Self-decision of sensor fault detection indicator for chiller based on principal component analysis

Accurate and timely monitoring of key system parameters is crucial for ensuring the safe and energy-efficient operation of refrigeration and air-conditioning equipment. However, sensor failures often occur due to equipment aging, environmental changes, and other factors. Fault detection methods based on operational data, such as principal component analysis (PCA), have shown promising adaptability. Yet, the inconsistency in sensor types and measurement ranges can lead to varying sensitivity and fault detection efficiency when using a single index, such as theQ-statistic. This study adopts a blind fault detection index to evaluate the performance of different fault detection indices on the same training dataset and sensor, enabling the selection of the most suitable index for improved fault detection sensitivity and recognition. The results demonstrate that utilizing the fault detection index with higher capability can enhance the sensors' fault detection sensitivity and recognition, thereby optimizing the energy-saving potential of the system.