A multilevel method for meshless solution of the poisson equation in heat transfer and fluid flow

Anand Radhakrishnan, -

Permalink

https://hdl.handle.net/2142/112966 Copy

Description

Title

A multilevel method for meshless solution of the poisson equation in heat transfer and fluid flow

Author(s)

Anand Radhakrishnan, -

Issue Date

2021-06-23

Director of Research (if dissertation) or Advisor (if thesis)

Vanka, Surya Pratap

Department of Study

Mechanical Sci & Engineering

Discipline

Mechanical Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Meshless
• Multigrid
• Radial Basis Function based Finite Difference
• Polyharmonic Spline
• Poisson Equation
• Incompressible Navier--Stokes Equation

Abstract

Meshless methods using radial basis functions (RBF) are an attractive alternative to grid based methods for solving partial differential equations in complex geometries. Gaussian, Multiquadratics and inverse Multiquadratics are some of the more popular RBF's, but the require a shape paramter for a stable and accurate solution and also face stagnation issues. Recently, Polyharmonic splines (PHS) with appended polynomials have overcome the aforementioned issues and offer spectral convergence of the discretization errors with the degree of appended polynomials. In this thesis, we present a non-nested multilevel algorithm using the PHS-RBF meshless method for the soluution of the Poisson equation, which commonly arises in numerous heat transfer and fluid flow applications. The PHS-RBF discretization of the Poisson equation leads to a sparse set of equations with unknown variables at each of the scattered point. The non-nested multilevel algorithm solves this set of equations by restricting and prolongating the values and corrections between multiple independently generated coarse set of points by making use of RBF interpolation. The performance of this algorithm is tested for the Poisson equation in three model geometries, using manufactured solutions. Rapid convergence of the residual is observed with Dirichlet boundary conditions using Successive Over-Relaxation(SOR) as the relaxation scheme . However, convergence is seen to be quite modest for the all-Neumann boundary condition, but this poor convergence is ameliorated by using the multilevel algorithm as a preconditioner to the GMRES, which is a Krylov Subspace Projection (KSP) method. This rapid convergence of the all-Neumann equation is then applied to the pressure Poisson equation arising in the fractional step method, with explicit convection and explicit diffusion. We demonstrate fast convergence, both with refinement of number of points and degree of appended polynomials, for various fluid flow problems in complex domains with high accuracy using the meshless fractional step algorithm.

Graduation Semester

2021-08

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/112966

Copyright and License Information

Copyright 2021 - Anand Radhakrishnan

Owning Collections