Analysis-Driven Geometry and Meshing for Large Scale Simulations

This presentation was made at the 2019 NAFEMS World Congress in Quebec Canada

Resource Abstract

Effective simulation of a variety of physics problems requires the initial model geometry to be altered or augmented during the simulation. Evolving domain problems such as ballistic flows and fracture mechanics require the model geometry to change as the simulation progresses. Additive manufacturing builds geometry up during the simulation. Accurate resolution of physics features such as shocks is possible by feature detection algorithms that capture the shock as a surface and insert it into the original geometry. Certain coupled problems such as fusion plasma simulations require CAD geometry to be combined with physics based geometry (e.g. flux surfaces). As the geometry is altered between analysis steps, the mesh must be faithfully evolved and adapted to these changes in geometry as well as to analysis needs such as error estimation results.

The scale of these problems often requires that these analyses run on large scale parallel computers. In order for the entire adaptive simulation to scale without bottlenecks, the parallel analyses must be coupled via functional interfaces with scalable, efficient geometry and meshing tools that also run on the same parallel computer.

This paper describes geometry and adaptive meshing capabilities that have been developed to support simulations where the analysis is a driver for the geometry. As the geometry is altered to satisfy analysis needs, the mesh is evolved and adapted as well, and maintains fidelity to the geometry. All mesh modifications for adaptation and quality are performed locally without a full remeshing for efficiency, and support efficient in-place solution transfer so that the solution data associated with the mesh is available for the next analysis step. These capabilities are supported in parallel, linking to the analysis code via functional interfaces, with automatic load balancing procedures ensuring parallel efficiency.

A number of state of the art simulations featuring analysis driven geometry are described. These include ballistic flows with focus on large projectile motion and multiphase propellant grain burn, additive manufacturing with focus on thermal simulation during selective laser melting, fracture mechanics with focus on crack insertion and propagation, and fusion plasma particle in cell simulations with focus on combining CAD and physics-based geometry.

Document Details

AuthorTendulkar. S
Date 18th June 2019
OrganisationSimmetrix Inc.


Back to Search Results