Supplementary material: Chapter 4 Matlab code

% comp2DDM_IterativeWithFluid.m
% This script couples a complementarity algorithm and the two dimensional
% displacement discontinuity method (DDM) to solve boundary value
% problems for isotropic, homogeneous, linear elastic solids containing
% fractures with user-defined frictional surface properties.
% --> Complementarity takes the form: f(Z)=M*Z+Q where f(Z)*Z=0 (e.g.
% Pang and Trinkle, 1996, Complementarity formulations and existence of
% solutions of dynamic multi-rigid-body contact problems with coulomb
% friction, Mathematical Programming, Springer).
% Complementarity script (pathlcp.m) and any other necessary files must
% reside in current directory.
% See http://pages.cs.wisc.edu/~ferris/path.html for more information.
% --> DDM is from Chap. 5 of Crouch and Starfield, 1983, Boundary Element
% Methods in Solid Mechanics, Unwin Hyman, London (abbreviated C&S).
% The number of elements making up each fracture and the element tip
% coordinates are stored in an Excel file comp2DDM_FluidTest.xls.
% Each fracture is digitized with regard to the (x,y)-global coordinate
% system such that Ne(i)+1 points define the tips of Ne(i) straight
% boundary elements making up fracture i.
% Code written by Ovunc (Uno) Mutlu (2006).
% Code modified by David D. Pollard (2007)
% Elizabeth (Libby) Ritz (2011)
% Michael Ferris & Libby Ritz (late 2012 - early 2013)

%% Clear variables, globals, functions, close figures

clear all; % Clear variables, globals, functions from memory.
close all; % Close all open figures
clc; % Clear the command window
toler = 10^-16; % Tolerance for rounding error

%% Input parameters from Excel worksheet

% Read row vector with input parameters from Excel worksheet.
Parm = xlsread('comp2DDM_FluidTest.xls','parm');
% Elastic constants: Poisson's ratio, pr, and Young's modulus, ym.
% Note: units for ym and stress components are consistent (MPa).
pr = Parm(1); ym = Parm(2);
% Three remote stresses in (x,y)-global coordinate system.
% Tension is taken as positive.
sxxr = Parm(3); syyr = Parm(4); sxyr = Parm(5);
% Number of fractures.
nf = Parm(6);
% Define the observation grid.
[XG,YG] = meshgrid(linspace(Parm(7),Parm(9),Parm(8)),...
linspace(Parm(10),Parm(12),Parm(11)));
% Read row vector with numbers of elements for each fracture.
Ne = xlsread('comp2DDM_FluidTest.xls','no');
ne = sum(Ne); % Total number of elements, including all fractures.
% Read rectangular arrays of the x- and y-coordinates of the digitized
% points along all fractures in the global coordinate system.
PX = xlsread('comp2DDM_FluidTest.xls','px');
PY = xlsread('comp2DDM_FluidTest.xls','py');
% Read rectangular array of friction coefficients for all elements.
MU = xlsread('comp2DDM_FluidTest.xls','mu');

1
% Read rectangular array(s) of frictional strength, Sf, coefficients.
SF = xlsread('comp2DDM_FluidTest.xls','Sf');
% Read value of possible normal stresses to be added along open elements.
% If negative, SnnOpen can be a proxy for fluid pressure.
snnOpen = xlsread('comp2DDM_FluidTest.xls','SnnOpen');
% Parse paramters into column vector of length ne, omitting nans.
Mu = MU(isfinite(MU));
Sf = SF(isfinite(SF));
SnnOpen = snnOpen(isfinite(snnOpen));
if ne == 2 % For 2 elements transpose Mu and Sf to column vectors.
Mu = Mu'; Sf = Sf'; SnnOpen = SnnOpen';
end

%Plot observation grid and fractures

figure, plot(XG,YG,'.k'), hold on
for i = 1:nf
plot(PX(:,i),PY(:,i))
end
axis equal, hold off
title('observation grid and fractures'), xlabel('x'), ylabel('y')

