view +scheme/Hypsyst2dCurve.m @ 1037:2d7ba44340d0 feature/burgers1d

Pass scheme specific parameters as cell array. This will enabale constructDiffOps to be more general. In addition, allow for schemes returning function handles as diffOps, which is currently how non-linear schemes such as Burgers1d are implemented.
author Vidar Stiernström <vidar.stiernstrom@it.uu.se>
date Fri, 18 Jan 2019 09:02:02 +0100
parents 706d1c2b4199
children 78db023a7fe3
line wrap: on
line source

classdef Hypsyst2dCurve < scheme.Scheme
    properties
        m % Number of points in each direction, possibly a vector
        n % size of system
        h % Grid spacing
        X,Y % Values of x and y for each grid point

        J, Ji % Jacobaian and inverse Jacobian
        xi,eta
        Xi,Eta

        A,B
        X_eta, Y_eta
        X_xi,Y_xi
        order % Order accuracy for the approximation

        D % non-stabalized scheme operator
        Ahat, Bhat, E

        H % Discrete norm
        Hxii,Hetai % Kroneckerd norms in xi and eta.
        I_xi,I_eta, I_N, onesN
        e_w, e_e, e_s, e_n
        index_w, index_e,index_s,index_n
        params % Parameters for the coeficient matrice
    end


    methods
        % Solving Hyperbolic systems on the form u_t=-Au_x-Bu_y-Eu
        function obj = Hypsyst2dCurve(m, order, A, B, E, params, ti)
            default_arg('E', [])
            xilim = {0 1};
            etalim = {0 1};

            if length(m) == 1
                m = [m m];
            end
            obj.params = params;
            obj.A=A;
            obj.B=B;

            obj.Ahat=@(params,x,y,x_eta,y_eta)(A(params,x,y).*y_eta-B(params,x,y).*x_eta);
            obj.Bhat=@(params,x,y,x_xi,y_xi)(B(params,x,y).*x_xi-A(params,x,y).*y_xi);
            obj.E=@(params,x,y,~,~)E(params,x,y);

            m_xi = m(1);
            m_eta = m(2);
            m_tot=m_xi*m_eta;

            ops_xi = sbp.D2Standard(m_xi,xilim,order);
            ops_eta = sbp.D2Standard(m_eta,etalim,order);

            obj.xi = ops_xi.x;
            obj.eta = ops_eta.x;

            obj.Xi = kr(obj.xi,ones(m_eta,1));
            obj.Eta = kr(ones(m_xi,1),obj.eta);

            obj.n = length(A(obj.params,0,0));
            obj.onesN=ones(obj.n);

            obj.index_w=1:m_eta;
            obj.index_e=(m_tot-m_e

        metric_termsta+1):m_tot;
            obj.index_s=1:m_eta:(m_tot-m_eta+1);
            obj.index_n=(m_eta):m_eta:m_tot;

            I_n = eye(obj.n);
            I_xi = speye(m_xi);
            obj.I_xi = I_xi;
            I_eta = speye(m_eta);
            obj.I_eta = I_eta;

            D1_xi = kr(I_n, ops_xi.D1, I_eta);
            obj.Hxii = kr(I_n, ops_xi.HI, I_eta);
            D1_eta = kr(I_n, I_xi, ops_eta.D1);
            obj.Hetai = kr(I_n, I_xi, ops_eta.HI);

            obj.e_w = kr(I_n, ops_xi.e_l, I_eta);
            obj.e_e = kr(I_n, ops_xi.e_r, I_eta);
            obj.e_s = kr(I_n, I_xi, ops_eta.e_l);
            obj.e_n = kr(I_n, I_xi,

        metric_termsops_eta.e_r);

            [X,Y] = ti.map(obj.xi,obj.eta);

