| Rotating stall is one of the types of stalling condition that may occur in the compressor section
of jet engine and gas turbine. Spike stall and modal stall are conditions of instability that can
cause serious harm to the engine. In this work, spike stall is predicted by studying those
changing dynamics within a blade passage of the compressor section using Autoregressive
(AR) models. In particular, the change in eigenvalues of the system describing the flow
dynamics within a blade passage is considered as a form of a precursor to the stall event. The
order of the stochastic AR model is determined by investigating different Information Criteria
such as the Akaike Information Criteria
(AIC), the Bayesian Information Criteria (BIC), the Kullback Information Criteria (KIC) and
the Conditional Model Estimator (CME). To test the proposed stall precursor prediction, a
number of experiments conducted near stall conditions are utilized. The experiments are
conducted on a one stage low speed axial compressor system. The proposed algorithm is
capable of predicting the onset of spike stall and – compared to current literature – is detecting
the precursor many revolutions earlier.
Key Words: Spike Stalls, Precursor, Autoregressive model, Order, Information criteria |