Pade-legendre Method for Uncertainty Quantification with Fluid Dynamics Applications

Book excerpt: Abstract: The Pade-Legendre (PL) method, a novel approach for uncertainty quantification is introduced. The proposed method uses a rational function expansion and is designed to effectively characterize uncertainties in strongly non-linear or discontinuous systems. The discontinuities can be either in the underlying functions (inherent discontinuities) or from lack of sufficient data resolution (multi-scale discontinuities). In the former case, PL method can produce an accurate response surface without spurious oscillations and does not require prior knowledge of the discontinuities. For the latter type of discontinuities, the PL method can help reduce the number of deterministic simulations required to accurately represent the response surface. If sufficient data resolution is achieved, the PL method degenerates to standard polynomial reconstruction. The present approach is illustrated in a number of applications as an uncertainty propagation technique. Moreover, the method is applied to an inference problem in which a sharp discontinuity in the system input is present. The PL method shows a considerable improvement over the traditional approach when discontinuities are present. In addition, an ongoing effort called the UQ Experiment in which we used the PL method to help design the experimental setup is discussed.