HPC APPLICATIONS: MULTIPHYSICS
➤
both large and small. They continue to offer per-core access to parallel processing, but also introduced a new concept called HPC packs, which are designed to offer virtually unlimited parallel processing of an individual simulation problem; this extreme scalability is priced to enable customers to use as much parallel processing as they need. Ansys also offers a unique product called DSO (Direct Solve Option), which is packaging for parametric modelling so you don’t have to run 10 instances of the software for 10 sweeps. ‘In the past, only the largest enterprises could afford HPC. Our goal is to enable users across the spectrum to consider more detailed, accurate and complete simulations than ever before,’ said Hutchings. Many major suppliers have been adding multiphysics capabilities, at least for the most common couplings. Algor, now part of Autodesk, ‘covers the range of standard physics, and we can couple any of them when it makes sense,’ says Bob Williams, product marketing manager for simulation for manufacturing industries. He adds that distributed computing has been a main point of emphasis at Algor for the last several years. Users work on a desktop PC to set up the problem, and the solvers can work on clusters. Some tasks, such as meshing, can today be done on multicore machines, but not in a distributed environment. He points out that his pricing structure involves no extra fee beyond the standard purchase price when working on as many as 30 nodes. He adds that there’s going to be a big push to cloud computing for two reasons. First, it removes the need for a company to invest in and maintain clusters; second, compute power becomes essentially ‘limitless’. For now, the company’s Inventor 3D mechanical simulation and Moldflow injection-moulding simulation products support this model, and he adds that this is the direction all the simulation products in the company’s portfolio will be taking. ‘Many people think of Abaqus as structural analysis software, but they don’t realise we’ve been doing multiphysics software for more than two decades,’ says Eric Weybrant, product manager for the Abaqus Analysis product line at Dassault Systemes. ‘In the late 1970s we initially developed our software based on
28
Parallel processing on computer clusters can reduce the turnaround time for large simulations such as this automotive powertrain assembly from days to just a few hours. (Image courtesy of Simulia, Dana Holding Corp. and R Systems)
customer funding, and we started with what is now known as FSI in coupling structural elements with wave loading for the offshore industry.’ Almost 20 years ago the firm added heat transfer couplings, and other long-standing capabilities include piezeoelectric effects, mass diffusion, acoustics and thermal-electrical. Through middleware, the Abaqus package can also tie into mainstream CFD codes such as those from CD-adapco and Ansys/Fluent, although recently the firm’s own code, Abaqus/CFD, came on the market and it represents ‘the largest expansion of our multiphysics capabilities in a long time’.
Middleware for other codes One piece of middleware Abaqus uses to accommodates other codes, both commercial and proprietary, is MpCCI (Mesh-based parallel Code Coupling Interface) from the Fraunhofer Institute for Algorithms and Scientific Computing (SCAI). The philosophy is that ‘there is a common feeling in the community that
SCIENTIFIC COMPUTING WORLD OCTOBER/NOVEMBER 2010
in most cases not a single (proprietary) simulation system can provide all necessary features, but that coupling the best codes of each discipline will enable more flexibility and simulation quality to the end user.’ MpCCI was developed to provide an application-independent interface for the coupling of different simulation codes, and this interface is claimed to have become a de-facto standard for doing so. However, feeling that they needed something even more flexible for general couplings, the developers at Simulia/ Abaqus have developed a direct coupling interface, which is being used by select partners, but has not been yet opened up to general release. ‘Depending on the physics,’ says Weybrant, ‘we can add special coupling algorithms or matrix operators to add stability, and having our own platform allows us to implement these algorithms more rapidly.’
As for HPC support, Abaqus takes several approaches using software written
www.scientific-computing.com ➤
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56