Aero-Engine Transient Acceleration Control Using DDPG
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for aero-engine transient state acceleration control are plagued by complex design processes, poor robustness, and limited operating ranges, failing to effectively optimize acceleration laws under various constraints.
Innovation Solution
A reinforcement learning-based acceleration control method for aero-engines, utilizing an Actor-Critic network and deep deterministic policy gradient algorithm to optimize fuel flow and rotor speed control, addressing high-dimensional state spaces and continuous action outputs, and incorporating noise for exploration and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dynamic programming method is used for optimization control, then the control accuracy is improved, but the calculation complexity and computational load increase significantly
Solution Approach 1:
The patent replaces traditional mechanical optimization methods (dynamic programming, nonlinear programming) with a neural network-based intelligent control system. The neural network learns optimal control policies through reinforcement learning, substituting complex iterative calculations with direct policy extraction from trained networks, thereby reducing computational burden while maintaining control accuracy.
Solution Approach 2:
The patent transforms the control problem from solving complex optimization equations to adjusting neural network parameters through reinforcement learning. By changing the approach from direct mathematical optimization to parameter-based learning, the system achieves comparable accuracy with significantly reduced computational complexity during operation.
2Productivity
If power extraction method is used to simplify the model, then the calculation speed is improved, but the robustness deteriorates due to ignoring volume effect and dynamic coupling
Solution Approach 1:
The patent enables the control system to automatically adapt to the full-engine dynamic characteristics through reinforcement learning. The neural network learns the complex coupling effects and volume effects directly from the engine model during training, allowing the simplified control structure to maintain robustness while achieving fast calculation speed during operation.
3Measurement precision
If traditional nonlinear programming methods are used, then the theoretical optimality is improved, but the design process complexity and implementation difficulty increase
Solution Approach 1:
The patent replaces complex nonlinear programming implementations with neural network-based reinforcement learning. The system achieves theoretical optimality through learned policies while simplifying the design process to standard deep learning workflows, making the implementation more accessible and easier to execute.
4Reliability
If reinforcement learning with Actor-Critic network is used, then the robustness and adaptability are improved, but the training time and computational resources increase
Solution Approach 1:
The patent performs the computationally intensive optimization work during the offline training phase using Actor-Critic reinforcement learning. Once trained, the neural network policies can be deployed for real-time control with minimal computational requirements, effectively moving the time consumption from operational phase to training phase.
Data Source
AI summary
The present invention provides an optimization control method for an aero-engine transient state based on reinforcement learning, and belongs to the technical field of aero-engine transient states. The method comprises: adjusting an existing twin-spool turbo-fan engine model as a model for invoking a reinforcement learning algorithm; to simultaneously satisfy high level state space and continuous action output of a real-time model, designing an Actor-Critic network model; designing a deep deterministic policy gradient (DDPG) algorithm based on an Actor-Critic frame, to simultaneously solve the problems of high-dimensional state space and continuous action output; training the model after combining the Actor-Critic frame with the DDPG algorithm; and obtaining the control law of engine acceleration transition from the above training process, and using the method to control an engine acceleration process.