Hello Dumuxers,
We found a crucial bug in the mpnc diffusion model in
dumux/implicit/mpnc/diffusion/fluxvariables.hh which will cause memory leaks if
a fluidsystem with more than 4 components is used in the 2D case. These are the
pieces of code which are relevant:
>From line 81:
const unsigned int i = face.i;
const unsigned int j = face.j;
from line 119:
// initialize the diffusion coefficients to zero
for (int i = 0; i < numComponents; ++i) {
porousDiffCoeffL_[i] = 0.0;
for (int j = 0; j < numComponents; ++j)
porousDiffCoeffG_[i][j] = 0.0;
}
from line 150:
if (phaseIdx == wPhaseIdx) {
// Liquid phase diffusion coefficients in the porous medium
for (int i = 0; i < numComponents; ++i) {
// -> arithmetic mean
porousDiffCoeffL_[i]
= 1./2*(red_i * elemVolVars[i].diffCoeff(wPhaseIdx, 0,
i) +
red_j * elemVolVars[j].diffCoeff(wPhaseIdx, 0,
i));
}
}
else {
// Gas phase diffusion coefficients in the porous medium
for (int i = 0; i < numComponents; ++i) {
for (int j = 0; j < numComponents; ++j) {
// -> arithmetic mean
porousDiffCoeffG_[i][j]
= 1./2*(red_i * elemVolVars[i].diffCoeff(nPhaseIdx,
i, j) +
red_j * elemVolVars[j].diffCoeff(nPhaseIdx,
i, j));
}
}
}
The problem here is that in each loop the variables for the i and j for the
scv corners (?) are reused as component indices. So if the number of components
exceeds the number of corners, elemVolVars[i] and elemVolVars[j] will not be
defined and valgrind will complain. Therefore, we suggest to follow the naming
convention used in dumux/implicit/mpnc/diffusion/volumevariables.hh. i should
be renamed to compIIdx and j to compJIdx. Note that the indices of
elemVolVars[i] and elemVolVars[j] stay unchanged. Here is the code:
// initialize the diffusion coefficients to zero
for (int compIIdx = 0; compIIdx < numComponents; ++compIIdx) {
porousDiffCoeffL_[compIIdx] = 0.0;
for (int compJIdx = 0; compJIdx < numComponents; ++compJIdx)
porousDiffCoeffG_[compIIdx][compJIdx] = 0.0;
}
if (phaseIdx == wPhaseIdx) {
// Liquid phase diffusion coefficients in the porous medium
for (int compIIdx = 0; compIIdx < numComponents; ++compIIdx) {
// -> arithmetic mean
porousDiffCoeffL_[compIIdx]
= 1./2*(red_i * elemVolVars[i].diffCoeff(wPhaseIdx, 0,
compIIdx) +
red_j * elemVolVars[j].diffCoeff(wPhaseIdx, 0,
compIIdx));
}
}
else {
// Gas phase diffusion coefficients in the porous medium
for (int compIIdx = 0; compIIdx < numComponents; ++compIIdx) {
for (int compJIdx = 0; compJIdx < numComponents;
++compJIdx) {
// -> arithmetic mean
porousDiffCoeffG_[compIIdx][compJIdx]
= 1./2*(red_i * elemVolVars[i].diffCoeff(nPhaseIdx,
compIIdx, compJIdx) +
red_j * elemVolVars[j].diffCoeff(nPhaseIdx,
compIIdx, compJIdx));
}
}
}
Could someone please check and commit the code?
Best regards
Georg Futter
--------------------------
German Aerospace Center (DLR)
Institute of Engineering Thermodynamics | Computational Electrochemistry |
Pfaffenwaldring 38-40 | 70569 Stuttgart
Dipl.-Ing. Georg Futter | Ph.D. student
Telefon 0711/6862-8135 | [email protected]<mailto:[email protected]>
www.DLR.de<http://www.dlr.de/>
_______________________________________________
Dumux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
