No abstract

Centrifugal compressors are required to increase their operating range and efficiency, which are limited at low mass flow rates by the rotating stall and surge. This paper presents a surrogate-based multi-objective optimization of a centrifugal compressor to improve its efficiency and stall margin. Curvatures of the blade, the impeller shroud, and the diffuser hub are selected as optimization parameters since they influence highly both the efficiency and the stall limit. The implemented optimization procedure starts by the construction of a metamodel, which is the radial basis function that uses a database composed of a well-selected set of geometries and their corresponding computational fluid dynamics predicted objectives using the Ansys-CFX 12 code. The NSGA-II optimization algorithm is used afterward to search the Pareto front based on radial basis function approximations. To improve the accuracy of the radial basis function and subsequently the Pareto front, a database refinement is sequentially achieved, using the leave-one-out-cross-validation uncertainty to select infill points. The present procedure is tested on the NASA lowspeed centrifugal compressor, showing its ability to increase both the compressor operating range and efficiency. Furthermore, the flow pattern analysis confirms the suppression of separations that lead to instability in the optimized compressor at the stall point of the baseline design.

No abstract

Centrifugal compressors have reached advanced stages of their development, and it is only through a detailed understanding of their complex airflows that improvements in the overall performance will be achieved. This article presents a numerical investigation of unsteady flows through the different components of a radial compressor, emphasizing on the impellervaneless-diffuser-scroll interactions. A transient rotor-stator simulation model was used for the calculations carried out by means of the code CFX-TASCflow, at design point and near stall and choke operating conditions. Spectral analyses of the pressure fluctuation-related different component interactions have resulted in the amplitudes and speeds of rotation of the main energetic modes and have revealed that the flow presents a space-time periodic behaviour that may be described by the double Fourier decomposition. The spatial mode analysis was therefore introduced to describe the interaction mechanisms, whereas the time mode analysis during the working time of the machine has permitted to determine different frequencies, and hence the most prevailing modes and their originating sources. These spectral analyses have permitted a good understanding of the flow interaction mechanisms and revealed the computations limits, depending on whether one-passage or a full-rotor simulation is considered. Principally, these limitations are related to an underestimation of pressure fluctuation amplitudes and a failure in detecting all harmonics lower than the number of passages per component. Finally, this study may help in envisaging the local treatments to reduce noise levels and increase the compressor stability.

No abstract