Performance Optimization Method of Turbocharged Solid Propellant Ramjet(TSPR)

摘要

Turbocharged Solid Propellant Ramjet (TSPR) is considered as one of the ideal propulsion system for tactical missiles in the atmosphere. It is the combination of Solid Rocket Ramjet (SRR) and Solid Propellant Air Turbo Rocket (SP-ATR).So, TSPR includes both advantages,and can achieve a high-performance application in a wide envelope. However, this new concept, which include the intake, turbocharging systems, gas generators, combustion chambers,etc,also increases the system complexity. In these circumstances,it making inlet, turbocharging system, gas parameters,chamber structure and other factors may affect its performance.Under this condition, it is difficult to find a quickly and accurately ways to improve the combustion performance through the empirical parameters method,and is not give full play to the motor performance. For this reason, the Haupt function,which is highly non-linear and multi-local extremum function, is used to verify the accuracy of the Polynomial Response Surface, Kriging Function and Artificial Neural Network(ANN). Among them, the Radial Basis Function ( RBF) neural network is not only meet the accuracy requirements, but also does well in describing the typical characteristics of the Haupt function. According to this results, this paper proposes an alternative model based on RBF. Then, the feasibility of the three optimization algorithms including Multi-Island Genetic Algorithm (MIGA), Simulated Annealing (SA) and Pointer Automatic Optimizer (Pointer) is also verified by Haupt function. Research has shown that: 1) the MIGA algorithm does not obtain the optimal solution within the set maximum iterations, and the biggest errors of the other two algorithms is similar, about 3%;2) Pointer's iterative number is 9540,which is about sixteen times that of the SA. Therefore, ASA algorithm is very appropriate to used in TSPR's performance optimization.Next, we was conducted the combustion performance optimization by combining the Optimized Hypercube Des