Performance Prediction Methods of Cascade Blades in Steam Turbines

Abstract

A better understanding of the mechanism of loss generation will aid steam path design and may lead to more reliable efficiency prediction when using a loss model. Loss estimation systems are still beneficial in the early stages of steam path design. A number of loss models for estimating turbine performance have been published over the past few decades. However, there have been surprisingly large deviations among them. They do not reflect the aerodynamic improvements towards modern blade design such as spline design and do not recognize the detailed refinements of the blade passage shape such as front loading and aft loading. In addition, it is not easy to set up new or modified methods in the loss models because it is necessary to have large quantity of test data. To establish a new loss estimation system, a method is presented for a turbine cascade blade operating in subsonic regions where the exit Mach number <= 0.8. A prediction method based on entropy generation is developed to analyze the loss mechanism for the cascade blade. In the profile loss model, the basic profile loss model is induced from isentropic Mach number distribution along the blade surface. The trailing edge thickness loss model was introduced from CFD results and experimental data. In the secondary loss model, a correlation model for different geometries (aspect ratio, deflection angle and contraction ratio) is developed from validated CFD methods compared to the measurements. In addition, inlet loss and downstream loss are derived using the concept of boundary layer and entropy generation respectively. To acquire a Reynolds number correction curve, the performance of blade profiles is examined with three linear cascade blades which represent hub, mid-span and tip sections using numerical analysis. The Reynolds number is varied from 10,000 to 10,000,000 to capture the operating regime for typical steam turbines. Twelve different levels of surface roughness on the same profile are calculated using the roughness model in ANSYS-CFX. The ratios of surface roughness to chord length are in the range of k_sc=4.2 ×10^(-5) ~ 8.3 ×10^(-4). The CFD result is compared to Reynolds number correction curves from published literature (AMDC-KO, Aungier, Craig & Cox, Traupel, Sanders and Denton) and available measurement data. Based on these comparisons, a Reynolds number correction curve is properly selected and newly correlated for the estimation of steam turbine performance. In addition, CFD results show that the proposed correlation for the Reynolds number, including roughness effects, can be adopted in both profile loss and secondary loss. The effect of an off-design incidence angle on the aerodynamic profile loss is examined experimentally with linear cascade blades which represent the hub, mid-span and tip sections. The cascade test covers a range of incidence angles from -20 deg to +20 deg. The results are evaluated with two existing kinds of incidence loss models (multiplier methods Ainley & Mathieson, Chen, Zehner / adder methods Stepanov, MK, MKT, BSM) and available measurement data (Zehner, Aronov, Hodson and Yamamoto). Based on these comparisons, it was found that the existing incidence model give quite good estimates in the mid-span profiles but there are some differences in the tip profile with a high inlet metal angle and the hub profile with a low inlet metal angle due to the different inlet geometry shape and a lack of data when developing the published models. To correlate the available test data, a new correlation method is proposed as a function of incidence, inlet metal angle, contraction ratio, leading edge diameter and inlet wedge angle. The results show that it is significantly more successful than the existing correlation methods. The linear cascade test data which represent hub, mid-span and tip sections are used to validate this loss model. The results show a good correlation with the cascade test data.