            [x_xi,x_eta] = gridDerivatives(X,ops_xi.D1,ops_eta.D1);
            [y_xi,y_eta] = gridDerivatives(Y,ops_xi.D1, ops_eta.D1);

            obj.X = reshape(X,m_tot,1);
            obj.Y = reshape(Y,m_tot,1);
            obj.X_xi = reshape(x_xi,m_tot,1);
            obj.Y_xi = reshape(y_xi,m_tot,1);
            obj.X_eta = reshape(x_eta,m_tot,1);
            obj.Y_eta = reshape(y_eta,m_tot,1);

            Ahat_evaluated = obj.evaluateCoefficientMatrix(obj.Ahat, obj.X, obj.Y,obj.X_eta,obj.Y_eta);
            Bhat_evaluated = obj.evaluateCoefficientMatrix(obj.Bhat, obj.X, obj.Y,obj.X_xi,obj.Y_xi);
            E_evaluated = obj.evaluateCoefficientMatrix(obj.E, obj.X, obj.Y,[],[]);

            obj.m = m;
            obj.h = [ops_xi.h ops_eta.h];
            obj.order = order;
            obj.J = obj.X_xi.*obj.Y_eta - obj.X_eta.*obj.Y_xi;
            obj.Ji = kr(I_n,spdiags(1./obj.J,0,m_tot,m_tot));

            obj.D = obj.Ji*(-Ahat_evaluated*D1_xi-Bhat_evaluated*D1_eta)-E_evaluated;

        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',General boundary conditions'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.
        function [closure, penalty] = boundary_condition(obj,boundary,type,L)
            default_arg('type','char');
            switch type
                case{'c','char'}
                    [closure,penalty] = boundary_condition_char(obj,boundary);
                case{'general'}
                    [closure,penalty] = boundary_condition_general(obj,boundary,L);
                otherwise
                    error('No such boundary condition')
            end
        end

        function [closure, penalty] = interface(obj, boundary, neighbour_scheme, neighbour_boundary, type)
            error('Not implemented');
        end

        function N = size(obj)
            N = obj.m;
        end

        function [ret] = evaluateCoefficientMatrix(obj, mat, X, Y,x_,y_)
            params = obj.params;

            if isa(mat,'function_handle')
                [rows,cols] = size(mat(params,0,0,0,0));
                x_ = kr(obj.onesN,x_);
                y_ = kr(obj.onesN,y_);
                matVec = mat(params,X',Y',x_',y_');
                matVec = sparse(matVec);
                side = max(length(X),length(Y));
            else
                matVec = mat;
                [rows,cols] = size(matVec);
                side = max(length(X),length(Y));
                cols = cols/side;
            end

            ret = cell(rows,cols);
            for ii = 1:rows
                for jj = 1:cols
                    ret{ii,jj} = diag(matVec(ii,(jj-1)*side+1:jj*side));
                end
            end
            ret = cell2mat(ret);
        end

        %Characteristic boundary conditions
        function [closure, penalty] = boundary_condition_char(obj,boundary)
            params = obj.params;
            X = obj.X;
            Y = obj.Y;
            xi = obj.xi;
            eta = obj.eta;

            switch boundary
                case {'w','W','west'}
                    e_ = obj.e_w;
                    mat = obj.Ahat;
                    boundPos = 'l';
                    Hi = obj.Hxii;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_w),Y(obj.index_w),obj.X_eta(obj.index_w),obj.Y_eta(obj.index_w));
                    side = max(length(eta));
                case {'e','E','east'}
                    e_ = obj.e_e;
                    mat = obj.Ahat;
                    boundPos = 'r';
                    Hi = obj.Hxii;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_e),Y(obj.index_e),obj.X_eta(obj.index_e),obj.Y_eta(obj.index_e));
                    side = max(length(eta));
                case {'s','S','south'}
                    e_ = obj.e_s;
                    mat = obj.Bhat;
                    boundPos = 'l';
                    Hi = obj.Hetai;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_s),Y(obj.index_s),obj.X_xi(obj.index_s),obj.Y_xi(obj.index_s));
                    side = max(length(xi));
                case {'n','N','north'}
                    e_ = obj.e_n;
                    mat = obj.Bhat;
                    boundPos = 'r';
                    Hi = obj.Hetai;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_n),Y(obj.index_n),obj.X_xi(obj.index_n),obj.Y_xi(obj.index_n));
                    side = max(length(xi));
            end

