http://hdl.handle.net/2268.2/1967 Wed, 14 Oct 2026 22:34:22 GMT 2026-10-14T22:34:22Z http://hdl.handle.net/2268.2/27556 Title: Development of a Method for the Aerothermal Pre-Design of an Air-Oil Heat Exchanger Safran Aero Boosters Abstract: Next-generation geared turbofan engines with ultra-high bypass ratios face severe thermal management constraints, as the power gearbox generates heat loads that cannot be dissipated without dedicated air-oil heat exchangers integrated directly into the bypass duct. Sizing these exchangers requires simultaneously maximizing heat rejection and limiting aerodynamic pressure losses within tight geometric envelopes. This is a trade-off that manual iterative workflows, combining simplified empirical tools and 3D CFD evaluations, struggle to explore efficiently. This thesis addresses this limitation by developing an automated aerothermal optimization framework for preliminary air-oil heat exchanger sizing, built around a 2D axisymmetric CFD formulation to keep computational costs compatible with industrial constraints. A systematic benchmark of Design of Experiments strategies and surrogate modeling techniques established that a 20-point Optimal Space-Filling design paired with a second-order polynomial response surface offers the best compromise between accuracy and computational cost. Kriging performed marginally better on smooth outputs but exhibited spurious oscillations on the highly nonlinear flow separation metric, making the polynomial surrogate the more reliable choice for industrial use. Application of the framework identified an optimal inlet-outlet height configuration achieving 70.73kW of heat rejection, with surrogate prediction errors below 1% for heat transfer and pressure loss. Validation against Safran's empirical tool yielded a discrepancy of only 2.34%, and comparison with 3D sector simulations confirmed that global performance metrics agree within 0.5%, with three-dimensional effects limited to a 3.9% additional pressure loss. A final comparative study between two matrix compactness levels showed that reducing the hydraulic diameter substantially increases internal flow resistance and triggers boundary layer separation downstream of the exchanger. Both configurations were evaluated based on their mass flow–thermal conductance characteristic curves to establish the optimal operating trade-off. Sun, 06 Sep 2026 22:00:00 GMT http://hdl.handle.net/2268.2/27556 2026-09-06T22:00:00Z http://hdl.handle.net/2268.2/27549 Title: Optimization of Calibration Parameters for Through-Flow Loss Correlations in Axial Flow Compressors Safran Aero Boosters Abstract: The preliminary aerodynamic design of axial-flow compressors relies on two-dimensional meridional through-flow solvers. While these tools allow rapid design space exploration, their accuracy is limited by empirical correlations used to model viscous and compressible losses. Conducted with Safran Aero Boosters, this Master's thesis establishes an automated numerical framework to recalibrate empirical profile and shock loss parameters against higher-fidelity two-dimensional cascade CFD data. Two optimization methods were evaluated following the validation of the cascade solver as a reference tool. First, a deterministic gradient-based method using a trust-region reflective algorithm calibrated loss parameters for individual and multi-streamline configurations. An enhanced objective function incorporating derivative errors was used to better capture loss polar curvature. Second, a surrogate-based optimization framework evaluated losses along constant speed lines, allowing the inlet Mach number and incidence angle to vary simultaneously and better match real compressor operation. The results show clear trade-offs between physical accuracy, computational speed, and calibration capability. In subsonic regimes, the gradient-based approach yielded rapid convergence and significant error reductions, but its performance in supersonic regimes was constrained by the formulation of the underlying correlations. The surrogate-based approach confirmed that constant Mach sweeps introduce minor errors away from choking conditions, though it faced difficulties in aligning the minimum loss location and matching the loss profile shape. This work provides automated calibration workflows and practical guidelines for early-stage compressor design, clarifying where parameter tuning is effective and where model reformulations are required. Future work should focus on refining objective function formulations, adding minimum loss alignment constraints, filtering near-wall streamlines, and testing surrogate-based optimization on supersonic rotor configurations. Sun, 06 Sep 2026 22:00:00 GMT http://hdl.handle.net/2268.2/27549 2026-09-06T22:00:00Z http://hdl.handle.net/2268.2/27448 Title: Operability Limits of Turbofan Engine Test Cells: A Comparative Investigation Across Two Distinct Architectures Safran Aero Boosters Abstract: Ground test cells are used for engine development, for post-assembly acceptance testing, and for maintenance testing. A non-uniform total pressure field at the engine inlet reduces the surge margin of the fan and compromises the repeatability of the measurement. Acceptance criteria therefore limit the distortion level admitted at the engine inlet. Modern turbofans feature increasing fan diameters and mass flow rates, whereas existing test cells keep their original cross-sectional area. This trend narrows the operability domain of these facilities. Two test cell architectures are compared: a 9 m wide cell designed by Safran Test Cells and a competitor cell with a 10 m wide test room. Steady Reynolds-averaged Navier-Stokes simulations are performed with Menter's shear stress transport turbulence model. Two campaigns are carried out, which differ by the inlet condition imposed upstream of the test cell. The first one varies the cell bypass ratio, defined as the ratio of the mass flow passing through the test cell to the engine mass flow, under a uniform inlet profile. The second one imposes a non-uniform inlet velocity profile representative of a crosswind. In both campaigns, the inlet flow quality is measured by the peak total pressure distortion at the engine inlet. The acceptance criterion is exceeded at a bypass ratio of 140 percent for the 9 m test cell and at 110 percent for the competitor cell. The competitor architecture therefore offers an operability margin that is 30 percentage points wider. The same gap is found under both inlet conditions, which strengthens the comparison. The degradation mechanisms, however, differ. In the 9 m test cell, the door wake is the dominant distortion source under both inlet conditions. In the competitor cell, the door governs the violation under a uniform inlet profile, whereas monorail wake becomes dominant once the inlet profile is distorted. Three geometric features explain the better behavior of the competitor architecture: a larger cross-sectional area, a door with a rounded edge, and a tapered test room that accelerates the bypass flow. These results also question the guidelines commonly used in the industry. A cell bypass ratio above 80 percent is usually recommended. All the simulated cases satisfy this value, yet the acceptance criterion is already violated at 140 percent for the 9 m test cell. The first violations take the form of intermittent peaks rather than a rise of the mean distortion level. This behavior is inferred from the iterative convergence of the steady simulations, and it suggests that the peak value is the relevant quantity to monitor. Improving the 9 m architecture therefore requires a dedicated study of the door, which is identified here as the dominant distortion source. Sun, 06 Sep 2026 22:00:00 GMT http://hdl.handle.net/2268.2/27448 2026-09-06T22:00:00Z http://hdl.handle.net/2268.2/27446 Title: Modelling Investigations of a 3D Supersonic Coherent Jet towards a Parametric Study for Steelmaking Applications in EAF Université de Liège Abstract: Supersonic coherent jets are used in EAF steelmaking to maintain oxygen-jet momentum over the distance separating the lance from the molten bath. Numerical predictions are strongly affected by turbulence, combustion and chemical modelling choices, but consistently show that a longer stoichiometric flame is associated with a longer potential core. Premature turbulent mixing remains a major limitation, particularly under hot-ambient conditions, where the interaction between the oxygen jet and the surrounding high-temperature gas becomes important. The turbulent-temperature-gradient correction provides the strongest improvement, while reduced chemistry should be reassessed against a detailed mechanism such as GRI-Mech~3.0 when operating conditions are modified. An automated framework, \textsc{HAstignis}, was developed to support systematic parametric studies of lance geometry and operating conditions Sun, 06 Sep 2026 22:00:00 GMT http://hdl.handle.net/2268.2/27446 2026-09-06T22:00:00Z