Introduction

3D finite element software for convection-dominated problems is developed within the TETRAD mesh adaptivity framework. The development process is enhanced by integrating the software with the IRIS Explorer visualisation system. A computational steering environment is developed by allowing parameters that control various aspects of the numerical simulation to be interactively modified from within the visualisation environment. Numerical solution and visualisation proceed simultaneously within IRIS Explorer. A collaborative environment is developed by making use of the COVISA applications allowing geographically separate users to share some or all of the dataflow pipeline. The sharing of data can be at several levels; including dial parameters, rendered images and the full solution data produced by the numerical software, allowing for a wide range of collaborative scenarios.

This has led to improvements in the implementation of some visualisation techniques for large unstructured 3D data sets within IRIS Explorer, and the development of a computational steering environment within IRIS Explorer for large scale numerical simulations by embedding the numerical software in IRIS Explorer as a module in a similar manner to the visualisation processes.

This computational steering environment is not restricted to the 3D finite element application considered here and is expected to be integrated into future research projects within the unit.

This work was done by Mark Walkley together with Jason Wood and Ken Brodlie from the Visualization Research Group.

Example Computations

Time dependent problem

Linear convection of a scalar through the domain. Solved with a linear SUPG FEM and explicit time-stepping. An isosurface tracks the position of the interface between high and low concentration. Visualisation of the solution isosurface, and isovolumes of non-physical solution extrema around the interface.

Steady problem

Steady analogue of the previous problem. Solved with a linear SUPG FEM and preconditioned GMRES iterative solver. Visualisation of the solution with a slice, and error estimate with tetrahedra exceeding a specified limit shown. The mesh is h-adapted and the solution slice and error estimate are shown on the new mesh, showing that only 2 tetrahedra now fail the computed error test.

Solution slice and tetrahedra failing the error test pre-adaptivity.	Solution slice and tetrahedra failing the error test post-adaptivity.