Utsav Purkait
I work on computational fluid dynamics — turbulence modelling, vortex dynamics, and the question of where physical models succeed and where they fail. My current research at IIT Bombay examines the structural limits of RANS closures in turbulent free-shear jets, using high-fidelity data to show why certain failures cannot be calibrated away. Turbulent Thoughts is where I write about this work in the open.
Background
I trained as a mechanical engineer at BITS Pilani, Hyderabad, where the problems that held my attention were the ones our models couldn’t quite solve. Turbulence stood out: a phenomenon governed by equations known for nearly two centuries, yet still impossible to predict without approximation.
My research follows that tension. At IIT Bombay, I study how RANS closures misrepresent Reynolds stress anisotropy in turbulent free-shear jets — failures that turn out to be structural rather than numerical. At the University of Wyoming, I designed and simulated a tornado-like vortex generator, testing how far standard models can be trusted in strongly swirling flow.
What draws me to CFD is less the answers it produces than the discipline it demands. Every result is a claim to be verified, validated, and questioned — and understanding precisely where our models break down seems to me the most honest way to improve them.
Why this blog
Most CFD resources either stop at button-clicking tutorials or jump straight into dense mathematics. Turbulent Thoughts tries to occupy the space in between — rigorous enough to be useful, clear enough to actually build intuition.
I write about the failures as much as the successes: solvers that silently break conservation, models that fail structurally rather than numerically, simulations that look right but aren't. That's where the real learning is.
Timeline
- Studying structural failure modes of RANS closure in representing turbulence anisotropy using high-fidelity LES data
- Analysing anisotropy states through Lumley triangle, barycentric map, and anisotropy tensor glyphs
- Identified systematic RANS bias toward plane-strain anisotropy — failure arises from structural constraints of the Boussinesq hypothesis, not model calibration
- Preparing manuscript for Physics of Fluids
- Led CFD simulation and validation of a dual-shell tornado simulator replicating Rankine vortex dynamics using an 8M-cell RANS k-ω SST model in Fluent
- Conducted parametric studies varying fan speed and guide vane angle, analysing influence on vortex intensity, core radius, and swirl ratio
- Physical experiments on fabricated model showed deviations less than 8.5%
- Developed 2D transient CFD models for flow past circular and square cylinders sweeping Re from 5 to 50,000
- Extracted shedding frequencies via FFT on lift coefficients; quantified Strouhal–Reynolds relationship against established data