This subproject is one of many research subprojects utilizing theresources provided by a Center grant funded by NIH/NCRR. The subproject andinvestigator (PI) may have received primary funding from another NIH source,and thus could be represented in other CRISP entries. The institution listed isfor the Center, which is not necessarily the institution for the investigator.We are implementing a two-scale model of collagen gel mechanics, with specific application to tissue engineering. The functional length scale of the bioartificial tissue - 1-10 mm - is much larger than the structural length scale - 1-10 um. Thus, on the functional scale, the tissue appears to be continuous, whereas on the structural scale, a distinct fibril network exists (diameters 100-300 nm). To model this system, we employ volume averaging theory. Continuous equations for the averaged stress are solved via FEM on the functional scale. At each Gauss point, a discrete (network) model is used to calculate the average stress in the network, which is then fed back up to the functional scale for the calculation. The two scales are fully coupled, with displacement information passing down from functional to structural scale, and stress information passing up. The approach allows us to solve for complex mechanical phenomena and composition-function relationships directly, but it is extremely demanding computationally (a run with 300 elements takes about a CPU-week). Preliminary scaling studies on our local machines suggest that the code will scale well as long as there are significantly more Gauss points than processors (we have tested up to 32 processors). For very large runs (> 1000 elements), we will require the more powerful resources available on the Teragrid. Our code is in MPI, so it can run on virtually any platform, and since our microproblems are largely independent, we should be able to run effectively on a distributed-memory system. Our primary goal with this grant would be to demonstrate that our code can move to the Teragrid system and get accurate timing data.
Showing the most recent 10 out of 292 publications
