Neural-network Based Analysis and Prediction of a 4-Stage Transonic Compressor's Characteristic Performance Map

Difficulties, due to lack of information about stage-by-stage transonic axial-compressor performance, used in f different aero vehicles and power plants are analyzed. To overcome these issues, a three-layer back-propagation neural-network of Levenberg-Marquardt algorithm is presented and discussed. Available experimental data are used for the neural-network training. Experimental data are limited to some special speeds; other data can be f predicted by neural network method and this method can be used for development of an off-design conditions or overall dynamic simulation of aero vehicles or gas-turbine power-plants.