基于主元分析的冷水机组传感器故障识别

PCA-based Fault Detection,Reconstruction and Identification for Sensors in the Chiller

测量数据的真实性和准确性是冷水机组安全运行和优化节能的必要条件。长期使用条件下,传感器故障极易发生而且很难识别。在基于主元分析的传感器故障研究中,以Q统计量为检测指标的常规故障检测、诊断与数据重构在故障源的识别上存在一定误判。采用基于数据重构的枚举甄别方法,分析了传感器故障检测、数据重构及故障识别的算法流程,并以实际工程数据进行验证。结果表明,问题传感器的重构数据的故障识别指标变化最明显,可以准确鉴别故障传感器。

The real and exact measurement data are the essential condition for the safety operation and optimal conservation for chillers. Unfortunately, the sensor faults are occurred easily and identified hardly due to the long term operation period. There are some misjudgments in the traditional method of fault detection, diagnosis and reconstruction by Q-statistics based on Principal Component Analysis. A reconstruction-based fault identification method by Q-statistics is presented and the flowchart for the sensor Fault detection, reconstruction and identification is demonstrated in this paper. Results show that since the fault identification index of reconstructed data is changed hugely, the faulty sensor is isolated easily when a sited data set was employed to validate the presented method.