Mercurial > repos > public > sbplib
view +rv/ResidualViscosity.m @ 1012:1e437c9e5132 feature/advectionRV
Create residual viscosity package +rv and generalize the ResidualViscosity class
- Generalize residual viscosity, by passing user-defined flux and calculating the time derivative outside of the update.
- Create separate RungekuttaRV specifically using interior RV updates
- Separate the artifical dissipation operator from the scheme AdvectionRV1D so that the same scheme can be reused for creating the diff op used by the ResidualViscosity class
author | Vidar Stiernström <vidar.stiernstrom@it.uu.se> |
---|---|
date | Wed, 05 Dec 2018 13:44:10 +0100 |
parents | |
children | e547794a9407 |
line wrap: on
line source
% class describing the viscosity classdef ResidualViscosity < handle properties D % Diff op approximating the gradient of the flux f(u) waveSpeed % Wave speed at each grid point, e.g f'(u). %TBD: Better naming? Cmax % Constant controling magnitude of upwind dissipation Cres % Constant controling magnitude residual dissipation h % Length scale used for scaling the viscosity. viscosity % Stores the computed viscosity. % Convenice (for verification and plotting) TBD: Decide on if it should be kept. u_t % Stores the latest approximated time derivative of the solution. grad_f % Stores the latest approximated gradient of the flux residual % Stores the computed residual end methods % TODO: - Consider passing residual normalization as a function handle. % or choosing a type of normalization on construction. % Could for example be 1, norm((v-mean(v),inf) or normInfNeighborhood(v) % but working % - Decide on how to treat waveSpeed. It would be nice to just pass a constant value without % wrapping it in a function. function obj = ResidualViscosity(D, waveSpeed, Cmax, Cres, h, N) obj.D = D; obj.waveSpeed = waveSpeed; obj.h = h; obj.Cmax = Cmax; obj.Cres = Cres; obj.viscosity = zeros(N,1); obj.u_t = zeros(N,1); obj.grad_f = zeros(N,1); obj.residual = zeros(N,1); end function obj = update(obj, v, dvdt) obj.u_t = dvdt; obj.grad_f = obj.D(v); obj.residual = obj.u_t + obj.grad_f; obj.viscosity = min(obj.Cmax*obj.h*abs(obj.waveSpeed(v)), obj.Cres*obj.h^2*abs(obj.residual)/norm(v-mean(v),inf)); end function [residual, u_t, grad_f] = getResidual(obj) residual = obj.residual; u_t = obj.u_t; grad_f = obj.grad_f; end function viscosity = getViscosity(obj) viscosity = obj.viscosity; end end % Remove or fix. Should be able to handle values close to zero. Should work in 2d and 3d. methods (Static) function R_norm = normInfNeighborhood(v) n = length(v); R_norm = zeros(n,1); R_norm(1,1) = norm(v(1:3), inf); R_norm(n,1) = norm(v(n-3:n), inf); for i = 2:n-1 R_norm(i,1) = norm(v(i-1:i+1), inf); end end end end