Browsing by Author "Turnquist, Brian"

Now showing 1 - 2 of 2

A fast, decomposed pressure correction method for an intrusive stochastic multiphase flow solver
(2021-05) Turnquist, Brian; Owkes, Mark
Solution of the pressure Poisson equation is often the most expensive aspect of solving the incompressible form of Navier–Stokes. For a single phase deterministic model the pressure calculation is costly. Expanded to an intrusive stochastic multiphase framework, the simulation expense grows dramatically due to coupling between the stochastic pressure field and stochastic density. To address this issue in a deterministic framework, Dodd and Ferrante (“A fast pressure-correction method for incompressible two-fluid flows” Journal of Computational Physics, 273, 416–434, 2014) discuss a decomposed pressure correction method which utilizes an estimated pressure field and constant density to modify the standard pressure correction method. The resulting method is useful for improving computational cost for one-fluid formulations of multiphase flow calculations. In this paper, we extend the decomposed pressure correction method to intrusive uncertainty quantification of multiphase flows. The work improves upon the original formulation by modifying the estimated pressure field. The new method is assessed in terms of accuracy and reduction in computational cost with oscillating droplet, damped surface wave, and atomizing jet test cases where we find convergence of results with the proposed method to those of a traditional pressure correction method and analytic solutions, where appropriate.
MultiUQ: A software package for uncertainty quantification of multiphase flows
(2021-11) Turnquist, Brian; Owkes, Mark
multiUQ is a novel tool that simulates gas-liquid multiphase flows and quantifies uncertainty in results due to variability about fluid properties and initial/boundary conditions. The benefit over a typical deterministic solver is that inexact information, such as variability in fluid properties or flow rates, can be included to determine the affect on simulation solutions. It is common to deploy non-intrusive methods which utilize many solutions from a deterministic solver to generate a distribution of possible results. Contrarily, multiUQ uses an intrusive uncertainty quantification method wherein variables of interest are functions of space, time, and additional uncertainty dimensions. The intrusive solver is run once, giving a distribution of solutions as an output, as well as desired statistics. We use polynomial chaos to create the stochastic variables, which represent a distribution of values at each grid point. The stochastic variables are substituted into the incompressible Navier-Stokes equations, which govern the stochastic fluid dynamics. A stochastic level set is used to capture the distribution of interfaces that are present in an uncertain multiphase flow. multiUQ is written in Fortran and uses a message passing interface (MPI) for parallel operation. Given the many applications of multiphase flows, including open flows, hydraulics, fuel injection systems, and atomizing jets, there is a massive potential benefit to calculating uncertainty information about these flows in a cost-effective manner.