After the emergence of turbo-machines in the industrial field, researchers have always been seeking to increase their efficiency. Centrifugal compressors are widely used in industries, hence understanding their flow and attempting to increase their efficiency has always been a focus. In recent years, using two splitter blades alongside the main blades has been one of the novel methods used to improve the performance of centrifugal compressors. In this study, after validating the NASA-CC3 centrifugal compressor, a centrifugal compressor design with two splitter blades was carried out. Then, by defining eight variables, an attempt was made to optimize the design. The genetic algorithm was used as the optimization algorithm in this process. Since the optimization process only using genetic algorithms is very time-consuming and has high computational costs, artificial neural networks were used to reduce costs. The objective function in this optimization process was to increase efficiency while maintaining the flow rate and pressure ratio at the design point. After the optimization process was completed, the performance curve of the optimized compressor was compared with the NASA-CC3 compressor at and outside the design point, and its advantages and disadvantages were stated. At the end of the above optimization process, the efficiency of the centrifugal compressor increased by 1.06% at the design point. To investigate the effect of the number of computational networks on the independence of results, further analysis is required.
Read full abstract