A Study on the Possibility of Developing Foundational Models for the Simulation of Physical Phenomena
Please login to view abstract download link
We investigate the potential of foundation models for simulating physical phenomena, particularly in continuum (solid and fluid) mechanics. While "learned simulators" have demonstrated success in specific tasks, it is still unclear how well they can handle significant alterations in domain shape, boundary conditions, and/or constitutive laws while delivering robust (i.e., hallucination-free) and accurate results. In this paper, we conduct a thorough examination of these aspects, placing ourselves in the worst-case scenarios to assess their resilience to substantial changes in their application domain.
