Scilab for uncertainty in simulations: some numerical applications of the Polynomial Chaos Expansion method

Gregorio Pellegrini

University of Verona


In many real world applications a certain lack of knowledge could characterize the model of a given process through uncertainties of the parameters involved. Hence a single deterministic numerical simulation may not be enough to describe a certain physical phenomenon. Thus we move toward Uncertainty Quantification (UQ) methods, in particular analyzing the so-called Polynomial Chaos Expansion (PCE) technique which provides an enhanced approach, with respect to standard Monte-Carlo methods, to deal with uncertainties in simulations.

We present such technique by means of non-trivial numerical applications, exploiting the Non-Intrusive Spectral Projection (NISP) toolbox for Scilab, which implements the PCE in such environment.

In particular, we focus our attentions on solving a 2-dimensional Computational Fluid Dynamics problem as well as a couple of 2-dimensional advections problems in presence of uncertain parameters, highlighting both the flexibility of NISP toolbox, which interacts with several data sources, and its efficiency in detecting the required solution.

Back to Program