            pos = signVec(1);
            zeroval = signVec(2);
            neg = signVec(3);

            switch boundPos
                case {'l'}
                    tau = sparse(obj.n*side,pos);
                    Vi_plus = Vi(1:pos,:);
                    tau(1:pos,:) = -abs(D(1:pos,1:pos));
                    closure = Hi*e_*V*tau*Vi_plus*e_';
                    penalty = -Hi*e_*V*tau*Vi_plus;
                case {'r'}
                    tau = sparse(obj.n*side,neg);
                    tau((pos+zeroval)+1:obj.n*side,:) = -abs(D((pos+zeroval)+1:obj.n*side,(pos+zeroval)+1:obj.n*side));
                    Vi_minus = Vi((pos+zeroval)+1:obj.n*side,:);
                    closure = Hi*e_*V*tau*Vi_minus*e_';
                    penalty = -Hi*e_*V*tau*Vi_minus;
            end
        end


        % General boundary condition in the form Lu=g(x)
        function [closure,penalty] = boundary_condition_general(obj,boundary,L)
            params = obj.params;
            X = obj.X;
            Y = obj.Y;
            xi = obj.xi;
            eta = obj.eta;

            switch boundary
                case {'w','W','west'}
                    e_ = obj.e_w;
                    mat = obj.Ahat;
                    boundPos = 'l';
                    Hi = obj.Hxii;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_w),Y(obj.index_w),obj.X_eta(obj.index_w),obj.Y_eta(obj.index_w));

                    Ji_vec = diag(obj.Ji);
                    Ji = diag(Ji_vec(obj.index_w));
                    xi_x = Ji*obj.Y_eta(obj.index_w);
                    xi_y = -Ji*obj.X_eta(obj.index_w);
                    L = obj.evaluateCoefficientMatrix(L,xi_x,xi_y,[],[]);
                    side = max(length(eta));
                case {'e','E','east'}
                    e_ = obj.e_e;
                    mat = obj.Ahat;
                    boundPos = 'r';
                    Hi = obj.Hxii;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_e),Y(obj.index_e),obj.X_eta(obj.index_e),obj.Y_eta(obj.index_e));

                    Ji_vec = diag(obj.Ji);
                    Ji = diag(Ji_vec(obj.index_e));
                    xi_x = Ji*obj.Y_eta(obj.index_e);
                    xi_y = -Ji*obj.X_eta(obj.index_e);
                    L = obj.evaluateCoefficientMatrix(L,-xi_x,-xi_y,[],[]);
                    side = max(length(eta));
                case {'s','S','south'}
                    e_ = obj.e_s;
                    mat = obj.Bhat;
                    boundPos = 'l';
                    Hi = obj.Hetai;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_s),Y(obj.index_s),obj.X_xi(obj.index_s),obj.Y_xi(obj.index_s));

                    Ji_vec = diag(obj.Ji);
                    Ji = diag(Ji_vec(obj.index_s));
                    eta_x = Ji*obj.Y_xi(obj.index_s);
                    eta_y = -Ji*obj.X_xi(obj.index_s);
                    L = obj.evaluateCoefficientMatrix(L,eta_x,eta_y,[],[]);
                    side = max(length(xi));
                case {'n','N','north'}
                    e_ = obj.e_n;
                    mat = obj.Bhat;
                    boundPos = 'r';
                    Hi = obj.Hetai;
                    [V,Vi,D,signVec] = obj.matrixDiag(mat,X(obj.index_n),Y(obj.index_n),obj.X_xi(obj.index_n),obj.Y_xi(obj.index_n));

