LIP6 2002/007
- Thesis
GranuLab : un système d'aide à la découverte scientifique appliqué à la physique des milieux granulaires - L.Breton
- 196 pages - 01/24/2002- document en - http://www.lip6.fr/lip6/reports/2002/lip6.2002.007.pdf - 2,301 Ko
- Contact : Laurent.Breton (at) nulllip6.fr, Laurent.Breton (at) nullemn.fr
- Ancien Thème : APA
- Keywords : human-computer collaboration, multi-agent based simulation, virtual laboratory, XML, granular physics
- Publisher : Ghislaine.Mary (at) nulllip6.fr
In granular physics, numerical simulations are often used to better explain the link between the microscopic interaction mechanisms (e.g. contact forces between beads) and macroscopic laws (e.g. the maximum load at the bottom of a silo). Up to now, the complexity of the numerical simulations prevented physicists to study large piles and their mechanical properties. Thus, we have first proposed a multi-agent based simulation: GranuSolve. This eco-solving method can compute piles in a static equilibrium. The complexity of our algorithm is linear with respect to the number of grains. Then, we show how relevant it is to model groups of agents for the distributed resolution of our problem. Theses performances have paved the way for the development of GranuLab, a human-computer collaboration system for scientific discovery in the domain of granular physics. To be more precisely, this is a virtual laboratory for experiments based on the solutions given by the GranuSolve component. Within this environment, a granular physic researcher can represent and memorize the scientific processes he is walking through. So, the system allows the researcher to state hypotheses and realize experiments in order to establish these hypotheses as laws. We have chosen a componential approach based on the XML language to represent and operationalize theses experiments. At first, this approach has contributed to easily achieve a virtual laboratory with the help of a limited number of elementary components. Then, we have also initiated the creation of a standardized representation of physical data related to granular piles. Finally, the system has allowed the verification of an analytical theory related to the propagative and diffusive behavior of sand piles from solutions given by the GranuSolve component.