Enhanced chiller sensor fault detection, diagnosis and estimation using wavelet analysis and principal component analysis methods

An enhanced sensor fault detection, diagnosis and estimation (FDD&E) strategy is developed for centrifugal chillers using wavelet analysis method and principal component analysis (PCA) method. A number of sensors of concern in chiller system monitoring and control are assigned into a PCA model, which can group these correlated variables and capture the systematic variations of chillers. Raw measurements or simple processing measurements of sensors may deteriorate the performance of sensor FDD&E strategy using PCA because of the embodied noises and dynamics. Wavelet analysis can extract the approximations of sensor measurements by separating noises and dynamics. Using these approximation coefficients for PCA modeling we can improve the capability and reliability of fault detection and diagnosis as well as the accuracy of sensor fault estimation. This wavelet–PCA-based sensor FDD&E strategy was validated using field operation data of an existing centrifugal chiller plant while various sensor faults of different magnitudes were introduced. The results demonstrate that this strategy can produce better performance of sensor FDD&E in terms of fault detection ratio, diagnosis ratio and estimation accuracy by comparing with conventional PCA-based sensor FDD&E strategy using raw or simple processing measurements for PCA modeling.