Mercurial > repos > public > sbplib
view +scheme/LaplaceCurvilinear.m @ 1198:2924b3a9b921 feature/d2_compatible
Add OpSet for fully compatible D2Variable, created from regular D2Variable by replacing d1 by first row of D1. Formal reduction by one order of accuracy at the boundary point.
author | Martin Almquist <malmquist@stanford.edu> |
---|---|
date | Fri, 16 Aug 2019 14:30:28 -0700 |
parents | 84200bbae101 |
children | 5ec23b9bf360 d1dad4fbfe22 |
line wrap: on
line source
classdef LaplaceCurvilinear < scheme.Scheme properties m % Number of points in each direction, possibly a vector h % Grid spacing grid order % Order accuracy for the approximation a,b % Parameters of the operator % Inner products and operators for physical coordinates D % Laplace operator H, Hi % Inner product e_w, e_e, e_s, e_n d_w, d_e, d_s, d_n % Normal derivatives at the boundary H_w, H_e, H_s, H_n % Boundary inner products Dx, Dy % Physical derivatives M % Gradient inner product % Metric coefficients J, Ji a11, a12, a22 x_u x_v y_u y_v % Inner product and operators for logical coordinates H_u, H_v % Norms in the x and y directions Hi_u, Hi_v Hu,Hv % Kroneckerd norms. 1'*Hx*v corresponds to integration in the x dir. Hiu, Hiv du_w, dv_w du_e, dv_e du_s, dv_s du_n, dv_n gamm_u, gamm_v lambda end methods % Implements a*div(b*grad(u)) as a SBP scheme % TODO: Implement proper H, it should be the real physical quadrature, the logic quadrature may be but in a separate variable (H_logic?) function obj = LaplaceCurvilinear(g ,order, a, b, opSet) default_arg('opSet',@sbp.D2Variable); default_arg('a', 1); default_arg('b', 1); if b ~=1 error('Not implemented yet') end % assert(isa(g, 'grid.Curvilinear')) if isa(a, 'function_handle') a = grid.evalOn(g, a); a = spdiag(a); end m = g.size(); m_u = m(1); m_v = m(2); m_tot = g.N(); h = g.scaling(); h_u = h(1); h_v = h(2); % 1D operators ops_u = opSet(m_u, {0, 1}, order); ops_v = opSet(m_v, {0, 1}, order); I_u = speye(m_u); I_v = speye(m_v); D1_u = ops_u.D1; D2_u = ops_u.D2; H_u = ops_u.H; Hi_u = ops_u.HI; e_l_u = ops_u.e_l; e_r_u = ops_u.e_r; d1_l_u = ops_u.d1_l; d1_r_u = ops_u.d1_r; D1_v = ops_v.D1; D2_v = ops_v.D2; H_v = ops_v.H; Hi_v = ops_v.HI; e_l_v = ops_v.e_l; e_r_v = ops_v.e_r; d1_l_v = ops_v.d1_l; d1_r_v = ops_v.d1_r; % Logical operators Du = kr(D1_u,I_v); Dv = kr(I_u,D1_v); obj.Hu = kr(H_u,I_v); obj.Hv = kr(I_u,H_v); obj.Hiu = kr(Hi_u,I_v); obj.Hiv = kr(I_u,Hi_v); e_w = kr(e_l_u,I_v); e_e = kr(e_r_u,I_v); e_s = kr(I_u,e_l_v); e_n = kr(I_u,e_r_v); obj.du_w = kr(d1_l_u,I_v); obj.dv_w = (e_w'*Dv)'; obj.du_e = kr(d1_r_u,I_v); obj.dv_e = (e_e'*Dv)'; obj.du_s = (e_s'*Du)'; obj.dv_s = kr(I_u,d1_l_v); obj.du_n = (e_n'*Du)'; obj.dv_n = kr(I_u,d1_r_v); % Metric coefficients coords = g.points(); x = coords(:,1); y = coords(:,2); x_u = Du*x; x_v = Dv*x; y_u = Du*y; y_v = Dv*y; J = x_u.*y_v - x_v.*y_u; a11 = 1./J .* (x_v.^2 + y_v.