2026-04-13T04:10:11Zhttps://recercat.cat/oai/request

oai:recercat.cat:2117/4467702025-11-23T05:49:15Zcom_2072_1033col_2072_452951

00925njm 22002777a 4500 dc Carrillo Vilà, Jordi author 2025-10-23 This thesis evaluates low-order loss-prediction methods for axial gas-turbine compressors by benchmarking them against high-fidelity CFD data to assess suitability for preliminary design. Four representative correlations are evaluated: Ainley & Mathieson (AM), Dunham & Came, Storer & Cumpsty (SC) and Banjac et al. (Ban). Pyhton-based post-processing of wall-resolving large eddy simulation (WRLES) data was conducted to obtain pressure loss coefficient and spanwise distributions. ParaView was used for diagnostic visualization. WRLESindicates a loss distribution dominated by the tip region, with smaller hub and midspan contributions. Two-dimensional velocity and pressure loss fields corroborate concentrated dissipation near endwall regions, particularly at the blade tip and casing, where intense flow mixing occurs. Comparisons reveal heterogeneous fidelity across models. While AM reproduces the overall loss magnitude reasonably well, it tends to misallocate the loss mechanism by underweighting tip-driven contributions. DC incorporated Mach and Reynolds number corrections with revised secondary and clearance terms. However, overall losses are overestimated while it still undervalues tip clearance and endwall effects. SC, which focuses on tip clearance mixing, predicts losses far below CFD results for the present geometry and operating point. Such discrepancy persists despite the efforts and verification checks. Ban formulation best captures isolated tip-clearance loss (approximately ten percent deviation) but returns unrealistically high spanwise distribution of additional losses. Collectively, the results show that legacy turbine-based correlations can predict reasonably accurate total compressor losses yet misattribute the dominant sources. In contrast, modern compressor-specific formulations display strong sensitivity to boundary layer conditions and input assumptions. While other loss-prediction methods may prevail, low-order correlations still remain a widespread tool for design validation to balance speed with high physical fidelity. http://hdl.handle.net/2117/446770 Àrees temàtiques de la UPC::Enginyeria mecànica::Motors::Turbomàquines Àrees temàtiques de la UPC::Física::Termodinàmica Gas-turbines Compressors Computational fluid dynamics Turbines de gas Compressors Dinàmica de fluids computacional Study on gas turbine compressor efficiency modelling