World-first supercomputer discovers invisible flaw in all jet engines that no human had spotted

[mahatmakanejeeves]

Tech

World-first supercomputer discovers invisible flaw in all jet engines that no human had spotted

Published on Feb 17, 2026 at 5:01 PM (UTC+4)
by Daisy Edwards

Last updated on Feb 17, 2026 at 9:38 PM (UTC+4)
Edited by Daisy Edwards

A world-first supercomputer has done something incredibly important: discovered a flaw in jet engines that no human had ever seen. ... The breakthrough came after scientists used the most powerful computer on Earth to simulate the brutal environment inside a working turbine. ... What they found was not a dramatic crack or snapped component, but something far smaller and far more subtle. ... A microscopic surface flaw that had secretly affected performance for decades was finally exposed.

A world-first supercomputer spots an invisible flaw in all jet engines

Tech researchers working with Frontier, the exascale supercomputer at Oak Ridge National Laboratory, ran ultra-high fidelity simulations of jet engine turbine blades operating under extreme heat and pressure. ... These blades operate in the hottest, most punishing part of the engine, where airflow, combustion, and cooling systems all collide at once.

As turbine blades age, their surfaces become rough at a microscopic level. ... That roughness is impossible to properly analyze with traditional testing or smaller-scale computing, but Frontier changed that. ...By modeling billions of grid points at once, the system recreated both the full environment of the jet engines and the tiny surface degradation happening on each blade. ... The result was the first truly accurate engine representative 3D simulation that captured how teeny tiny amounts of wear affect the entire turbine’s performance.

The simulations revealed that this barely visible roughness can push airflow to shift earlier from smooth movement to turbulence. ... That transition increases aerodynamic issues and causes heat transfer on the blade surface. ... In simple terms, rough blades can reduce fuel efficiency and increase thermal stress on already extreme components, over time, that means higher operating costs, reduced durability, and more frequent maintenance. ... Even with Frontier’s immense power, each simulation took weeks to complete. ... On a standard laptop, the same calculations would have taken centuries; that scale of computing finally allowed engineers to see how tiny imperfections affect the entire system.

{snip}

[OAITW r.2.0]

1. So what did it expose and how do we fix it?

[Whyisthisstillclose]

2. Awesome! Can it tell us what is wrong with 37% of the American people yet?

[Beartracks]

3. Right? "tiny imperfections affect the entire system"

Each MAGA voter or disinterested voter is a tiny imperfection.

[Norrrm]

4. Alas! No matter how fast, logic cannot prove nor fix such a problem.

[OAITW r.2.0]

5. I Groked that question.

And it said Democrats are the reason everything is going to hell.

[FiveFifteen]

6. HA!

Racist syntax error?

[usonian]

7. Found it in more detail.

Most articles are a lot of hand-waving.

Here is the announcement from Oak Ridge National Lab with some detail.

https://www.olcf.ornl.gov/2026/01/27/frontier-provides-high-fidelity-insights-into-turbine-aerothermal-performance/
Researchers at the University of Melbourne in Australia collaborated with GE Aerospace. By the way, my first job after the service was at GE Aerospace.

Article at University of Melbourne.
https://eng.unimelb.edu.au/ingenium/engineers-keep-cool-on-jet-engines,-thanks-to-frontier-supercomputerengineers-keep-cool-on-jet-engines,-thanks-to-frontier-supercomputer

And they all reference a paper by ASME which you can have for a mere $25, UNLESS you know Usonian, who found the paper (or one almost the same) for free at the University of Melbourne.

https://people.eng.unimelb.edu.au/imarusic/publications/Edited%20Papers%202023/High-fidelity%20computational%20study%20of%20roughness%20effects%20on%20high%20pressure_Int%20J%20Heat%20Fluid%20Flow.pdf

High-fidelity computational study of roughness effects on high pressure turbine performance and heat transfer
Thomas O. Jelly ∗, Massimiliano Nardini, Marco Rosenzweig, John Leggett, Ivan Marusic, Richard D. Sandberg
Department of Mechanical Engineering, University of Melbourne, Victoria 3010, Australia
ARTICLE INFO
Keywords:
Roughness Heat transfer Turbulence

ABSTRACT
While blade surface roughness arising from in-service wear and/or the manufacturing process greatly affects aero-thermal performance, the detailed underlying physical mechanisms remain far from fully understood. In this study, a series of highly-resolved Large-Eddy Simulations of compressible flow past a high-pressure turbine vane with systematically varied levels of blade surface roughness have been performed, along with a smooth-blade simulation at matched flow conditions for comparison. Three non-dimensional roughness amplitudes have been investigated, namely, 𝑘s ∕𝑐 = {1.0, 2.0, 3.0} × 10−3 , where 𝑘𝑠 is an equivalent value of Nikuradse’s sandgrain roughness for an irregular, multi-scale near-Gaussian height distribution, and 𝑐 is the axial blade chord. All simulations have been performed at an axial chord Reynolds number of 0.59 × 106 and a Mach number of 0.9, based on the exit conditions of the reference smooth vane, and with synthetic inflow turbulence to mimic unsteady, three-dimensional disturbances from an upstream combustion chamber. The present investigation highlights the profound impact that blade surface roughness can have upon boundary- layer transition mechanisms, wall shear stress and blade surface heat flux, as well as the levels of turbulence kinetic energy and total pressure losses incurred in the wake. While blade surface roughness leads to major aero-thermal differences between the suction-side of the smooth and rough vanes, the pressure-side surface remains relatively unaffected — even for the largest roughness amplitude investigated here.


https://doi.org/10.1016/j.ijheatfluidflow.2023.109134
Received 31 January 2023; Received in revised form 6 March 2023; Accepted 14 March 2023
Available online 31 March 2023
0142-727X/© 2023 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

The paper is dated 2023, but seems to describe the study well.
and you can find out how well for $25.
Or not.

[ret5hd]

8. Pro tip: contact an author directly (email, etc)...

they can and often will send a copy to you for free.

[usonian]

9. I've had decent luck just searching.

Researchers gotta publish, so chances are good finding articles at the institution or at arxiv.

Except when DuckDuckGo sends me to a nearby restaurant instead.

[mahatmakanejeeves]

10. Thanks. This was a slipup on my part. I hadn't paid attention to the date.

Last edited Wed Feb 18, 2026, 09:30 AM - Edit history (2)

All I saw was the "February 17, 2026." I didn't look further to see that this went back to 2023. I'll try to pay closer attention next time.

And good morning.

[NNadir]

11. Frontier is not the world's first super computer. It's simply the most powerful as of now.

There has been a lot of angst about computational power because of power demands. Interestingly Oak Ridge is the place where the solution to the problem is actively being researched.

The lab is a powerhouse, pun intended, in materials design for extreme environments, jet engine Brayton devices being a small subset of these types.