Recent trends in engineering computing increasingly rely on multi-scale, multi-physics computer simulations to enhance predictive capabilities, by replacing conventional, largely empirically based methods, with a more scientifically based methodology. Traditional approaches employ a suite of stand-alone codes to simulate individual physical phenomena that are actually, and thus introduce errors. Take an example in nuclear engineering. The nuclear fuel code should have been coupled with the fission product release code. Without the coupling, the fuel code does not have a model of fission gas release. Hence, during transients, the internal pressure in the fuel element is calculated assuming that no fission gas be released to the gap. With coupling considered, the fission gas release code requires temperature as input from the fuel code. Meanwhile, the fuel code requires information of transient fission gas release calculated by the fission product release code to update the internal gas pressure in the fuel sheath so that it can model the appropriate mechanical response. This research will focus on improving our existing HPC software framework and code synchronization algorithms to offer the multi-physics computer simulations while boosting the performance of engineering computing.
The proposed framework can have broader applications in diverse fields such as nuclear safety, biomedical engineering and aerospace engineering. Students will explore different parallel and distributed programming models on multi-core/many-core computer systems to implement and evaluate this framework.