Pattern Computer, the global leader in Pattern Discovery, announced a historic milestone in computational physics: successfully reconstructing the foundational Navier-Stokes equations for incompressible fluids in 2D directly from raw water-flow data. Starting with a dataset of simulated observational measurements of fluid flow dynamics over a short time period from a single set of initial and boundary conditions, and without assuming a single equation governs the system, Pattern rediscovered the Navier-Stokes equation using its innovative NeuralXAI product. In a validation framework, the Company’s partial differential equation architecture reverse-engineered the well-established Navier-Stokes equations solely by analyzing a fluid's vorticity profile under specific Reynolds number constraints.
The successful demo provides a vetted framework for researchers and engineers looking to model complex physical realities where governing equations do not yet exist.
By feeding its specialized Pattern Discovery Engine (PDE) spatiotemporal velocity vectors, Pattern naturally isolated and mathematically formulated the continuous-medium relationships that govern fluid dynamics. Operating with no prior physical assumptions, the system independently derived the exact partial differential equation governing vorticity that has formed the bedrock of fluid mechanics since the 19th century.
Mark R. Anderson, co-founding Chair and CEO, noted: “Pattern Computer’s work represents a notable advance in data-driven scientific discovery. By regenerating the foundational Navier-Stokes equations from raw water flow data, Pattern has shown how proprietary machine learning and pattern-recognition architectures can extract the underlying physics of fluid dynamics directly from empirical data, without relying on human-derived mathematical models. By treating a known, hyper-complex physical system as a target to be learned from behavior, Pattern has verified that its tools can accurately reconstruct foundational mathematical truths without relying on predefined formulas.
“This architecture serves as a critical proof of concept for automated physics discovery. We are not claiming to advance fluid dynamics or introduce new physics; the Navier-Stokes equations are already known to science. Rather, the significance lies in two critically important arenas: first, using a neural network to input physical data and then prove our ability to break through the “black box” and deliver true explainability using our NeuralXAI product; and, second, validation. By successfully reconstructing an established, chaotic system from raw observation, we can benchmark our AI’s output against scientific reality to confirm it identifies genuine physical laws instead of superficial data correlations.”
“Having validated this data-driven methodology on a known system, the logical next step is deployment against the unknown,” Anderson added. “This benchmark proves that under real-world, continuous conditions, our architecture can reliably turn raw empirical data into a computationally tractable, immediately actionable engineering tool. If applied to less understood or unmapped continuous systems, this same technique can be extrapolated to discover entirely new governing equations that are currently hidden from science.”
