Turbine Preliminary Design Using Artificial Intelligence and Numerical Optimization Techniques

Abstract

This paper describes a new software approach to the preliminary design of aircraft engine turbines. A hybrid artificial intelligence and numerical-optimization-based design shell called Engineous was used to capture some basic turbine preliminary design knowledge, manipulate turbine design parameters, execute a turbine performance prediction program and its preprocessors, and analyze results. Engineous automatically supplements incomplete human design knowledge with symbolic and numerical search techniques when needed. This approach produced designs with higher predicted performance gains than the existing manual design process in a tenth of the turnaround time and has yielded new insights into turbine design. A comparison of turbine designs obtained by designers and by Engineous is presented here along with an overview of Engineous system architecture.