The blended-wing-body aircraft exhibits complex and highly nonlinear flight dynamics, especially in post-stall conditions. It poses significant challenges for stability analysis and control system design. Traditional aerodynamic models, which are often based on static or quasi-steady assumptions, fail to capture the critical unsteady effects, such as hysteresis, that govern these flight regimes. This paper proposes a new data-driven framework to overcome these limitations. First, we generate a high-fidelity dataset that captures unsteady aerodynamics. We achieve this using a control-based virtual flight test method that maintains stable periodic motion within the aircraft’s unstable flight envelope. Next, we construct a high-fidelity unsteady aerodynamic model from these data. This model is based on a recurrent neural network with a real-time recursive learning algorithm and an extended Kalman filter. We then integrate this enhanced model into the aircraft’s nonlinear equations of motion. Bifurcation analysis performed on the improved model accurately predicts stall boundaries and instability characteristics, showing strong agreement with experimental bifurcation results. The analysis reveals that the primary mechanism for longitudinal instability is the complex hysteresis of the pitching moment, which causes a critical change in the sign of the aerodynamic stability derivatives. This data-driven framework provides a more precise and robust method for analyzing post-stall dynamics. It also offers vital insights for designing flight control systems for advanced aircraft configurations.
Fu et al. (2026) studied this question.