Aircraft dynamic boundary analysis based on PSO-ELM and improved BACTM

Aircraft stall, characterized by abrupt lift reduction due to severe aerodynamic flow separation on the wing surface, represents a safety critical phenomenon that can trigger problems such as stall roll and stall spin, posing significant threats to flight safety. Establishing high-precision dynamic models and conducting stall characteristic analysis are pivotal technical measures for preventing stall risks and enabling effective stall recovery. To address the trade-off between model complexity and accuracy in unsteady aerodynamic force modeling, a hybrid modeling approach integrating Particle Swarm Optimization and Extreme Learning Machine (PSO-ELM) was developed. By fusing multisource wind tunnel test data, this method constructs an unsteady coupled aerodynamic force model with high predictive accuracy, low computational complexity, and robust adaptability to diverse operating conditions, thereby enhancing the reliability of stall characteristic analysis and boundary computation. To overcome the limitations of conventional bifurcation analysis algorithms, including initial-value dependency and fixed-step rigidity that hinder thorough exploration of solution spaces, a modified bifurcation analysis algorithm incorporating “random state-point generation” and “adaptive step-size adjustment” was proposed. Combined with saddle-node manifold theory, this approach enables global analysis of longitudinal nonlinear dynamic characteristics in aircraft. Using a statically unstable aircraft's longitudinal model as a case study, simulations of both open-loop and backstepping control-integrated closed-loop systems were conducted. The focus was on analyzing global longitudinal dynamic behavior and solving stall boundaries, providing dynamic support for mitigating unsteady hysteresis effects, expanding stable flight envelopes, and suppressing unintended stalls.