Efficient uncertainty propagation for network multiphysics systems

摘要

We consider a multiphysics system with multiple component models coupledtogether through network coupling interfaces, i.e., a handful of scalars. Ifeach component model contains uncertainties represented by a set of parameters,a straightfoward uncertainty quantification (UQ) study would collect alluncertainties into a single set and treat the multiphysics model as a blackbox. Such an approach ignores the rich structure of the multiphysics system,and the combined space of uncertainties can have a large dimension thatprohibits the use of polynomial surrogate models. We propose an intrusivemethodology that exploits the structure of the network coupled multiphysicssystem to efficiently construct a polynomial surrogate of the model output as afunction of uncertain inputs. Using a nonlinear elimination strategy, we treatthe solution as a composite function: the model outputs are functions of thecoupling terms which are functions of the uncertain parameters. The compositestructure allows us to construct and employ a reduced polynomial basis thatdepends on the coupling terms; the basis can be constructed with many fewersystem solves than the naive approach, which results in substantialcomputational savings. We demonstrate the method on an idealized model of anuclear reactor.