^2); a12 = -1./J .* (x_u.*x_v + y_u.*y_v); a22 = 1./J .* (x_u.^2 + y_u.^2); lambda = 1/2 * (a11 + a22 - sqrt((a11-a22).^2 + 4*a12.^2)); obj.x_u = x_u; obj.x_v = x_v; obj.y_u = y_u; obj.y_v = y_v; % Assemble full operators L_12 = spdiag(a12); Duv = Du*L_12*Dv; Dvu = Dv*L_12*Du; Duu = sparse(m_tot); Dvv = sparse(m_tot); ind = grid.funcToMatrix(g, 1:m_tot); for i = 1:m_v D = D2_u(a11(ind(:,i))); p = ind(:,i); Duu(p,p) = D; end for i = 1:m_u D = D2_v(a22(ind(i,:))); p = ind(i,:); Dvv(p,p) = D; end % Physical operators obj.J = spdiag(J); obj.Ji = spdiag(1./J); obj.D = obj.Ji*a*(Duu + Duv + Dvu + Dvv); obj.H = obj.J*kr(H_u,H_v); obj.Hi = obj.Ji*kr(Hi_u,Hi_v); obj.e_w = e_w; obj.e_e = e_e; obj.e_s = e_s; obj.e_n = e_n; %% normal derivatives I_w = ind(1,:); I_e = ind(end,:); I_s = ind(:,1); I_n = ind(:,end); a11_w = spdiag(a11(I_w)); a12_w = spdiag(a12(I_w)); a11_e = spdiag(a11(I_e)); a12_e = spdiag(a12(I_e)); a22_s = spdiag(a22(I_s)); a12_s = spdiag(a12(I_s)); a22_n = spdiag(a22(I_n)); a12_n = spdiag(a12(I_n)); s_w = sqrt((e_w'*x_v).^2 + (e_w'*y_v).^2); s_e = sqrt((e_e'*x_v).^2 + (e_e'*y_v).^2); s_s = sqrt((e_s'*x_u).^2 + (e_s'*y_u).^2); s_n = sqrt((e_n'*x_u).^2 + (e_n'*y_u).^2); obj.d_w = -1*(spdiag(1./s_w)*(a11_w*obj.du_w' + a12_w*obj.dv_w'))'; obj.d_e = (spdiag(1./s_e)*(a11_e*obj.du_e' + a12_e*obj.dv_e'))'; obj.d_s = -1*(spdiag(1./s_s)*(a22_s*obj.dv_s' + a12_s*obj.du_s'))'; obj.d_n = (spdiag(1./s_n)*(a22_n*obj.dv_n' + a12_n*obj.du_n'))'; obj.Dx = spdiag( y_v./J)*Du + spdiag(-y_u./J)*Dv; obj.Dy = spdiag(-x_v./J)*Du + spdiag( x_u./J)*Dv; %% Boundary inner products obj.H_w = H_v*spdiag(s_w); obj.H_e = H_v*spdiag(s_e); obj.H_s = H_u*spdiag(s_s); obj.H_n = H_u*spdiag(s_n); % Misc. obj.m = m; obj.h = [h_u h_v]; obj.order = order; obj.grid = g; obj.a = a; obj.b = b; obj.a11 = a11; obj.a12 = a12; obj.a22 = a22; obj.lambda = lambda; obj.gamm_u = h_u*ops_u.borrowing.M.d1; obj.gamm_v = h_v*ops_v.borrowing.M.d1; end % Closure functions return the opertors applied to the own doamin to close the boundary % Penalty functions return the opertors to force the solution. In the case of an interface it returns the operator applied to the other doamin. % boundary is a string specifying the boundary e.g. 'l','r' or 'e','w','n','s'. % type is a string specifying the type of boundary condition if there are several. % data is a function returning the data that should be applied at the boundary. % neighbour_scheme is an instance of Scheme that should be interfaced to. % neighbour_boundary is a string specifying which boundary to interface to. function [closure, penalty] = boundary_condition(obj, boundary, type, parameter) default_arg('type','neumann'); default_arg('parameter', []); e = obj.getBoundaryOperator('e', boundary); d = obj.getBoundaryOperator('d', boundary); H_b = obj.getBoundaryQuadrature(boundary); gamm = obj.getBoundaryBorrowing(boundary); switch type % Dirichlet boundary condition case {'D','d','dirichlet'} tuning = 1.2; % tuning = 20.2; b1 = gamm*obj.lambda./obj.a11.