%% DDM formulation (from C&S chapter 5, p. 79-93)

% Parse coordinate data on element tips from padded arrays into single
% column vectors of beginning and ending coordinates of all elements
% in the (x,y)-global system.
% Note: the arrangement of elements into fractures is not relevant to the
% procdure for solving the boundary value problem in which each
% element is treated independently and simultaneously.
Xbeg = PX(1:Ne(1),1); Ybeg = PY(1:Ne(1),1);
Xend = PX(2:Ne(1)+1,1); Yend = PY(2:Ne(1)+1,1);
for i = 2:nf
Xbeg = [Xbeg ; PX(1:Ne(i),i)]; Ybeg = [Ybeg ; PY(1:Ne(i),i)];
Xend = [Xend ; PX(2:Ne(i)+1,i)]; Yend = [Yend ; PY(2:Ne(i)+1,i)];
end

% In what follows several square arrays are defined such that:

% the row number varies from i=1 to i=ne, and
% the column number varies from j=1 to j=ne.
% Define column vectors of projected element lengths onto the
% (x,y)-global coordinate axes.
Xpro = Xend-Xbeg; Ypro = Yend-Ybeg;
% Define column vector of actual half lengths of each element.
H = 0.5*sqrt(Xpro.*Xpro + Ypro.*Ypro);
% Define square array where each row contains all element half lengths
% ordered from j=1 to j=ne.
HJ = repmat(H',ne,1);
% Define column vector of element midpoint coordinates in the
% global (x,y)-system.
Xmid = (Xbeg + Xend)/2; Ymid = (Ybeg + Yend)/2;
% Define square array where each column contains all midpoint coordinates.
XI=repmat(Xmid,1,ne); YI=repmat(Ymid,1,ne);
% Define square array where each row contains all midpoint coordinates.
XJ=repmat(Xmid',ne,1); YJ=repmat(Ymid',ne,1);

% Calculate orientations of elements in (x,y)-global coordinate system.

% Beta is (smaller) ccw angle (in radians) between Ox and element trace.

2
% This is the convention of C&S, Figures 4.7, p. 60 and 5.6, p. 91.
Beta = atan2(Ypro,Xpro);
% Define square array where each row contains all the angles Beta.
BJ = repmat(Beta',ne,1);
% Define square array of local coordinates for the ith element midpoint
% with respect to an origin at the jth element midpoint and
% the Ox axis aligned with the jth element (Figure 4.7, C&S, p. 60).
% These are eqs. 4.6.2a and b of C&S, p. 60.
XB = (XI-XJ).*cos(BJ) + (YI-YJ).*sin(BJ);
YB = -(XI-XJ).*sin(BJ) + (YI-YJ).*cos(BJ);

% Calculate derivatives of the function f(x,y), eq. 5.2.5 of C&S, p. 81,

% which are used to calculate the stress influence coefficients.
% The following functions are used in these calculations.
% It is understood that X and Y refer to XB and YB.
Y2 = YB.^2;
XMH = XB-HJ; XPH = XB+HJ;
XMH2 = XMH.^2; XPH2 = XPH.^2;
R1S = XMH2 + Y2; R1S2 = R1S.^2;
R2S = XPH2 + Y2; R2S2 = R2S.^2;
% The shear modulus, sm, is related to the prescribed elastic constants.
sm = ym/(2*(1+pr));
% Define material constant used in calculating influence coefficients.
con = 1/(4*pi*(1-pr));
% The following derivatives are eqs. 4.5.5c and d of C&S, p. 58.
FF4 = con*(YB./R1S - YB./R2S);
FF5 = con*(XMH./R1S - XPH./R2S);
% The following derivatives are eqs. 5.5.3a and b of C&S, p. 91.
FF6 = con*((XMH2 - Y2)./R1S2 - (XPH2 - Y2)./R2S2);
FF7 = 2*con*YB.*(XMH./R1S2 - XPH./R2S2);

