Multi-Objective Optimization Method for Performance Prediction Loss Model of Centrifugal Compressors

The selection of loss models has a significant effect on the one-dimensional mean streamline analysis for obtaining the performance of centrifugal compressors.In this study,a set of optimized loss models is proposed based on the classical loss models suggested by Aungier,Coppage,and Jansen.The proportions and variation laws of losses predicted by the three sets of models are discussed on the NASA Low-Speed-Centrifugal-Compressor(LSCC) under the mass flow of 22 kg/s to 36 kg/s.The results indicate that the weights of Skin friction loss,Diffusion loss,Disk friction loss,Clearance loss,Blade loading loss,Recirculation loss,and Vaneless diffuser loss are greater than 10%,which is dominant for performance prediction.Therefore,these losses are considered in the composition of new loss models.In addition,the multi-objective optimization method with the Genetic Algorithm(GA) is applied to the correction of loss coefficients to obtain the final optimization loss models.Compared with the experimental data,the maximum relative error of adiabatic the three classical models is 7.22%,while the maximum relative error calculated by optimized loss models is 1.22%,which is reduced by 6%.Similarly,compared with the original model,the maximum relative error of the total pressure ratio is also reduced.As a result,the present optimized models provide more reliable performance prediction in both tendency and accuracy than the classical loss models.