Coordinator: mgr inż. Mateusz Sitko.
Numerical modelling based on the digital material representation idea, which introduces microstructure elements in a direct manner (grains, grains boundaries, inclusions, particles, phase boundaries, etc.), is being rapidly developed in the leading scientific centers. Simulations based on this approach give opportunity to analyze material behavior under conditions, which were not possible to monitor in the conventional approaches where microstructure is treated as a homogeneous material. The digital material representation concept combined with the discrete modelling techniques e.g. cellular automata (CA) provide opportunity to perform material evolution simulations during various metalforming and heat treatment operations. This kind of simulation is a powerful computational tool because it allows control of many aspects of material microstructure evolution during various metalforming processes. A lot of publications dedicated to simulation based on this approach can be found in the scientific literature, but the major problem of these models lies in the high computational cost, which is a limiting factor in practical application during designing of new metalforming technologies. Thus, presently, application of the approach at industrial scale is practically impossible and is limited only to scientific investigations. However, there are two approaches that can minimize CA model computational complexity and simulation time. First, is based on frontal cellular automata concept, which assumes that computations are realized only in selected areas of the investigated microstructure e.g. at the grain boundaries. The second approach takes advantages of modern computer centers that can additionally work in the grid environment. They are equipped with hundred and sometimes even thousands of computational units, providing enormous computational power.
In the current project authors were focused on the second approach based on the parallel and distributed computing concept to minimize computational time of the CA models. During the work series of fundamental research related to redesigning of the CA algorithms in order to take advantages of mentioned computing power were performed. Within the project new algorithms dedicated for high-performance computing including: communication and synchronization protocols both for logical and physical computing units were developed and their behavior during simulations was in details investigated (Fig. 1).
As a result of the project, parallel and distributed computations of the investigated CA microstructure evolution model can be performed on two types of computing units: multi-core processors or computing clusters working in the grid environment. of the major outcome of the project is fast and efficient cellular automata static recrystallization model that can be used as a support of experimental and industrial research. Eventually, developed computationally efficient CA model will provide in the near future possibility to support industrial processes with online simulations, performed during manufacturing cycle, and will give an immediate adjustment of process conditions to precisely control final product properties.
This research was supported in part by PLGrid Infrastructure.