                    Ji_vec = diag(obj.Ji);
                    Ji = diag(Ji_vec(obj.index_n));
                    eta_x = Ji*obj.Y_xi(obj.index_n);
                    eta_y = -Ji*obj.X_xi(obj.index_n);
                    L = obj.evaluateCoefficientMatrix(L,-eta_x,-eta_y,[],[]);
                    side = max(length(xi));
            end

            pos = signVec(1);
            zeroval = signVec(2);
            neg = signVec(3);

            switch boundPos
                case {'l'}
                    tau = sparse(obj.n*side,pos);
                    Vi_plus = Vi(1:pos,:);
                    Vi_minus = Vi(pos+1:obj.n*side,:);
                    V_plus = V(:,1:pos);
                    V_minus = V(:,(pos)+1:obj.n*side);

                    tau(1:pos,:) = -abs(D(1:pos,1:pos));
                    R = -inv(L*V_plus)*(L*V_minus);
                    closure = Hi*e_*V*tau*(Vi_plus-R*Vi_minus)*e_';
                    penalty = -Hi*e_*V*tau*inv(L*V_plus)*L;
                case {'r'}
                    tau = sparse(obj.n*side,neg);
                    tau((pos+zeroval)+1:obj.n*side,:) = -abs(D((pos+zeroval)+1:obj.n*side,(pos+zeroval)+1:obj.n*side));
                    Vi_plus = Vi(1:pos,:);
                    Vi_minus = Vi((pos+zeroval)+1:obj.n*side,:);

                    V_plus = V(:,1:pos);
                    V_minus = V(:,(pos+zeroval)+1:obj.n*side);
                    R = -inv(L*V_minus)*(L*V_plus);
                    closure = Hi*e_*V*tau*(Vi_minus-R*Vi_plus)*e_';
                    penalty = -Hi*e_*V*tau*inv(L*V_minus)*L;
            end
        end

        % Function that diagonalizes a symbolic matrix A as A=V*D*Vi
        % D         is a diagonal matrix with the eigenvalues on A on the diagonal sorted by their sign
        %                                    [d+       ]
        %                               D =  [   d0    ]
        %                                    [       d-]
        % signVec   is a vector specifying the number of possitive, zero and negative eigenvalues of D
        function [V,Vi, D,signVec] = matrixDiag(obj,mat,x,y,x_,y_)
            params = obj.params;
            syms xs ys
            if(sum(abs(x_)) ~= 0)
                syms xs_
            else
                xs_ = 0;
            end

            if(sum(abs(y_))~= 0)
                syms ys_;
            else
                ys_ = 0;
            end

            [V, D] = eig(mat(params,xs,ys,xs_,ys_));
            Vi = inv(V);
            syms xs ys xs_ ys_

            xs = x;
            ys = y;
            xs_ = x_;
            ys_ = y_;

            side = max(length(x),length(y));
            Dret = zeros(obj.n,side*obj.n);
            Vret = zeros(obj.n,side*obj.n);
            Viret = zeros(obj.n,side*obj.n);
            for ii = 1:obj.n
                for jj = 1:obj.n
                    Dret(jj,(ii-1)*side+1:side*ii) = eval(D(jj,ii));
                    Vret(jj,(ii-1)*side+1:side*ii) = eval(V(jj,ii));
                    Viret(jj,(ii-1)*side+1:side*ii) = eval(Vi(jj,ii));
                end
            end

            D = sparse(Dret);
            V = sparse(Vret);
            Vi = sparse(Viret);
            V = obj.evaluateCoefficientMatrix(V,x,y,x_,y_);
            D = obj.evaluateCoefficientMatrix(D,x,y,x_,y_);
            Vi = obj.evaluateCoefficientMatrix(Vi,x,y,x_,y_);
            DD = diag(D);

            poseig = (DD>0);
            zeroeig = (DD==0);
            negeig = (DD<0);

            D = diag([DD(poseig); DD(zeroeig); DD(negeig)]);
            V = [V(:,poseig) V(:,zeroeig) V(:,negeig)];
            Vi = [Vi(poseig,:); Vi(zeroeig,:); Vi(negeig,:)];
            signVec = [sum(poseig),sum(zeroeig),sum(negeig)];
        end
    end
end