Spike Stall Precursor Detection in a Single-Stage Axial Compressor: A Data-Driven Dynamic Modeling Approach

Operational safety and fuel efficiency are critical, yet often conflicting, objectives in modern civil get engine designs. Optimal efficiency operating conditions are typically close to unsafe regions, such as compressor stalls, which can cause severe engine damage. Consequently, engines are generally operated below peak efficiency to maintain a sufficient stall margin. Reducing this margin through active control requires stall precursor detection and mitigation mechanisms. While several algorithms have shown promising results in predicting modal stalls, predicting spike stalls remains a challenge due to their rapid onset, leaving little time for corrective actions. This study addresses this gap by proposing a method to identify spike stall precursors based on the changing dynamics within a compressor blade passage. An autoregressive time series model is utilized to capture these dynamics and its changes are related to the flow condition within the blade passage. The autoregressive model is adaptively extracted from measured pressure data from a one-stage axial compressor test stand. The corresponding eigenvalues of the model are monitored by utilizing an outlier detection mechanism that uses pressure reading statistics. Outliers are proposed to be associated with spike stall precursors. The model order, which defines the number of relevant eigenvalues, is determined using three information criteria: the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), and the Conditional Model Estimator (CME). For prediction, an outlier detection algorithm based on the Generalized Extreme Studentized Deviate (GESD) Test is introduced. The proposed method is experimentally validated on a single-stage low-speed axial compressor. Results demonstrate consistent stall precursor detection, with future application for timely control interventions to prevent spike stall inception.