Thrust optimization control for gas turbine engine afterburner state via model-based deduction learning

Optimization control of gas turbine engines, based on a baseline control schedule, serves as a highly effective means to enhance engine thrust. However, conventional methods directly apply the optimal nozzle control schedule from the intermediate state to the afterburner state, resulting in the drawback that thrust gains are not maximized across different augmentation ratios. To address this, the paper proposes a model-based deduction learning (MDL) method for optimizing afterburner thrust control. The method comprises the following steps: 1) a model-based deductin optimization module that generates a comprehensive dataset of optimal control schedule based on component-level model deduction; 2) an offline experience learning module that refines and prunes the dataset, subsequently using a multilayer perceptron network to learn and derive the optimal control schedule; 3) a smooth switching module that integrates the optimal control schedule with the baseline schedule for online implementation. Simulation results demonstrate that, compared with conventional methods, the MDL method increases the average thrust gain over the full flight envelope from 1.62 % to 2.78 % at 50 % afterburner, and from 0.99 % to 1.81 % at maximum afterburner. The MDL method achieves an average computation time of 4.3 ms on a microcontroller with a clock frequency of 132 MHz. And the method shows excellent adaptability when confronted with individual engine differences and performance degradation.