^2; b2 = gamm*obj.lambda./obj.a22.^2; tau1 = tuning * spdiag(-1./b1 - 1./b2); tau2 = 1; tau = (tau1*e + tau2*d)*H_b; closure = obj.a*obj.Hi*tau*e'; penalty = -obj.a*obj.Hi*tau; % Neumann boundary condition case {'N','n','neumann'} tau1 = -1; tau2 = 0; tau = (tau1*e + tau2*d)*H_b; closure = obj.a*obj.Hi*tau*d'; penalty = -obj.a*obj.Hi*tau; % Unknown, boundary condition otherwise error('No such boundary condition: type = %s',type); end end % type Struct that specifies the interface coupling. % Fields: % -- tuning: penalty strength, defaults to 1.2 % -- interpolation: type of interpolation, default 'none' function [closure, penalty] = interface(obj,boundary,neighbour_scheme,neighbour_boundary,type) defaultType.tuning = 1.2; defaultType.interpolation = 'none'; default_struct('type', defaultType); switch type.interpolation case {'none', ''} [closure, penalty] = interfaceStandard(obj,boundary,neighbour_scheme,neighbour_boundary,type); case {'op','OP'} [closure, penalty] = interfaceNonConforming(obj,boundary,neighbour_scheme,neighbour_boundary,type); otherwise error('Unknown type of interpolation: %s ', type.interpolation); end end function [closure, penalty] = interfaceStandard(obj,boundary,neighbour_scheme,neighbour_boundary,type) tuning = type.tuning; % u denotes the solution in the own domain % v denotes the solution in the neighbour domain e_u = obj.getBoundaryOperator('e', boundary); d_u = obj.getBoundaryOperator('d', boundary); H_b_u = obj.getBoundaryQuadrature(boundary); I_u = obj.getBoundaryIndices(boundary); gamm_u = obj.getBoundaryBorrowing(boundary); e_v = neighbour_scheme.getBoundaryOperator('e', neighbour_boundary); d_v = neighbour_scheme.getBoundaryOperator('d', neighbour_boundary); H_b_v = neighbour_scheme.getBoundaryQuadrature(neighbour_boundary); I_v = neighbour_scheme.getBoundaryIndices(neighbour_boundary); gamm_v = neighbour_scheme.getBoundaryBorrowing(neighbour_boundary); u = obj; v = neighbour_scheme; b1_u = gamm_u*u.lambda(I_u)./u.a11(I_u).^2; b2_u = gamm_u*u.lambda(I_u)./u.a22(I_u).^2; b1_v = gamm_v*v.lambda(I_v)./v.a11(I_v).^2; b2_v = gamm_v*v.lambda(I_v)./v.a22(I_v).^2; tau1 = -1./(4*b1_u) -1./(4*b1_v) -1./(4*b2_u) -1./(4*b2_v); tau1 = tuning * spdiag(tau1); tau2 = 1/2; sig1 = -1/2; sig2 = 0; tau = (e_u*tau1 + tau2*d_u)*H_b_u; sig = (sig1*e_u + sig2*d_u)*H_b_u; closure = obj.a*obj.Hi*( tau*e_u' + sig*d_u'); penalty = obj.a*obj.Hi*(-tau*e_v' + sig*d_v'); end function [closure, penalty] = interfaceNonConforming(obj,boundary,neighbour_scheme,neighbour_boundary,type) % TODO: Make this work for curvilinear grids warning('LaplaceCurvilinear: Non-conforming grid interpolation only works for Cartesian grids.'); % User can request special interpolation operators by specifying type.interpOpSet default_field(type, 