Preliminary Design of Axial Compressors Using Artificial Intelligence and Numerical Optimization Techniques

Abstract

This paper discusses an automated method for optimizing multistage axial compressor designs using a generic design software shell called Engineous®, coupled with an aerodynamic compressor design program called CUS5 (Compressor Unification Study 5). Engineous combines heuristic searching and numerical optimization techniques, and expert systems to explore the design space of turbomachinery components by iterating preliminary design analysis programs written in languages such as FORTRAN and C. The application design and development process and the lessons learned are discussed for the Engineous/CUS5 axial compressor application. This particular application calculates and maintains compressor stall prediction values while optimizing at the operating (design) point, and requires a design vs. stall comparison during each iteration. A trial compressor design problem was investigated using the Engineous application, and the preliminary results of the study are included. Also introduced is a method for varying flowpath walls using shape criteria rather than radii points.