The centrifugal compressors are used in a wide variety of turbo machines ranging from low-pressure fans for cooling of electric motors to high-pressure ratio gas turbine compressors, from tiny cryogenic coolers to large industrial petrochemical compressor stations. An automated design scheme making use of Navier Stokes equation and Artificial Neural Network is envisaged as in figure 1. The present work describes a user friendly scheme to carry out blade geometry generation and to carry out preliminary designs that are two of the most important steps of the automated design scheme. Present paper describes a robust method to carry out these two steps. This paper also describes a method adopted to decide on viable design space. In the blade generation steps, present method was applied to that obtained by other methods reported in literature. It was found that the present scheme could replicate the centrifugal compressor geometry with just six parameters as compared to many more by others. In Mihai[1], use of adjoint method is used for optimization of NASA rotor. The design had 28 parameters. In Trigg[2], a systematic approach for optimization of a two dimensional blade design is reported. Two dimensional profile is defined by seventeen parameters. In Wang Hong liang[3] an optimization based on particle swarm principle is made use of to optimize the blade angle distribution at hub and shroud and had seven control points to do this. In order to minimize the number of control points, Asimara and Goto[4], used inverse method. Here blade loading at hub and shroud is the input to a inverse method to generate the blade geometry. Optimizer works on the blade loading. The generated geometry is analyzed by CFD calculations. From the above references, it is clear that the present scheme is a user friendly general method.
展开▼