% Define square array with inclinations of ith element with respect to jth.
% Gamma is ccw angle between local Ox for jth and Ox for ith elements.
Gamma = repmat(Beta,1,ne) - BJ;
% Calculate trigonometric functions of the angle Gamma for each element,
% which are also used to calculate the stress influence coefficients.
SG = sin(Gamma); CG = cos(Gamma); SG2 = SG.^2;
S2G = 2*SG.*CG; C2G = CG.^2 - SG2;
% Calculate the shear and normal stress influence coefficients for
% the midpoint of the ith element with respect to the orientation of
% the ith element plane.
% These are the stress components caused by unit shear and unit normal
% displacement discontinuities of the jth element.
% Sign conventions for element displacement discontinuity and stress
% are from C&S Figure 5.4, p. 87.
% The stress influence coefficients are from eqs. 5.6.2 of C&S, p. 93.
% The element self-effects are given by eqs. 5.6.4 of C&S, p. 93.
Ass = 2*sm*(-S2G.*FF4-C2G.*FF5-YB.*(S2G.*FF6-C2G.*FF7));
Asn = 2*sm*(-YB.*(C2G.*FF6+S2G.*FF7));
Ans = 2*sm*(2*SG2.*FF4+S2G.*FF5-YB.*(C2G.*FF6+S2G.*FF7));
Ann = 2*sm*(-FF5+YB.*(S2G.*FF6-C2G.*FF7));
% Regroup the four stress influence coefficient arrays to form
% a square array of size 2*ne by 2*ne such that the shear and
% normal stress components, Ss and Sn, on the ith element are related
% to the shear and normal displacement discontinuities, Ds and Dn,
% on the jth element as given by eqs 5.4.3 of C&S, p. 88.

3
% |Ss| = |Ass Asn||Ds|
% |Sn| |Ans Ann||Dn|

% The 3 remote stress components are resolved into shear stress, Ssr,
% and normal stress, Snr, acting on the plane of each element.
% These components are referred to the element local coordinates (s,n)
% where n is directed at angle beta + pi/2 to the global Ox axis and
% s is directed to the right w/re to n (Figure 4.7, C&S, p. 60).
% Calculate trigonometric functions of the angle beta for each element.
Sinb = Ypro./(2*H); Cosb = Xpro./(2*H); % sine and cosine beta
Sbcb = Sinb.*Cosb; % sine beta times cosine beta
Sinb2 = Sinb.^2; Cosb2 = Cosb.^2; % sine and cosine beta squared
% The transformation equations are 2.8.8c and 2.8.8b from C&S, p. 24.
Ssr = -(sxxr-syyr).*Sbcb + sxyr.*(Cosb2-Sinb2);
Snr = sxxr.*Sinb2 - 2*sxyr.*Sbcb + syyr.*Cosb2;

%% Complementarity formulation
% Reformulate as a complementarity problem.
% Scale factors to improve conditioning of MCP.
scaleDs = 1e-3; scaleDn = 1e-4;
% Calculate complementary equation.
M = [speye(ne) sparse(-[scaleDn*Asn scaleDs*Ass]) sparse(ne,3*ne); ...
sparse(ne,ne) sparse(-[scaleDn*Ann scaleDs*Ans]) speye(ne)
sparse(ne,2*ne); ...
sparse(ne,2*ne) speye(ne) sparse(ne,ne) speye(ne) -speye(ne); ...
sparse(ne,ne) speye(ne) sparse(ne,4*ne); ...
-speye(ne) sparse(ne,2*ne) -spdiags(Mu,0,ne,ne) sparse(ne,2*ne); ...
speye(ne) sparse(ne,2*ne) -spdiags(Mu,0,ne,ne) sparse(ne,2*ne)];
Q = [-Ssr; -Snr; zeros(2*ne,1); Sf; Sf];
lo = [-inf*ones(4*ne,1); zeros(2*ne,1)];
up = [inf*ones(3*ne,1); zeros(ne,1); inf*ones(2*ne,1)];
Z = [Ssr; zeros(2*ne,1); Snr; zeros(2*ne,1)];

open = []; nopen = 1; % Initiate count of open elements.

