Design And Implementation Of Low Thrust Turbofan Engine Control Algorithm Based On Neural Network

Small turbofan engines have significant research value in the international aviation field due to their potential in reducing size,weight,and improving fuel efficiency,providing power support for unmanned aerial vehicles,micro aircraft,and high-altitude platforms.Moreover,the study of engine control technologies is crucial for achieving precise control,optimizing combustion performance,and reducing emissions of small turbofan engines,thereby promoting environmental protection,sustainable development,and technological innovation in the aviation industry.Therefore,this article focuses on the research of component-level modeling and neural network control algorithms for small turbofan engines.Firstly,using aerothermodynamics principles and component-based methodology,a nonlinear SIMULINK model is established.This method divides the engine into multiple sub-components,each described by characteristic curves formed by predicted data.Aerothermodynamic characteristics of each component are mathematically abstracted by replacing them with corresponding aerothermodynamics equations.To ensure the engine components can perform thermodynamic calculations normally within the entire envelope,common working equations under different working conditions need to be satisfied and solved using the Newton-Raphson method.This provides a model basis for subsequent neural network control.Secondly,the BP neural network is introduced into the turbofan engine power turbine PID controller,and the Particle Swarm Optimization(PSO)algorithm is used to further improve the shortcomings of the BP neural network,forming a BP neural network single-loop PID control method.Control simulation results under various operating conditions show that the proposed control strategy can meet the requirements of the control law in both steady-state and load-increasing or load-decreasing conditions.However,some parameter control curves still have minor fluctuations and overshoot,and the smoothness of the curves needs to be enhanced.Lastly,a cascade PID controller based on RBF neural network is designed and optimized through a genetic algorithm.The L2 regularization method is used in the genetic algorithm to optimize its smoothness,and the traditional genetic algorithm’s crossover and mutation rates selection methods are improved to optimize the RBF parameters.The GA-RBF neural network is introduced for parameter self-tuning in the cascade PID control structure established for both high and low-pressure turbines.Simulation results show that the GA-RBF is optimized in terms of smoothness and overshoot improvement,but its control performance in terms of maximum error is inferior to the PSO-BP neural network control.Both neural network control algorithms can meet the control law requirements under different working conditions and have their own advantages and disadvantages in controlling different parameter curves.They can be chosen according to specific needs in practical work.