Benchmarks

Benchmarking is essential if engineering software is to be trusted. HeavyMetal therefore includes an automated verification and regression suite that can be run in a few minutes, providing a rapid check that changes to the code have not altered previously verified behavior. Each case is evaluated against quantitative acceptance criteria, making verification part of the routine development process rather than an occasional exercise. This testing is one component of a broader quality-assurance program intended to protect the integrity of the software as HeavyMetal continues to evolve.

Most benchmark problems compare finite element predictions directly with exact analytical solutions. Some of these solutions are deliberately simple, because simple problems are often the best way to isolate and test individual capabilities. Other cases reproduce classic analytical solutions from the literature. Where suitable benchmarks did not exist, we developed analytical solutions specifically to test new HeavyMetal capabilities. Together, these problems test both individual components of the formulation and the interactions among them.

The benchmarks presented here are selected examples from a substantially larger verification suite. They were chosen because they exercise some of the more challenging aspects of the multiphysics framework, including surface diffusion, stress-driven transport, thermomigration, evolving geometry, contact, and coupled thermal, mechanical, and diffusional behavior. In these problems, many parts of HeavyMetal must operate correctly together for the finite element solution to reproduce the analytical result.

The purpose is not simply to demonstrate interesting calculations. It is to demonstrate that HeavyMetal is developed as engineering software: tested systematically, quantitatively, and repeatedly, with verification built into the development process.