'interpOpSet', @sbp.InterpOpsOP); interpOpSet = type.interpOpSet; tuning = type.tuning; % u denotes the solution in the own domain % v denotes the solution in the neighbour domain e_u = obj.getBoundaryOperator('e', boundary); d_u = obj.getBoundaryOperator('d', boundary); H_b_u = obj.getBoundaryQuadrature(boundary); I_u = obj.getBoundaryIndices(boundary); gamm_u = obj.getBoundaryBorrowing(boundary); e_v = neighbour_scheme.getBoundaryOperator('e', neighbour_boundary); d_v = neighbour_scheme.getBoundaryOperator('d', neighbour_boundary); H_b_v = neighbour_scheme.getBoundaryQuadrature(neighbour_boundary); I_v = neighbour_scheme.getBoundaryIndices(neighbour_boundary); gamm_v = neighbour_scheme.getBoundaryBorrowing(neighbour_boundary); % Find the number of grid points along the interface m_u = size(e_u, 2); m_v = size(e_v, 2); Hi = obj.Hi; a = obj.a; u = obj; v = neighbour_scheme; b1_u = gamm_u*u.lambda(I_u)./u.a11(I_u).^2; b2_u = gamm_u*u.lambda(I_u)./u.a22(I_u).^2; b1_v = gamm_v*v.lambda(I_v)./v.a11(I_v).^2; b2_v = gamm_v*v.lambda(I_v)./v.a22(I_v).^2; tau_u = -1./(4*b1_u) -1./(4*b2_u); tau_v = -1./(4*b1_v) -1./(4*b2_v); tau_u = tuning * spdiag(tau_u); tau_v = tuning * spdiag(tau_v); beta_u = tau_v; % Build interpolation operators intOps = interpOpSet(m_u, m_v, obj.order, neighbour_scheme.order); Iu2v = intOps.Iu2v; Iv2u = intOps.Iv2u; closure = a*Hi*e_u*tau_u*H_b_u*e_u' + ... a*Hi*e_u*H_b_u*Iv2u.bad*beta_u*Iu2v.good*e_u' + ... a*1/2*Hi*d_u*H_b_u*e_u' + ... -a*1/2*Hi*e_u*H_b_u*d_u'; penalty = -a*Hi*e_u*tau_u*H_b_u*Iv2u.good*e_v' + ... -a*Hi*e_u*H_b_u*Iv2u.bad*beta_u*e_v' + ... -a*1/2*Hi*d_u*H_b_u*Iv2u.good*e_v' + ... -a*1/2*Hi*e_u*H_b_u*Iv2u.bad*d_v'; end % Returns the boundary operator op for the boundary specified by the string boundary. % op -- string % boundary -- string function o = getBoundaryOperator(obj, op, boundary) assertIsMember(op, {'e', 'd'}) assertIsMember(boundary, {'w', 'e', 's', 'n'}) o = obj.([op, '_', boundary]); end % Returns square boundary quadrature matrix, of dimension % corresponding to the number of boundary points % % boundary -- string function H_b = getBoundaryQuadrature(obj, boundary) assertIsMember(boundary, {'w', 'e', 's', 'n'}) H_b = obj.(['H_', boundary]); end % Returns the indices of the boundary points in the grid matrix % boundary -- string function I = getBoundaryIndices(obj, boundary) assertIsMember(boundary, {'w', 'e', 's', 'n'}) ind = grid.funcToMatrix(obj.grid, 1:prod(obj.m)); switch boundary case 'w' I = ind(1,:); case 'e' I = ind(end,:); case 's' I = ind(:,1)'; case 'n' I = ind(:,end)'; end end % Returns borrowing constant gamma % boundary -- string function gamm = getBoundaryBorrowing(obj, boundary) assertIsMember(boundary, {'w', 'e', 's', 'n'}) switch boundary case {'w','e'} gamm = obj.gamm_u; case {'s','n'} gamm = obj.gamm_v; end end function N = size(obj) N = prod(obj.m); end end end