Researchers have developed a novel method for simulating turbulent systems using probability concepts.
A team from the University of Oxford has introduced an innovative technique to simulate turbulent systems centered on probabilities. Their research was published today (29 January) in the journal Science Advances.
Forecasting the behavior of turbulent fluid flows has been a key objective for scientists and engineers. However, despite advancements in computing technology, accurately simulating all but the simplest turbulent flows remains a daunting challenge.
This difficulty arises because turbulence consists of eddies and swirls of various forms and sizes interacting in chaotic and unpredictable ways. These fluctuations are so complex that even the most advanced supercomputers struggle to replicate them accurately, especially in engineering applications or weather predictions.
Collaborating with peers from Hamburg, Pittsburgh, and Cornell, the Oxford team redefined the problem to bypass the complications of directly simulating these turbulent fluctuations. Instead of tackling these challenging fluctuations head-on, they treated them as random variables that follow a probability distribution function. By simulating these probability distributions, the researchers could derive all essential metrics from the fluid flow (such as lift and drag) without getting bogged down by the chaotic nature of turbulence.
Typically, simulating probability distributions of turbulence necessitates solving high-dimensional Fokker-Planck equations, which are not feasible using traditional methods. To address this, the researchers utilized a quantum-inspired computing technology developed at the University of Oxford. This innovative approach employs ‘tensor networks’ to represent turbulence probability distributions in a highly compressed format that allows for effective simulation.
In their research, the quantum-inspired computing algorithm running on a single CPU core was able to perform computations in just a few hours, a task that would take several days for an equivalent classical algorithm on a full supercomputer.
Moreover, this increase in computational speed is just the beginning. Greater advantages are anticipated when the quantum-inspired tensor network algorithm is run on specialized hardware like tensor processing units and fault-tolerant quantum chips in the future.
The researchers believe this new approach not only pushes the boundaries of current turbulence simulations but also paves the way for simulating other chaotic systems that can be described probabilistically.
Lead researcher Dr. Nikita Gourianov from the Department of Physics at the University of Oxford stated: “The demonstrated and future computational benefits not only allow us to explore previously inaccessible domains of turbulence physics but also herald the development of next-generation computational fluid dynamics codes. These advancements could enhance weather predictions, improve vehicle aerodynamics, boost efficiencies in chemical industries, and much more.”
