Robert Kowalski

2024-10

OpenFOAM Python automation framework — parameterized case generation, mesh refinement control, job submission to HPC cluster, and result extraction to pandas DataFrames for automated reporting.

Challenge: OpenFOAM case files use a custom dictionary format — not JSON or YAML. Wrote a Python parser for OpenFOAM dictionary syntax to enable programmatic case modification without string regex fragility.

Learned: Investing in a proper OpenFOAM dictionary parser pays back quickly when running parametric studies. Ad-hoc regex-based case file modification breaks for edge cases (nested dictionaries, lists) and produces bugs that are hard to track.

2024-04

Ground effect analysis for engine pod — CFD assessment of ground clearance effects on intake distortion and nacelle surface pressure at low speed / high angle of attack typical of takeoff.

Challenge: Modeling ground proximity required a moving ground plane boundary condition — OpenFOAM slip-wall with matching freestream velocity. Static ground simulation (zero-velocity wall) gave 12% different intake distortion coefficient.

Learned: Ground effect CFD must use the correct ground boundary condition (moving wall matching flight velocity) to accurately capture the vortex-ground interaction. Static wall gives non-physical vortex behavior near the ground plane.

2023-11

Nacelle drag decomposition analysis — pressure drag, friction drag, and interference drag components quantified using surface integral post-processing in OpenFOAM. Compared against ESDU design charts.

Challenge: Interference drag between nacelle and pylon required separate computations — with-pylon and without-pylon. The pylon interaction created a complex horseshoe vortex system that required finer mesh than baseline nacelle alone.

Learned: Installation effects (pylon, wing interference) can contribute 15-25% of total nacelle drag and cannot be assessed from isolated nacelle analysis. Always include pylon geometry in installation drag studies.

2023-07

Wind tunnel test correlation — compared OpenFOAM RANS results with pressure tap data and PIV velocity fields from a 1:4 scale nacelle model test. Evaluated prediction accuracy for drag and separation onset.

Challenge: OpenFOAM k-omega SST underpredicted separation onset angle by 4 degrees — too optimistic. k-kl-omega transition model gave better agreement for the leading-edge separation bubble but was computationally 3x more expensive.

Learned: For blunt-body aerodynamics near separation, turbulence model choice significantly affects accuracy. Fully turbulent RANS models are often too optimistic about separation. Transition-sensitive models or scale-resolving approaches (LES/DES) are needed for accurate separation prediction.

2023-02

External aerodynamics of engine nacelle using OpenFOAM with k-omega SST — drag coefficient, pressure distribution, and boundary layer behavior at cruise and high angle of attack.

Challenge: snappyHexMesh generated poor quality cells in the nacelle lip region (high aspect ratio, skewness >0.9). Required manual castellated mesh refinement levels and feature edge snapping configuration. Quality check using OpenFOAM checkMesh utility was essential before running.

Learned: snappyHexMesh for complex aerodynamic shapes requires iterative refinement of castellatedMesh and snap settings. The default settings are rarely suitable. Invest time in the checkMesh workflow — bad cells lead to divergence or wrong results silently.