Physical Systems Modeled Without Physical Laws

TL;DR

This paper explores using tree-based machine learning models to emulate complex physics simulations, aiming to predict detailed spatial-temporal data and improve resolution without costly recalculations.

Contribution

It demonstrates that machine learning can effectively approximate physics simulations, reducing computational costs and generalizing to finer grids without explicit physical laws.

Findings

Tree-based models accurately emulate Navier-Stokes, stress, and electromagnetic simulations.

Models successfully predict intermediate spatial-temporal data.

Approach reduces computational costs for high-resolution physics predictions.

Abstract

Physics-based simulations typically operate with a combination of complex differentiable equations and many scientific and geometric inputs. Our work involves gathering data from those simulations and seeing how well tree-based machine learning methods can emulate desired outputs without "knowing" the complex backing involved in the simulations. The selected physics-based simulations included Navier-Stokes, stress analysis, and electromagnetic field lines to benchmark performance as numerical and statistical algorithms. We specifically focus on predicting specific spatial-temporal data between two simulation outputs and increasing spatial resolution to generalize the physics predictions to finer test grids without the computational costs of repeating the numerical calculation.