(1/8)
Are you interested in reaction-diffusion systems? Also want to know more about coupling them with moving interface problems? Especially if it’s related to #HighPerformanceComputing and elaborated on it? Then, don't miss our recent publication😉👇
doi.org/10.1177/109434…
Are you interested in reaction-diffusion systems? Also want to know more about coupling them with moving interface problems? Especially if it’s related to #HighPerformanceComputing and elaborated on it? Then, don't miss our recent publication😉👇
doi.org/10.1177/109434…
(2/8)
Reaction-diffusion systems coupled with moving interface problems (in which the boundary of the domain is part of the solution 🙂) have great importance in various real-world scenarios in chemistry and chemical engineering as well as environmental and life sciences.
Reaction-diffusion systems coupled with moving interface problems (in which the boundary of the domain is part of the solution 🙂) have great importance in various real-world scenarios in chemistry and chemical engineering as well as environmental and life sciences.
(3/8)
As an example of such systems, we developed an #InSilico model of the #biomaterials degradation phenomena, in which the loss of material due to #corrosion (#biodegradation) leads to movement of the material-medium interface.
As an example of such systems, we developed an #InSilico model of the #biomaterials degradation phenomena, in which the loss of material due to #corrosion (#biodegradation) leads to movement of the material-medium interface.
(4/8)
In addition to a step-by-step elaboration of the employed numerical technique (#FiniteElement), we explained the details of the high-performance computing (#HPC) implementation as much as what was allowed for the length of the paper 🙃.
In addition to a step-by-step elaboration of the employed numerical technique (#FiniteElement), we explained the details of the high-performance computing (#HPC) implementation as much as what was allowed for the length of the paper 🙃.
(5/8)
We used the cool #PETSc toolkit to parallelize the simulations, allowing us to distribute the computation to hundreds or thousands of cores (if needed). As part of this, the mesh is partitioned into smaller overlapping subdomains, each of which is assigned to one CPU core.
We used the cool #PETSc toolkit to parallelize the simulations, allowing us to distribute the computation to hundreds or thousands of cores (if needed). As part of this, the mesh is partitioned into smaller overlapping subdomains, each of which is assigned to one CPU core.
(6/8)
To evaluate the performance of the implemented parallelization, we performed weak- and strong-scaling tests to measure the speedup, parallel efficiency, and linear #scalability, the result of which was appropriate and expected regarding the studied problem size.
To evaluate the performance of the implemented parallelization, we performed weak- and strong-scaling tests to measure the speedup, parallel efficiency, and linear #scalability, the result of which was appropriate and expected regarding the studied problem size.
(7/8)
Various combinations of preconditioners and iterative solvers were also tested to find the most efficient ones for solving the obtained linear system of equations. We found out that the HYPRE BoomerAMG #preconditioner and GMRES solver work quite well for our system.
Various combinations of preconditioners and iterative solvers were also tested to find the most efficient ones for solving the obtained linear system of equations. We found out that the HYPRE BoomerAMG #preconditioner and GMRES solver work quite well for our system.
(8/8)
And as always, I thank my supportive supervisor, @LiesbetGeris, for all the great help and guidance she provided me with during my PhD project and while writing this paper. 🥰
And as always, I thank my supportive supervisor, @LiesbetGeris, for all the great help and guidance she provided me with during my PhD project and while writing this paper. 🥰
@threadreaderapp unroll
• • •
Missing some Tweet in this thread? You can try to
force a refresh