while nopen > 0,
% Update bounds on sigma.
lo(open+3*ne) = SnnOpen(open);
up(open+3*ne) = SnnOpen(open);
% Update Sf and MU within Q and M matrices (eliminate frictional
% properties). May not be applicable to some values SnnOpen -
% depends on what this parameter is being used to simulate.
Q(4*ne+open) = 0; Q(5*ne+open) = 0; % For Sf
M(4*ne+open,3*ne+open) = 0; M(5*ne+open,3*ne+open) = 0; % For mu

% Run PATH.
Z=pathlcp(M,Q,lo,up,Z);

% Calculate Ss, Sn, Ds, Dn on each element: from Z.

Ss = Z(1:ne);
Dn = scaleDn*Z(ne+1:2*ne);
Ds = scaleDs*Z(2*ne+1:3*ne);
Sn = Z(3*ne+1:4*ne);

% Find number of open elements.

fopen = length(open);
% Correct for rounding errors.

4
 Dn(Dn>-toler) = 0;
open = find(Dn < 0);
nopen = length(open) - fopen;
end

% Plot displacement discontinuities on each fracture.

figure, plot(Xmid,Dn,'b.-',Xmid,Ds,'r.-');
title('displacement discontinuity'), xlabel('x (m)')
ylabel('Dn (blue) and Ds (red) (m)')
% % Plot displacement discontinuities on each element.
% figure, plot(1:ne,Dn,'b.-',1:ne,Ds,'r.-')
% title('displacement discontinuity'), xlabel('element number')
% ylabel('Dn (blue) and Ds (red) (m)')

%% Compute displacement and stress components on the observation grid.

% Initialize the displacement and stress components on the grid.
UX = zeros(size(XG)); UY = zeros(size(XG));
SXX = sxxr*ones(size(XG)); SYY = syyr*ones(size(XG));
SXY = sxyr*ones(size(XG));
% Loop over all elements, superimposing their contributions to each field.
for i = 1:ne
Dxb = Ds(i); Dyb = Dn(i);
sb = sin(Beta(i)); cb = cos(Beta(i));
s2b = sin(2*Beta(i)); c2b = cos(2*Beta(i));
% Define array of local coordinates for the observation grid relative to
% the midpoint and orientation of the ith element.
% Refer to (Figure 5.6, C&S, p. 91) and eqs. 4.5.1 of C&S, p. 57.
XB = (XG-Xmid(i))*cb + (YG-Ymid(i))*sb;
YB = -(XG-Xmid(i))*sb + (YG-Ymid(i))*cb;

% Calculate derivatives of the function f(x,y), eq. 5.2.5 of C&S, p. 81.

% which are used to calculate the displacement and stress components.
% It is understood that X and Y refer to XB and YB.
% First abbreviate repeated terms in the derivatives of f(x,y):
Y2 = YB.^2;
XMH = XB-H(i); XPH = XB+H(i);
XMH2 = XMH.^2; XPH2 = XPH.^2;
R1S = XMH2 + Y2; R1S2 = R1S.^2;
R2S = XPH2 + Y2; R2S2 = R2S.^2;
% The following derivatives are eqs. 4.5.5a thru d of C&S, p. 58.
FF2 = con*(log(sqrt(R1S)) - log(sqrt(R2S)));
FF3 = -con*(atan2(YB,XMH) - atan2(YB,XPH));
FF4 = con*(YB./R1S - YB./R2S);
FF5 = con*(XMH./R1S - XPH./R2S);
% The following derivatives are eqs. 5.5.3a and b of C&S, p. 91.
FF6 = con*((XMH2 - Y2)./R1S2 - (XPH2 - Y2)./R2S2);
FF7 = 2*con*YB.*(XMH./R1S2 - XPH./R2S2);

% Define material constants used in calculating displacements.

