This study is a collaboration between Siemens Industrial Turbomachinery(SIT) and Royal Institute of Technology(KTH). It is aimed to study and compare the outputs of two different computational approaches in axial gas turbine design procedure with the data obtained from experimental work on a test turbine.
The main focus during this research is to extend the available test databank and to further understand and investigate the turbine stage efficiency, mass flow parameters and reaction degree under different working conditions. Meanwhile the concept and effect of different loss mechanisms and models will be briefly studied.
The experimental part was performed at Heat and Power Technology department on a single stage test turbine in its full admission mode. Three different pressure ratios were tested.
For the medium pressure ratio a constant temperature anemometry (CTA) method was deployed in two cases, with and without turbulence grid, to determine the effect of free-stream turbulence intensity on the investigated parameters. During the test campaign the raw gathered data was processed with online tools and also they served as boundary condition for the computational codes later.
The computational scope includes a one-dimensional design approach known as mean-line calculation and also a two-dimensional method known as throughflow calculation. An in-house SIT software, CATO, generated the stage geometry (vane, blade and the channel) and then two other internal computational codes, MAC1 and BETA2, were employed for the one-dimensional and two-dimensional computations respectively.
It was observed that to obtain more accurate mass flow predictions a certain level of channel blockage should be implemented to represent the boundary layer development and secondary flow which is typically around 2%.
The codes are also equipped with two options to predict the friction loss: One is a more empirical correlation named as the Old approach in SIT manuals and the other works based on allocation of boundary layer transition point, named as BL in the present thesis. Simulations were done by use of both approaches and it turned out that the latter works more accurately if it is provided with appropriate transition point and blockage estimation.
The measured data also suggests the idea that the transition point of the vane and blade is not affected by a change in turbulence intensity at least up to 6% in the tested Reynolds numbers, . Amongst different solutions the one which used BL approach and constant transition point (while the turbulence intensity changed) managed to predict this behavior.
Also it was investigated and revealed that the codes inherently predict poor results in off-design loadings which is mainly due to positive incidence angle in addition to high spanwise gradient of the flow parameters.
Author: Mikaillian, Navid