Gas turbine blades operate in very high temperatures to achieve a high thermal efficiency of the engine. For this reason, the blades have to be cooled to prevent degradation or even melting. The blades can be cooled using various techniques, both by cooling the inside of the blade with cooling channels, and by protecting the outside of the blade from the hot environment. One way to cool the blades from the inside is with rib turbulated channels. Straight square channels lined with 90◦, 45◦ and V-shaped ribs in a staggered configuration are investigated in this thesis. Computational fluid dynamics (CFD), among other methods, can be used to predict important parameters such as heat transfer and pressure loss for different ribbed channel geometries. In this thesis a CFD model using RANS simulations with the turbulence models Lag Elliptic Blending k − ε, Realizable k − ε two-layer and SST k − ω is established and validated against experimental data by Taslim et al [1]. This is done by comparing the Nusselt number between a pair of ribs as well as the channel friction factor for 90◦, 45◦ and V-shape ribs. Different sensitivities are also investigated to get an understanding of the uncertainties found during the CFD implementation. These include the effect of mesh resolution, inlet turbulence intensity, rounded rib edges, wall roughness and temperature used for Reynolds number calculations. The Nusselt number and friction factor predictions of the turbulence models are also compared with existing empirical correlations. The results of the investigation show that the CFD results for 90◦ ribs deviate the most from experimental results, while closer results are seen for the 45◦ and V-shape ribs. In conclusion, the Lag Elliptic Blending k−ε model generally produces results closest to experimental data, especially for 90◦ ribs, but it shows some differences in Reynolds number trends. It proves to predict heat transfer and pressure loss closer to the experiment than the other models in flows where recirculation and reattachment has a significant impact. The Lag EB model is relatively stable and mesh independent. The SST k − ω model produces results rather similar to experimental data, but is unstable and sensitive to mesh resolution. The Realizable k − ε two-layer model produces results that are slightly less consistent with experimental data, but is very stable and insensitive to mesh resolution. The Nusselt number and friction factor from the investigated empirical correlations are closer to experimental results than the turbulence models for 90◦ inline ribs.
Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:liu-205501 |
Date | January 2024 |
Creators | Broberg, Viktor, Eklöw, Georg |
Publisher | Linköpings universitet, Mekanisk värmeteori och strömningslära |
Source Sets | DiVA Archive at Upsalla University |
Language | English |
Detected Language | English |
Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
Format | application/pdf |
Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0017 seconds