection and Diagnosis for Variable Refrigerant Flow Air Conditioning System based on Principal Component Analysis

作为多元数据分析方法之一,主元分析(PCA)被广泛运用于诊断制冷空调系统的传感器故障。本文首先结合热平衡原理以及多联机运行的控制逻辑,筛选系统中常用的18个传感器变量,建立多联机(VRF)传感器的故障分析(FDD)模型。然后结合主元分析的算法原理,给出以Q统计量和Q贡献率为检验标准的传感器故障检测与诊断流程。利用实测数据验证工作,引入不同类型和程度的传感器故障,分析得到不同故障条件下的故障检测和诊断特性。结果表明:总体上,主元分析应用于多联机传感器故障检测与诊断过程是可靠的。具体特征表现为:不同类型的传感器在不同故障类型及程度条件下,故障检测效果差异明显;在小偏差故障条件下,基于主元分析的传感器故障检测方法的故障检测效率较低,并且针对个别传感器而言,其整体故障检测效率偏低。鉴于故障诊断是基于故障检测的结果,因此上述故障检测方法在FDD过程中将起到重要的作用。

As one of the multivariate data analysis methods, principal component analysis （PCA） is widely used for sensor fault diagnosis in refrigeration and air conditioning systems. First, the 18 sensors commonly used in a variable refrigerant flow （VRF） system are selected to establish sensor fault detection and diagnosis （FDD） models according to the thermal equilibrium principles and control logics of the system. Then, the process of sensor FDD is presented with the Q statistic and Q contribution as test standards, combined with the principles of a PCA algorithm. Next, validation is conducted using the measured data after introducing sensor faults of different types and degrees. Finally, the characteristics of sensor FDD are obtained under different fault conditions. As a whole, the results prove the reliability of applying a PCA to the sensor FDD process for VRF systems. Specific performance characteristics are as follows： fault detection efficiency has big differences for different sensors under different types and extents of faults ; the fault detection efficiency of the PCA-based sensor fault detection method under the conditions with small deviation faults is low; and for individual sensors, the fault detection efficiency is integrally low. Since fault diagnosis is based on fault detection, the above-mentioned fault detection method may play important role in the FDD process.