Safety-constrained transient control for aero-engines: A data-driven diffeomorphic ADP framework.

Area of Science:

• Aerospace Engineering
• Control Theory
• Artificial Intelligence

Background:

• Aero-engine control requires managing complex transient conditions while adhering to strict safety limits.
• Existing methods struggle with wide-range transients and explicit constraint enforcement.
• Developing robust control strategies for safe and efficient aero-engine operation is critical.

Purpose of the Study:

• To develop a novel data-driven adaptive dynamic programming (ADP) framework for safety-constrained aero-engine control.
• To explicitly enforce state and input safety constraints during wide-range transient operations.
• To enhance control performance and reduce computational complexity in aero-engine applications.

Main Methods:

• Utilizing diffeomorphic transformations to eliminate explicit state constraints, reformulating the problem with virtual input saturation.
• Designing an inverse hyperbolic tangent barrier function to handle input constraints and apply Bellman's optimality principle.
• Employing a data-driven policy iteration method to approximate the Hamilton-Jacobi-Bellman equation and derive the optimal control law.

Main Results:

• The proposed ADP framework successfully enforces state and input safety constraints.
• Simulations on a JT9D engine demonstrated safe and rapid operating condition transitions.
• The method achieved superior control performance, reducing acceleration time by 24.6%, compared to PID and PSO-MPC.

Conclusions:

• The developed data-driven diffeomorphic ADP framework offers a feasible and stable solution for safety-constrained aero-engine control.
• The approach significantly improves control performance and efficiency during transient conditions.
• This research presents a practical advancement for modern aero-engine control systems.