Affiliation:
1. Mechanical Engineering Department, University of Natal, Durban, South Africa
Abstract
The accuracy of the streamline-curvature method has hitherto not been evaluated for curved-axis convergent-divergent flow as found in transonic turbine cascades and for flow in cascades with flow divergence or flare. To provide the data for evaluation, two simple single-channel rectangular-section wind tunnels were built to simulate the flow between turbine blades. The supersonic channel was successfully and economically powered by a large steam ejector. In transonic and supersonic flow, an examination of the sonic and M = 1·1 lines was used as a basis for evaluating the predictions. The correlation was found to be good near the convex surface, but showed a deterioration towards the concave surface. Both test and prediction revealed the unexpected fact that the sonic line contour is not smooth, but wavy. The correlation at the M = 1·1 line was positionally the same as the sonic line, but the experimental contour was considerably more wavy. The accuracy deteriorated further into the supersonic region. The discrepancies in the transonic region are thought to be caused by a very slow rate of change of area, which produces a complex boundary layer interaction in the throat region, affecting both the location of the throat and the area ratios in that zone. The prediction method was shown to be extremely accurate for flared-cascade flow even up to the very exacting situation of a 40 per cent blade height increase, the only error being that due to boundary layers in the apparatus that effect the velocity distribution over the latter half of the suction surface in a highly staggered cascade.