Detection and diagnosis of compressors' instabilities: A mixed model-based and signal-based approach

Issues concerning detection and diagnosis of the surge and rotating stall in a compression system are presented. Due to the importance of identifying the type of instabilities of a compressor for instability recovery, this paper combines a model-based fault identification filter (FIDF) and a signal-based adaptive fuzzy logic system (AFLS) for the detection and recognition of the type of system instabilities. A residual signal is first generated on-line by an FIDF scheme to detect the occurrence of instabilities. The feature of the residual signal regarding the amplitude of a specific frequency is then extracted in real-time and transferred to an AFLS mechanism for the prediction of the type of system instabilities. By properly adjusting the threshold of FIDF for generating the alarm signal, this study may not only act as a reliable precursor but also provide a precise diagnosis of the type of instability at its onset so that a corrective action can be made on-line promptly. Simulation results using Moore and Greitzer's compressor model (1986) are obtained to demonstrate the effectiveness of the proposed scheme.