pr1 = 1-2*pr; pr2 = 2-2*pr;
% Calculate the displacement components using eqs. 5.5.4 of C&S, p. 91.
UX = UX + Dxb*(-pr1*sb*FF2 + pr2*cb*FF3 + YB.*(sb*FF4 - cb*FF5))...
+Dyb*(-pr1*cb*FF2 - pr2*sb*FF3 - YB.*(cb*FF4 + sb*FF5));
UY = UY + Dxb*(+pr1*cb*FF2 + pr2*sb*FF3 - YB.*(cb*FF4 + sb*FF5))...
+Dyb*(-pr1*sb*FF2 + pr2*cb*FF3 - YB.*(sb*FF4 - cb*FF5));

5
% Calculate the stress components using eqs. 5.5.5 of C&S, p. 92.
SXX = SXX + 2*sm*Dxb*(2*cb*cb*FF4 + s2b*FF5 + YB.*(c2b*FF6-s2b*FF7))...
+2*sm*Dyb*(-FF5 + YB.*(s2b*FF6 + c2b*FF7));
SYY = SYY + 2*sm*Dxb*(2*sb*sb*FF4 - s2b*FF5 - YB.*(c2b*FF6-s2b*FF7))...
+2*sm*Dyb*(-FF5 - YB.*(s2b*FF6 + c2b*FF7));
SXY = SXY + 2*sm*Dxb*(s2b*FF4 - c2b*FF5 + YB.*(s2b*FF6+c2b*FF7))...
+2*sm*Dyb*(-YB.*(c2b*FF6 - s2b*FF7));
end

%% Additional calculations
% Calculate principal stresses, S1 and S2.
% Equations from Pollard and Fletcher, 2005, Fundamentals of Structural
% Geology, Cambridge University Press (abbreviated P&F)
% P&F page 220, equations 6.72.
S1 = 0.5*(SXX+SYY)+sqrt(0.25*(SXX-SYY).^2+SXY.^2);
S2 = 0.5*(SXX+SYY)-sqrt(0.25*(SXX-SYY).^2+SXY.^2);

%% Figures
%Plot vector arrows for displacements.
figure, hold on, axis equal tight
quiver(XG,YG,UX,UY, 'linewidth', 1.2),
plot(PX,PY,'k', 'linewidth', 2)
axis([Parm(7), Parm(9), Parm(10), Parm(12)])
title('displacement field'), xlabel('x/w'), ylabel('y/w')

% % Plot the magnitude of the Cartesian stress components.

% figure, hold on, axis equal tight
% contourf(XG,YG,SXX,18), plot(PX,PY,'k', 'linewidth', 2)
% axis([Parm(7), Parm(9), Parm(10), Parm(12)])
% title('SXX (MPa)'), xlabel('x (m)'), ylabel('y (m)'), colorbar
%
% figure, hold on, axis equal tight
% contourf(XG,YG,SYY,18), plot(PX,PY,'k', 'linewidth', 2)
% axis([Parm(7), Parm(9), Parm(10), Parm(12)])
% title('SYY (MPa)'), xlabel('x (m)'), ylabel('y (m)'), colorbar
%
% figure, hold on, axis equal tight
% contourf(XG,YG,SXY,18), plot(PX,PY,'k', 'linewidth', 2)
% axis([Parm(7), Parm(9), Parm(10), Parm(12)])
% title('SXY (MPa)'), xlabel('x (m)'), ylabel('y (m)'), colorbar

% % Plot the magnitude of the principal stresses.

% figure, hold on, axis equal tight
% contourf(XG,YG,S1-SnnOpen(1,1),350), plot(PX,PY,'k', 'linewidth', 4)
% axis([Parm(7), Parm(9), Parm(10), Parm(12)])
% title('S1 (MPa)'), xlabel('x (m)'), ylabel('y (m)')
% colorbar, shading flat
% caxis([-400 100])
% box on
%
% figure, hold on, axis equal tight
% contourf(XG,YG,S2-SnnOpen(1,1),18), plot(PX,PY,'k', 'linewidth', 2)
% axis([Parm(7), Parm(9), Parm(10), Parm(12)])
% title('S2 (MPa)'), xlabel('x (m)'), ylabel('y (m)')
% colorbar