Influence of Low-Solidity Cascade Diffuser on Spike Stall Inception in a Centrifugal Compressor

Abstract

Stall inception patterns for three low-solidity cascade diffusers were experimentally investigated to extend the operating range of centrifugal compressor. Pressure fluctuations on the wall and flow distributions at the diffuser inlet were measured, and static pressure-rises across the diffuser vane were also measured to clarify the mechanism of rotating stall inception. In the case of the original diffuser (LSD_O), which had the radius ratio between the impeller outlet and the diffuser vane inlet of 1.10 and a flat plate wing section, a short length-scale stall cell known as a spike first appeared in the positive slope of the diffuser static pressure-raise (after the flat slope was maintained), and then a surge occurred as flow-rate decreased. The propagation speed of short length-scale stall cell was about 13% of the impeller rotation speed. The measured flow distribution at the inlet of the LSD_O suggested that the flow separation occurred in the vaneless space and this separation influenced the rotating stall. In the case of the LSD_A, which had the radius ratio of 1.10 and the NACA 63 wing section, a short length-scale stall cell that was smaller than that of LSD_O was appeared in the positive slope of the diffuser static pressure-raise (without the flat slope). A surge also occurred as further flow-rate decreased. The propagation speed of short length-scale stall cell was about 13% just prior to surge point. In contrast, in the case of LSD_B, which had the radius ratio of 1.05 and the NACA 63 wing section, a small disturbance appeared without spike just prior to surge point and rotating stall did not occur. The small disturbance did not propagate in the circumferential direction. Moreover, the flow separation did not occur in the semi-vaneless space. At the stall inception point of LSD_O and LSD_A, the averaged absolute flow angle at the diffuser inlet agreed with the onset flow angel of rotating stall in vaneless diffuser predicted by Senoo. Therefore, it was considered that the flow separation also occurred in the vaneless space and this separation influenced the rotating stall in the case of LSD_A. Moreover, it was considered that the difference of two stall cell patterns between LSD_O and LSD_A depended on the static pressure-rise characteristic of diffuser determined by the design specifications in case of that the rotating stall occurred in the vaneless space. In contrast, at the point of the small disturbance appeared of LSD_B, the averaged absolute flow angle at the diffuser inlet is smaller than that predicted by Senoo. Therefore, in case of small the radius ratio, the reverse flow in vaneless diffuser is suppressed, and the rotating stall is also suppressed.