Performance Improvement of a Return Channel in a Multistage Centrifugal Compressor Using Multiobjective Optimization

Multistage centrifugal compressors are widely used in applications in the oil and gas fields where compressors are operated for long periods, and hence their reliability is very important. For its operation to be cost effective, a compressor is required to exhibit high efficiency and a wide operating range. To improve the aerodynamic performance of the centrifugal compressor, many investigations on impellers and diffusers have been conducted [1–5]. Previous investigations suggested that higher efficiency can be achieved by improving the blade loading distribution of the impeller in the case of high and medium flow coefficients. Other papers suggested that wedge-type impellers applied in the case of a low flow-coefficient region also provide higher efficiency. Previous investigations on diffusers reported that half-guide vane-type diffusers provided high efficiency in the case of high flow coefficients. The velocity in the return channel is considerably lower than that in the impeller and diffuser. Therefore, the affect of total pressure loss in the return channel on the overall performance at the design flow rate is relatively small. However, it has been confirmed that the residual swirl flow at the outlet of the return channel leads to insufficient head rise in the next impeller stage [6–8]. Therefore, optimization of the return channel is necessary to minimize the loss in its passage and the residual swirl flow at its outlet. Hildebrandt [9] optimized the return vane and return bend separately by using a multiobjective optimization method to minimize the loss coefficient. Aalburg et al. [10,11] optimized the return channel using the design of experiment (DOE) by decreasing the diffuser outlet-to-inlet radius ratio while maintaining the return channel performance. Aalburg et al. also experimentally confirmed that by using a diffuser with a smaller outlet-to-inlet radius ratio, the efficiency and power could be further decreased. However, in the above studies, the residual swirl flow was not considered in the objective functions used in the optimization. Therefore, in this study a multiobjective optimization based on a genetic algorithm was performed to determine the optimum shape of the return channel to minimize the loss in the passage and the residual swirl flow at the outlet of the return channel. The improved performance of the optimum return channel was experimentally confirmed using a two-stage test compressor.