Real-Time Instability Detection of Centrifugal Compressors Based on Motor Speed Measurements

Flow instability in the centrifugal compressor should be detected and avoided for stable and safe operation. Due to the popularity of electric centrifugal compressors, instability detection could be achieved by measuring motor signals instead of traditional aerodynamic signals. In this paper, the feasibility of instability detection by motor signals (i.e. rotating speed and phase current) was studied experimentally. The physical structure and control method of the electric centrifugal compressor were discussed to reveal the potential of instability detection by motor signals. Dynamic pressure signals and motor signals measured during unsteady experiments were analyzed in the time domain and frequency domain. Characteristics of these signals were then compared under different operating conditions to indicate the feasibility of instability detection by motor signals. Finally, the ability of Short-Time Fourier Transform (STFT) of rotating speed signals in real-time instability detection was discussed. Results showed that the rotating speed signal is a good alternate for instability detection in spite of signal distortion, while the phase current signal can only detect surge due to the low resolution of the controller. Based on the variations of the amplitude and frequency of rotating speed signals, the real-time instability can be captured accurately by STFT with a window size of 0.5 s. Besides, the interference caused by the controller can be removed by STFT.