Abstract: To address insufficient model accuracy, delayed unstart warning, and stringent operating boundary constraints in scramjet engines, a control method integrating model correction and adaptive protection is proposed. First, an offline-online collaborative correction mechanism is constructed to address the insufficient accuracy of the geometrically reduced-order model. Offline correction compen-sates for the systematic bias of the nominal model using a forward-mapping method, while online correction employs a dual extend-ed Kalman filter to identify key parameters in real time and compensate for their time-varying characteristics. Second, to overcome the delay inherent in conventional single-channel warning methods, a combined dual-channel warning method is proposed. The model channel continuously estimates risk, while the sensor channel provides real-time physical verification; a switching mechanism is used to achieve risk perception. Furthermore, to resolve the conflict between engine unstart risk and performance optimization, a protection control strategy is designed. Uncertainty quantification information is incorporated into model predictive control decisions, and adaptive scheduling is used to balance performance and safety. Finally, simulation results show that the proposed method im-proves estimation accuracy and warning lead time while optimizing performance under unstart safety constraints, thereby validating the synergistic effectiveness of the correction-warning-control architecture.

Key words: scramjet engine, quasi-one-dimensional model, model correction, parameter identification, unstart warning, protection control