🧪 I took on an impinging jet CFD project where I customized the velocity inlet using my own User-Defined Function (UDF). I tried out different turbulence models—each with their own oddities—and compared the outcomes with published data and other CFD tools to see who really had the best predictions.

🧩 Since ANSYS CFX doesn't really handle pure 2D simulations, I got around that by squeezing my grid to just one cell wide—basically faking a 2D setup so I could still compare my results to what’s out there in the literature and experiments.

🧵 A big takeaway for me: the mesh matters—a lot. Depending on how fine or coarse the grid was, I’d see huge changes in my results, and the turbulence models could tell very different stories about the jet’s behavior. It was a real eye-opener for how critical it is to choose the right setup, and I also bumped into a few quirks and limits with CFX along the way.

What came out of it?

• 🔍 First, I saw firsthand how mesh quality can make or break your results.
• 🌪️ I got a clearer sense of how turbulence models behave and why it’s worth trying more than one—sometimes they disagree a lot.
• 🛠️ I now know how to set up and run custom UDFs, making simulations much more flexible.
• 📏 It was also a real exercise in checking and benchmarking results—comparing with published data taught me a lot about what to trust (and what not to).

Lessons learned

• ✍️ Writing my own velocity-inlet UDF let me really control the inflow and keep things repeatable.
• 📐 Going “pseudo-2D” in CFX is a neat trick if you want to run 2D-style cases in a software that’s not built for it.
• 🧵 The turbulence model you pick seriously changes your results—SST, for example, gave different predictions for heat transfer and flow details near the wall compared to some others.
• 🧱 Mesh around the jet core and stagnation point needs to be tight if you want stable and accurate results.
• 🧪 Double-checking everything with published work and other CFD packages made it clear where CFX shines—and where it doesn’t.

🚀 In the end, the project was a deep dive into practical CFD, code tweaks, and getting creative with software. I learned a lot by trying, failing, adapting, and then testing everything against real-world data—not just trusting what the computer spits out.

Aeromechanics Engineering Experience

During my time as an aeromechanics engineer at the Warsaw Institute of Aviation collaborating with GE Aerospace, I focused exclusively on commercial aircraft engine compressors and turbines. My role combined computational analysis with real-world test data to ensure blade reliability and engine safety under demanding flight conditions.

Campbell Diagram Analysis for Engine Rotating Assemblies

I routinely generated Campbell diagrams for compressor and turbine blade rows to pinpoint where blade natural frequencies intersected with engine excitation harmonics. By mapping these intersections across the speed range, I identified critical speed bands that could trigger forced response or flutter. This work enabled design adjustments—such as tweaking blade mass or stiffness—to steer clear of dangerous resonance zones without compromising efficiency.

Modal Analysis of Compressor and Turbine Blades

My modal analysis characterized how individual compressor and turbine blades vibrate under aerodynamic and thermal loads. For compressor stages, I examined how inlet distortions and stage interactions influenced blade modes. For turbine stages, I incorporated temperature‑dependent material properties to capture shifts in natural frequencies during operation. Validating these models against blade tip‑timing and strain‑gauge measurements ensured accurate prediction of vibrational behavior across engine sections.

Fatigue Life Assessment Using Goodman Diagrams

To predict blade service life, I applied Goodman diagram techniques tailored for aircraft engine conditions. By plotting alternating stress amplitudes from vibration against mean stresses from centrifugal and thermal loads, I could estimate high cycle fatigue (HCF) life for both compressor and turbine blades. This approach supported maintenance planning by defining safe inspection intervals based on realistic loading scenarios.

High Cycle Fatigue (HCF) Analysis

My fatigue work focused solely on high cycle fatigue (HCF), addressing the elastic‑deformation cycles induced by steady‑state engine vibrations. Using experimental data from operational engines and test rigs, I refined damage‑accumulation models to improve life‑prediction accuracy. This ensured blade designs met GE’s stringent durability requirements under continuous high‑frequency loading.

