Page 16 - Oil and Gas Simulations
P. 16


USING HIGH                                                                                  and HPC are allowing organizations
PERFORMANCE                                                                                 to develop more accurate insights into
COMPUTING TO RUN                                                                            anticipated product behavior earlier in the
LARGE SIMULATIONS                                                                           development cycle than ever before.

High performance computing (HPC) enables the                                                èTHE VALUE OF HIGH
computation of large models by allowing design teams to                                     PERFORMANCE COMPUTING
analyze multiple design ideas simultaneously.
                                                                                            Greater computational resources can
by ALEXANDER GATZEMEIER                                                                     significantly reduce simulation time
                                                                                            where needed. An example of this can
In today’s markets, engineers are             more accurately represent the physical        be seen in the geothermal dublet model
continuously challenged to deliver            world, leading to more complex and            shown in Figure 1, where HPC was used
products that meet market demand,             larger models that move beyond single-        to reduce model computation time. For
improve operational efficiency, and           component or single-physics systems.          this benchmark simulation, additional
exceed customer expectations. In order to     This is especially true for the oil and gas   computational resources allow for more
stay ahead of the competition, simulation     industry, where large scales, nonlinear       simulation through-put within the same
is playing an increasingly important role in  material properties, extensive parametric     timeframe (see Figure 1).
product development.                          studies, and coupled physics and processes    	 HPC on cluster hardware allows for
	 During the product development              need to be efficiently analyzed.              efficient and comprehensive simulation
process, a design concept evolves from        	 Multiphysics simulations require            since it enables a greater number of
a general model to a more complete            computer hardware that scales with the        calculations to be run in the same amount
virtual prototype, making it a real-          model, reducing memory limitations. High      of time. Using HPC, engineers are able
world representation of the physical          performance computing (HPC) is key to         to analyze not only a single design
world. Throughout the design process,         enabling the use of large-scale simulations   idea, but can concurrently compare and
it is beneficial to explore various design    in this respect. The increasing availability  analyze many design variations. This helps
modifications and what-if scenarios in        of multicore processors allows for            research and design teams to improve and
order to fully understand a component or      impressive performance gains to be made       optimize their designs early in the process
problem. Growing markets put pressure         in the computation of medium to large         in a way that is more effective than
on engineers to use simulation tools to       numerical analyses. Multiphysics simulation   physical prototyping alone.
                                                                                            	 COMSOL Multiphysics® supports fully
                                                                                            parallel computing to find the solution for
                                                                                            large, complex models.

                                                                                            èPARALLEL COMPUTING

                                                                                            Parallel computing basically consists of
                                                                                            two different classical set-ups: shared
                                                                                            memory and distributed memory (see

 FIGURE 1. Left: Hydro-geothermal doublet system after 30 years of production showing the flow path and displacement. Right:
 Benchmark of deep geothermal reservoir model using the Fujitsu RTG+ solution platform (2.47MDOF, Transient, Direct Solver, 210GB).
 Red: High-end workstation (2x Intel® Xeon® E5-2643@3.30 GHz, 256GB RAM).

     comsol news industry focus
   11   12   13   14   15   16   17   18   19   20   21