Abhilash

From supercomputers of the 1980s to today’s digital twins, Computational Fluid Dynamics (CFD) has transformed how we understand and design the world around us. In Prof Mahesh Panchagnula's Podcast's recent episode, Prof. S. Balachandar, University of Florida, traces this remarkable journey—exploring the Navier–Stokes equations at the heart of CFD, the rise of simulations as indispensable design tools, and their power to tackle the “impossible experiments” of nature, from volcanic eruptions to dust storms. He emphasises that CFD is more than just high-fidelity simulations—it’s about asking the right questions, balancing detail with purpose, and delivering useful, actionable predictions. From industrial applications in oil, gas, and aerodynamics to cutting-edge roles in medicine and climate science, CFD connects scales as small as blood flow to as vast as the atmosphere. Prof. Bala also shares how digital twins and hybrid physics-AI models are shaping the future, why turbulence remains a grand scientific challenge, and why validation and scepticism are essential for trustworthy predictions. For aspiring engineers, he highlights the multidisciplinary nature of simulation and the importance of mastering fundamentals while exploring modern tools and pathways. A deep dive into the evolution, applications, and future of CFD. Watch the full episode here: youtube.com/watch?v=XGQQBZ…. @iitmadras #ComputationalFluidDynamics #DigitalTwins #EngineeringInnovation #IITMadras

I shouldn't be surprised that Gemini Pro thinks Chinese remainder theorem is 𝑑𝑖𝑣𝑖𝑠𝑖𝑣𝑒...cue CSI:Miami sunglasses and screech!

Very proud of everyone's presentations @ USNCCM in Chicago! Taeouk Kim, Victoria Sturgess, @meden96 , Andres Fernandez, Abhilash Malipeddi, Liam Wang and Haizhou Yang all did a great job! @umichCVC @UoMVascSurgery @UMichSurgery @UM_MICDE

Solution to Hilbert 6!!!! A medal is on its way!!! arxiv.org/abs/2503.01800

@SoRheology Bravo! May your trick-or-treat bag defy viscoelastic limits, filled with candies with a range of relaxation times.

Happy Halloween, everyone!

@costa_p buoyancy?

Temperature contours in two numerical simulations of the turbulent flow between a cold and a hot wall (flow from left to right), with the same Reynolds and Prandtl numbers. Can you guess what causes the differences? :) #FluidsFriday

@miniapeur Chirplet transform

There is the Fourier transform and then the wavelet transform. Is anything better than wavelets?

@costa_p Very impressive! The parallel TDMA is it a Schur complement based one?

Summer holidays bring nice news on the performance of our DNS solver CaNS. I am happy to share that we managed to speedup of Poisson solver by a factor of about 2 or more on many GPUs at scale by adopting a parallel TDMA! #cfd #gpu #hpc 1/

@Quasilocal so its not 1/𝛿 ∫𝓊 𝒹𝒞 then

Sneak peek into how math research is done

@bernat_font ditto on the meaning on smoother. My previous comment was due to this discussion scicomp.stackexchange.com/a/850

@DrDiffusion smoother in the sense that we smooth the residual, not solve up to the divergence-free tolerance - this is only done on the coarser level so it's not very costly. We previously used GS as smoother, but it was suffering from race-conditions on the GPU, so we switched to PCG.

dammit pressure solver...

@bernat_font PCG as smoother is interesting! Does this work better than a typical method like using Jacobi or sym. Gauss-Seidel smoother based multigrid as a preconditioner with CG outer solve? I was under the impression that Krylov methods generally don't make good smoothers for MG methods.

@DrDiffusion geometric multigrid with PCG smoother, and Jacobi preconditioner.

@LorenaABarba Calle de Alcalá, 16, Madrid, España.

Challenge question, tweeps—where am I?

Register for @CRIMSONsoftware workshop at #SB3C2024 ! Focus on particle-laden flows Free for registered attendees to SB3C Zoom option available. 'Virtual' registration required Register here: whova.com/portal/registr… @jesse_caps @umichCVC @UMichSurgery @DrDiffusion @meden96

@almormd Valid Perl program, no doubt.

Cloud computing

Anyone want to play Where's WaldAbhilash?

Taken at Avon Lake, OH.

Phone+Bino solar photography dry run. It was bit cloudy so no sunspots

@amirvaxman_dgp for linear regression of [t,y]: A = np.vstack([t, np.ones(len(t))]).T m, c = np.linalg.lstsq(A, y,rcond=-1)[0]

Nice question: what is the (small) piece of code you wrote the most times in your career? For me that would likely be the Laplacian matrix (or related subtasks like the d0 differential matrix).

Holi was during Holy Week this year.