Aspects of Parallelization of Multiphase Computational Fluid Dynamics System Models of a Textile Suction Device

摘要

Textile industries are among the largest industries in a world wide context. Therefore increasing the efficient energy usage has a big impact globally. During textile wet finishing, for example scouring, bleaching, washing, desizing or colouring, suction devices are integrated for removing contaminants and their energy consumption is of major concern. Computational Fluid Dynamics (CFD) system models and simulations can be used to lower the energy consumption of textile suction devices. In this paper we give an introduction into CFD simulations for the dewatering of a technical fabric and discuss the computing times and scaling on different Cluster. The CFD-simulations were performed with the commercial code ANSYS Fluent. We explain shortly the 2D system model consisting of the solution domain of the suction slot in a suction pipe together with a moving mesh for the textile fabric. We used the k-ω turbulence model and the VOF multiphase model for the two phase flow of water and air. A complicated multiphase system model like the one utilized in this investigation poses severe problems with regards to computing times. This issue limits the implementation of CFD-simulations in industrial optimization processes. With two compute clusters, different versions of the commercial code Fluent and varying number of compute cores the computing times for the described system model were explored. We discuss the scaling of each configuration in computing time and the total memory needed as a function of the number of processes. Furthermore we look at the effect of the usage of different numbers of memory channels per processor as well as the memory speed on the computing time.