Integrated Computational and Experimental Workflow

I combined CFD‑based aerodynamic loading with finite element (FE) structural models to simulate compressor and turbine blade dynamics within full-engine assemblies. These multiphysics simulations captured interactions between adjacent blade rows and accounted for real boundary conditions such as inlet pressure distortions and combustor exit temperature patterns. All results were rigorously validated against full‑scale engine tests and component‑level rig data to confirm model fidelity.

Field Issue Investigations

A highlight of my work was leading investigations into real engine issues reported by operators—like unexpected compressor blade cracking on high‑flight‑cycle engines. By correlating field vibration data with wear patterns, I identified root causes (e.g., particle erosion amplifying vibratory stress) and recommended mitigation measures such as updated blade coatings and optimized inspection schedules. These solutions directly improved engine reliability and reduced unscheduled maintenance events.

This hands-on experience with aircraft engine compressors and turbines honed my skills in vibration analysis, fatigue life prediction, and multiphysics modeling—expertise that’s directly applicable to advancing the safety and performance of modern jet engines.

Full-Flight Mach Regime Analysis

I ran CFD simulations in ANSYS Fluent for:

• Subsonic (M < 0.8): Mapped pressure distributions during liftoff roll and early ascent.
• Transonic (0.8 < M < 1.2): Investigated shock–boundary layer interactions over the nose cone and interstage fairings, optimizing shapes to delay separation peaks.
• Supersonic (1.2 < M < 5): Studied shock patterns and expansion fans on the stainless-steel airframe, ensuring stability through Mach-dependent center-of-pressure shifts.

These studies highlighted transonic separation hotspots and aftward shifts in aerodynamic center at supersonic speeds.

Python Automation for Post-Processing

I built Python scripts to parse Fluent output, automatically extracting:

• Surface pressure (Cp) distributions
• Aerodynamic force and moment coefficients
• Shock location metrics

Automated plotting and report generation accelerated result analysis and design feedback.

Key Aerodynamic Insights

• Shock-Boundary Layer Interaction: Improved fairing transitions to reduce drag spikes in transonic flight.
• Center-of-Pressure Migration: Adjusted fin sizing and mass distribution to maintain stability at varying Mach.
• Plume-Induced Flow Fields: Refined nozzle-aft contours to smooth recirculation zones, enhancing separation reliability.

This internship sharpened my expertise in ANSYS Fluent simulations, Python-driven post-processing, and comprehensive Mach-regime aerodynamics—skills that directly support cutting-edge launch vehicle development.

🚀 Aerodynamic Performance Analysis of Starship Reentry

Warsaw University of Technology (WUT)

I conducted an extensive computational study on Starship's aerodynamic performance during atmospheric reentry, focusing on how its body shape, forward canards, and aft flaps interact to control descent stability, lift, and drag under extreme conditions. The work combined high‑fidelity CFD simulations (with a 15.5‑million‑cell mesh validated through sensitivity studies) using ANSYS Fluent to accurately resolve shock‑wave structures and boundary‑layer dynamics throughout subsonic, transonic, and hypersonic flow regimes.

🧪 Study Objective & Method

• Mapped how angle of attack and control‑surface deflections affect forces and moments across reentry regimes where loads and heating are critical.
• Simulated a matrix of canard/flap settings—both independently and in counterposition.
• Confirmed that a superposition principle reliably predicts combined effects, reducing compute cost while maintaining accuracy.
• Executed mesh‑independence and solution‑convergence checks.

📊 CFD Setup Highlights

• Mesh: 15.5M cells, refined shock/BL resolution; prism layers tuned for y+ targets.
• Regimes: Subsonic → Transonic → Hypersonic; density‑based solver with regime‑appropriate CFL control.
• Physics: Shock capturing, separation/reattachment behavior, control‑surface hinge‑line effects.
• Post‑processing: Automated force/moment extraction, Cp lines, L/D maps, trim analysis.

🔧 Control‑Surface Matrix & Trim

• Quantified lift surges, transonic drag rise, and trim requirements (Cm trends, center‑of‑pressure migration).
• Identified optimal canard/flap strategies for controllability, heat load management, and reduced post‑flight wear.
• Established trim tables for guidance and entry corridor targeting.

