Aircraft engine preliminary design utilizing a cascade optimization strategy with neural network and regression approximations

Preliminary design optimization of aircraft engine has been attempted utilizing a cascade optimization strategy along with neural network and regression approximation concepts. The design optimization required a coupling of the engine analyzer or NASA Engine performance program (NEPP) to the CometBoards testbed. Replacement of PowelPs algorithm in the NEPP code by the cascade strategy of CometBoards improved the optimization segment of the engine cycle code. Convergence difficulties encountered during engine balancing were alleviated through a use of neural network and regression approximators. This paper illustrates the results and insights gained from the improved version of the NEPP code considering two examples: a supersonic mixedflow turbofan engine and a subsonic wave-rotor topped subsonic engine. The performance of the regression and neural network methods were found to be satisfactory for the analysis and operation optimization of airbreathing propulsion engines. Both linear regression and neural networks performed at about the same level. 'Engineer, AIAA Associate Fellow Engineer, AIAA Senior Member ^Mathematician, AIAA Member Copyright @2000 by the American Institute of Aeronautics and Astronautics Inc. No copyright is asserted in the United States under Title 17, US Code CometBoards