Single Matrix Solution and Multi-level Smart Multiphysics
Five Questions for Detlef Schneider, Senior Vice President – Solver Products, Altair
During the development process of highly engineered products, CAE can help in solving a broad range of problems and making the right decisions. This can be done at a global systems level down to a component level. In most of the cases, engineers look at single physical phenomena individually, and that’s often sufficient to proceed with good confidence in the process. However, because demands on product performance are becoming more and more challenging, while lead time is being shortened, in an increasing number of cases a product needs to be developed with a more holistic approach, considering all the relevant physical effects at the same time.
Altair has developed a solution that allows engineers to obtain the efficiency and accuracy typical of single-phenomena analysis in the world of multiphysics simulation. In this interview, Detlef Schneider, Altair’s senior vice president for solver products, talks about Altair’s “Smart Multiphysics™” approach and how HyperWorks’ comprehensive solvers platform can resolve product development issues.
It seems that the multiphysics industry has come a long way in the past five years, e.g., handling models between analysis types without resorting to a neutral format, or offering automated adaptive meshing. What are some other multiphysics analysis capabilities that were difficult or time-consuming a few years ago that are now more straightforward?
Lots of progress has been made on true multiphysics simulation platforms, such as HyperWorks, that can serve different physical disciplines in one environment. Having one platform allows you to model, run, visualize and optimize different disciplines seamlessly. On the analysis/solver side, there was and is a big emphasis on developing robust and scalable solutions. Only when single-discipline solutions can run robustly, with high performance and with high-quality results on detailed models, can true multiphysics deliver the necessary impact in the development process. So, basically, by continuously enhancing single-physics components, having them “talk” to each other and including the use of optimization technology, multiphysics simulation delivers high value for the development process.
Mathematically coupled multiphysics solutions solve all of the relevant equations in a single matrix that includes all of the DOF parameters, converging to a single solution. Are there still cases where it’s mathematically easier to solve two separate matrices, handing the results from one to the other?
Putting all equations in one single matrix and coupling them means compromising on many ends. Useful multiphysics analysis only needs to exchange the relevant data, which depends on what the actual engineering problem is that needs to be solved. To get high performance as well as high quality from multiphysics simulation, each discipline involved needs to deliver high performance and high quality on its own. Altair is focusing on getting high-quality results with scalable solutions on single physics and, at the same time, making sure that the relevant data to solve the multiphysics problem is seamlessly exchanged during the simulation run. This could be addressed by applying strong or weakly coupled interaction or by mapping data from one discipline to another, such as using manufacturing simulation results as an input for structural analysis.
Also, having a coupling strategy rather than integrating in one matrix allows you to implement third- party tools into the process. An open architecture has always been a key part of Altair’s strategy.
Mathematically, single monolithic code is more stable for solving highly coupled problems. However, this comes with the disadvantage of problem size, as well as an optimum solution for each physics, making the method impractical for large industry problems. On the other hand, weakly coupled interaction seems less stable in theory. But in practice, we have developed technologies that give us the stability of the strongly coupled code, without its drawbacks. In addition, having close collaboration between CFD, structural and MBD developers (including having access to each other’s source code) has added a measurable advantage to our solutions.
Meshing in the different physics realms is a key step. What are some features in Altair’s multiphysics software that help users manage this task, especially when more than two physics types are involved (e.g., can you do FSI plus heat)?
Consider the Altair CFD solution AcuSolve, which is based on a unique finite element approach that, unlike other CFD finite elements codes, fully preserves all the physical quantities, producing highly accurate results. The finite element approach provides two advantages when it comes to multiphysics: 1) the solution is insensitive with respect to meshes, which makes meshing much easier, and 2) coupling finite element CFD code with a finite element structural code is quite straightforward. This allows you to solve, for example, a structure + heat transfer + flow problem.
Moreover, with advanced meshing technologies, including batch meshing, the time for the meshing task has been reduced to a minor part of the simulation process.
What guidance do you offer to help users determine/judge when various levels of deeper multiphysics are required (maybe with very nonlinear materials, such as particles suspended in fluids, or very fast changes in conditions, such as with deep-forming of metals)? What kind of questions can help determine the tradeoffs for choosing a strongly coupled or weakly coupled situation, if they are looking for a simpler (faster?) solution?
From a technology point of view, Altair provides different approaches to solving multiphysics problems. The right approach depends highly on the physics, which need to be understood by the users. However, through the Altair Support and the Altair ProductDesign organizations, we can provide high-level knowledge to make sure that customers use the best available technology. Once the best available technology is determined, a typical approach is to capture the process in a workflow, which can be exercised for other parts or by other people later on. The highly customizable HyperWorks environment allows you to provide easy access to complex processes.
What are some of your more esoteric physics capabilities (e.g., can you handle smoke flow)?
HyperWorks provides a generic solver environment with different numerical capabilities, such as structural FE (explicit, implicit), CFD, thermal, or multi-body dynamics, supported by numerical optimization technology. Lots of development in the past few years has been about simulation of composite materials, which requires new material models, etc. Recent developments with extended finite elements (X-FEM) allow you to do simulations, such as windshield cracking or rupture simulation, more precisely. We also can simulate smoke flow.
Is there any additional question you’d like to ask? Please submit your comment in the form below. Looking forward to reading your opinion on the subject.