Experimental investigation and multivariate recurrence plot analysis of flow instabilities in a centrifugal compressor

Abstract

Flow instabilities such as stall and surge pose significant challenges to the safe and efficient operation of centrifugal compressors. A comprehensive understanding of these unstable phenomena is essential for improving compressor performance. To this end, an experimental investigation is conducted to study the propagation of flow instabilities and surge characteristics in a high-speed centrifugal compressor. High-response pressure transducers are installed along the streamwise passage to capture transient pressure signals. In addition to conventional time-domain analysis, multivariate recurrence plots are employed for the first time to explore the nonlinear characteristics during the transition from pre-surge to deep surge conditions. The results show that mild and deep surge occur successively at the surge point, and the vaneless region exhibits the most pronounced pressure fluctuations. Multivariate recurrence plots, including joint and cross recurrence plots, successfully reveal synchronization and simultaneous recurrence patterns between circumferential pressure signals. As the compressor approaches deep surge, the recurrence structures exhibit a clear shift from uniform, sparse patterns to highly non-uniform distributions. By identifying specific structures in the plots, recurrence-based metrics can quantitatively capture dynamical variations associated with flow instabilities. Moreover, compared with the average diagonal line length and joint recurrence rate, the entropy exhibits higher sensitivity in distinguishing between stable and unstable states. These findings demonstrate the potential of recurrence-based techniques to detect and identify the evolution of flow instabilities in centrifugal compressors.