Izabela Nowacka

2024-08

1D-3D coupled simulation — 1D thermodynamic cycle model (in-house code) coupled with 3D CFX compressor stage. 1D provides total conditions at compressor inlet; 3D returns stage performance map; iteration until convergence.

Challenge: Coupling stability — 1D and 3D codes use different timestep conventions and convergence criteria. Implemented relaxation on the coupling variables (total pressure, total temperature) to prevent oscillation between the two solvers.

Learned: 1D-3D coupling requires careful design of the coupling interface and relaxation strategy. The coupling loop converges much more slowly than either standalone model. Expect 20-50 coupling iterations for <1% convergence on coupled variables.

2024-03

Comparison of RANS (k-omega SST in CFX) vs Scale-Resolving Simulation (SRS — SAS-SST) for high-load compressor rotor. Evaluated prediction accuracy of separation bubble on blade suction surface.

Challenge: SRS required time-stepping with very small physical timestep (1 blade passing period / 20 timesteps) and 10 flow-through times for statistical convergence. Total cost was 40x the RANS equivalent. Justified only for validation case, not for design iteration.

Learned: Scale-resolving simulations are valuable for understanding flow physics (separation, secondary flows, transition) but are not practical for daily design work. Use RANS for design iteration and SRS for targeted validation and physical insight.

2023-11

Circumferential total pressure distortion analysis — applied DC60 distortion descriptor inlet condition per ARP1420, predicted compressor response in terms of local stage work redistribution and stability margin reduction.

Challenge: Full annular computation with 360-degree distortion inlet was 20x more expensive than a single-passage calculation. Required HPC with 200 cores for overnight turnaround. Sector-based approach with periodic boundaries was not possible due to non-periodic distortion pattern.

Learned: Full annular distortion simulations are necessary for non-periodic inlet conditions. Sector modeling with approximate distortion representation can miss important inter-blade channel interactions that affect stability margin.

2023-07

Tip clearance parametric study — ran 6 clearance values (0.3% to 1.5% tip chord) through full stage CFD, extracted efficiency and pressure ratio sensitivity to clearance degradation.

Challenge: Very small tip clearances (< 0.4%) required extremely fine mesh in the tip gap — 15+ layers of elements in a gap of ~0.2mm. Mesh generation time exceeded 8 hours per case. Developed automated TurboGrid template that generated all 6 cases from a parameter file.

Learned: Tip clearance CFD mesh density requirements are extreme — at least 10 elements across the gap are needed for mesh-independent results. Automated mesh generation is essential; manual meshing 6 tip clearance variants is not practical.

2023-02

Axial compressor stage analysis using ANSYS CFX with mixing plane interface between rotor and stator passages. Computed stage performance at design point and ±20% mass flow.

Challenge: Mixing plane stability near the stall line was poor — simulation diverged within 50 iterations at mass flows 15% below design. Required conservative upwind scheme activation and reduced false timestep for near-stall points.

Learned: Compressor CFX simulations near stall require different numerical settings than design point runs. Lower physical timestep (10-100x lower than design point), aggressive convergence monitoring, and sometimes a good initial field from a converged nearby point are all necessary.