Optimal Design of a Centrifugal Compressor using CFD, Multiobjective Optimization and Response Surface Methodology


CHALLENGE -  CFD based optimization of turbomachinery requiring high accuracy is computationally time consuming and is incompatible with industrial applications. In order to perform optimization without losing accuracy variable reduction and response surface metamodels can be used.

SOLUTION - FRIENDSHIP-framework and ANSYS Workbench have been used to parameterize the geometry and perform the CFD simulation. modeFRONTIER has been used to perform statistical and correlation analysis and multi-objective optimization. After running a DOE, results have been analyzed, enabling to reduce the number of input variables from 21 to 13. In the second step has been used the FAST algorithm to enable a multi-objective optimization . This algorithm efficiently combines real and virtual optimization to enhance convergence speed, find more Pareto designs and save computational time.









BENEFITS - The use of modeFRONTIER and the FAST algorithm enabled to perform a multi-objective optimization considering both impeller and volute of a centrifugal compressor. Total-to-total isentropic efficiency has been improved of 1.84 % and total-to-total pressure ratio of 5.34 %, considering a commercial compressor as baseline.