Comparison of Rotating Stall Warning by Different Methods for Variable Speed Configurations in a Contra-Rotating Compressor

Stall in compressors can cause performance degradation and even lead to disasters.These unacceptable consequences can be avoided by timely monitoring stall inception and taking effective measures.This paper focused on the rotating stall warning in a low-speed axial contra-rotating compressor.Firstly,the stall disturbance characteristics under different speed configurations were analyzed.The results showed that as the speed ratio(RR)increased,the stall disturbance propagation speed based on the rear rotor speed gradually decreased.Subsequently,the standard deviation(SD)method,the cross-correlation(CC)method,and the discrete wavelet transform(DWT)method were employed to obtain the stall initiation moments of three different speed configurations.It was found that the SD and CC methods did not achieve significant stall warning results in all three speed configurations.Besides,the stall initiation moment obtained by the DWT method at RR=1.125 was one period after the stall had fully developed,which was unacceptable.Therefore,a stall warning method was developed in the present work based on the long short-term memory(LSTM)regression model.By applying the LSTM model,the predicted stall initiation moments of three speed configurations were at the 557th,518th,and 333rd revolution,which were44,2,and 74 revolutions ahead of stall onset moments,respectively.Furthermore,in scenarios where a minor disturbance preceded the stall,the stall warning effect of the LSTM was greatly improved in comparison with the aforementioned three methods.In contrast,when the pressure fluctuation before the stall was relatively small,the differences between the stall initiation moments predicted by these four methods were not significant.

Dynamical pattern recognition for univariate time series and its application to an axial compressor

In this paper, a learning and recognition approach is proposed for univariate time series composed of output measurements of general nonlinear dynamical systems. Firstly, a class of dynamical systems in the canonical form is derived to describe the univariate time series by introducing coordinate transformation. An observer-based deterministic learning technique is then adopted to achieve dynamical modeling of the associated transformed systems of the training univariate time series, and the modeling results in the form of radial basis function network (RBFN) models are stored in a pattern library. Subsequently, multiple observer-based dynamical estimators containing the RBFN models in the pattern library are constructed for a test univariate time series, and a recognition decision scheme is proposed by the derived recognition indicator. On this basis, more concise recognition conditions are provided, which is beneficial for verifying the recognition results. Finally, simulation studies on the Rossler system and aero-engine stall warning verify the effectiveness of the proposed approach.