Wavelet-Based FDI (Fault Detection And Identification) In Oil Refinery

摘要

Fault detection is currently done by experienced operators. However, for a variety of reasons, some faults go undetected or are mis-identified. There are two obstacles to be overcome, when we develop automatic fault-handling tools: First, the poor effectiveness of current fault detection algorithms. Second, the difficulty in articulating operators' experience through well formulated linguistic expressions. To address these concerns, we have developed a wavelet-based procedure for FDI in our oilrefinery, we characterize each fault and normal states out of power spectrums obtained through wavelet transform, and define a "distance" which represents the difference between the two states.