✅ Validation & Deliverables

• Compared trends against available flight‑test data and literature benchmarks.
• Delivered a reusable aerodynamic database (AoA × deflections × Mach) and Python tools for interpolation and envelope checks.
• Provided guidance‑law inputs and control‑surface schedules for closed‑loop reentry concepts.

📈 Impact for Rocket Employers

• Cost & schedule: Superposition‑based prediction cut compute time while preserving fidelity.
• Risk reduction: Early detection of transonic drag spikes and trim gaps informs heat‑load margins and actuator sizing.
• Ops readiness: Clean handoff artifacts (ADB + scripts) for GNC, testing, and flight readiness reviews.

🔭 What’s Next

• Expand flap/canard matrices; incorporate wind‑tunnel and flight‑test cross‑validation.
• Integrate ADB into trajectory optimization and entry guidance algorithms.

🏷️ #Starship #CFD #Reentry #ANSYSFluent #Hypersonics #ControlSurfaces #Guidance #GNC #WUT

🏙️ Aerodynamics of Urban Structures

Warsaw University of Technology (WUT)

Led a comprehensive CFD study to evaluate how introducing a high-rise building affects pedestrian-level wind conditions within an existing urban block. This work combined numerical simulation with experimental validation to ensure accurate representation of real-world airflow and to refine best-practice guidelines for urban planning.

🎯 Project Scope and Objectives

• Develop a scaled 1:1500 model of an urban district featuring two main building clusters to match wind-tunnel test conditions.
• Simulate wind flow at pedestrian height (1–2 m) for three wind directions (0°, 45°, 90°) using ANSYS Fluent.
• Compare CFD velocity contours against wind-tunnel measurements to assess model fidelity and identify divergent regions.
• Optimize mesh design and boundary conditions to achieve robust, grid-independent results.

🛠️ Methodology

Geometry and Scaling: Converted real-size building layouts into a 0.6 m-tall scaled model based on a maximum building height of 90 m.

Mesh Generation: Created both coarse (~23k cells) and fine (~1.26M cells) unstructured meshes, enforcing y⁺ ≈ 30–300 for SST k–ω modeling. Mesh quality metrics met skewness (< 0.7) and orthogonality (> 0.3) targets.

UDF Implementation: Developed User-Defined Functions to impose logarithmic velocity-inlet profiles and altitude-dependent turbulence intensity, matching terrain categories III and IV. Additional UDFs recorded time-averaged velocity probes and pressure coefficients for automated data extraction.

Solver Settings: Employed the RANS SST k–ω “realizable” model with second-order discretization and SIMPLE pressure-velocity coupling under incompressible, steady-state assumptions. Side and top walls used specified-shear conditions, while building walls applied no-slip boundaries.

📊 Key Findings and Validation

• Fine-mesh simulations reproduced wind-tunnel velocity distributions within ±10% in most pedestrian zones, confirming CFD’s capability to predict critical acceleration hotspots.
• Notable discrepancies arose behind tall building edges for 90° flow, prompting further mesh refinement and turbulence-intensity sensitivity checks.
• Wind directions of 0° and 45° showed strong alignment between CFD and experimental data, validating the inlet profile and turbulence modeling approach.

🧠 Challenges and Lessons Learned

• Mesh Sensitivity: Coarse meshes failed to capture fine-scale vortices between closely spaced buildings. Targeted local refinement proved more efficient than global densification.
• UDF Performance: Initial UDFs introduced solver instability due to heavy in-loop computations. Refactoring to minimize hook frequency and using efficient array operations stabilized convergence.
• Turbulence Intensity Calibration: Accurate pedestrian-level predictions required iterative tuning of turbulence intensity (optimal TI ≈ 25%) to match wind-tunnel inlet conditions.
• Automated Data Extraction: Embedding probe data logging within UDFs significantly streamlined post-processing and improved repeatability of validation procedures.

✅ This project delivered validated CFD workflows for urban wind-comfort assessment and provided actionable insights for architects and planners on mitigating adverse wind effects through building placement and design.