Olivier Byrde

Optimization of a Kenics static mixer for non-creeping flow conditions

Computational Fluid Dynamics (CFD) has been employed for the optimization of the mixing efficiency of a Kenics static mixer. A series of numerical simulations has been undertaken for non-creeping flow conditions to determine the optimal twist angle of the mixing elements. The mixer efficiency has been assessed by considering the computed pressure drop along the mixer and the size of the fluid structures remaining at the mixer outlet. Contrary to the results of previous investigations for creeping flows, it is shown that for the present non-creeping Bow conditions, the twist angle of 180 degrees employed in the standard Kenics design is optimal. It is demonstrated that CFD provides an invaluable tool for mixer design optimization, despite the significant computational resources necessary to undertake the present study

Parallel computation and analysis of the flow in a static mixer

The use of a high-performance parallel computer system has been investigated to compute the incompressible flow and mixing efficiency of a static mixer. The flow computation is performed using a conventional Eulerian approach, by resolving the governing flow equations on a block-structured computational mesh. To examine the mixing process, a Lagrangian approach involving particle tracking is employed. The parallelization of both the flow computation and particle tracking phases of the numerical simulation, which are performed independently, is described. It is shown that a highperformance parallel computer system provides the possibility for more detailed and accurate simulations, leading to greater insights into the flow behaviour and mixing properties

Parallel multi-block computations of incompressible flows for industrial applications

Reference LIN-CONF-2007-015View record in Web of Science Record created on 2007-07-20, modified on 2016-08-08

Parallel multi-block computations of three-dimensional incompressible flows

Parallel computation has shown to provide considerable potential for the numerical simulation of complex three-dimensional flows. A number of studies have shown that CFD codes can be parallelized for efficient use on present-day parallel computer systems. However, of particular importance for industrial applications is the total time to solution, comprised not only of the resolution of the flow equations but also the pre- and post-processing phases. Results are presented of a study of the use of high-performance parallel computing to facilitate such numerical simulations. This study is being undertaken using a 256-processor Cray T3D system, within the framework of the joint Cray Research-EPFL Parallel Application Technology Program.

Parallel computation of flow in static mixers

Publisher Summary Static mixers are in-line mixing devices that consist of mixing elements inserted in a length of pipe. A variety of element designs are available from various manufacturers, with the number of elements required for a particular application being dependent on the difficulty of the mixing task (more elements are necessary for difficult tasks). The optimization of chemical mixers is traditionally performed by trial and error, with a lot depending on previous experience and wide safety margins. The ability to simulate the flow through a mixer numerically can contribute significantly to the understanding of the mixing process and provide better, faster, and cheaper design optimization. The numerical method is based on a cell-centered finite-volume discretization with an artificial compressibility method to couple the pressure and velocity fields. Parallelism is achieved by dividing the computational domain into a number of blocks (subdomains), with the flow equations being resolved in all blocks in parallel by assigning one block to each processor. Communication among processors is necessary to exchange data at the interface of neighboring blocks.

Portability issues for parallel computational fluid dynamics

Publisher Summary On the hardware level, parallel computers can be distinguished by whether the memory is shared or distributed, or both, and by their memory access rates. Within each of these basic classifications, much different specific architecture is currently available. Two basic concepts are associated with a useful definition of code portability on these different parallel computer systems. The fact that a code will run on different computer platforms does not guarantee that its performance will be satisfactory on all systems. The fact that message passing is not restricted to distributed memory systems, but also available on most shared-memory SMP systems, enhances its choice as the currently preferred parallel programming model. It can thus be concluded that communication overhead appears not to limit portability of the 2D Euler code, except when the data parallel programming model is employed. In general, synchronization overhead resulting from computation and communication imbalance among processors is often a more major source of performance limitation. The appropriate definition of acceptable performance for a particular user will thus depend on the relative importance of running the code on different computer platforms.