For the past few years, most of the simulation and analysis industry has been abuzz concerning multi-disciplinary (MD) simulations, and rightly so. MD analysis offers a far more realistic view of product performance than one-off simulations. But despite the increasing availability of MD simulation capabilities, it’s not all roses. In fact, there are some serious barriers from a modeling and management perspective. Read on and let me explain.

The Necessity of Abstractions and Simplifications

A contributing factor to the root of the problem lies in the reality that simulations often require abstractions. For example, if you are running a structural analysis on an engine block, you can remove the outer rounds on some of the outer edges. It reduces the complexity of the geometric model that needs to be meshed in a way that does not affect the simulation’s results. This case is helpful but not required, meaning that you don’t have to abstract the model in this way. But there are other cases that are absolutely necessary. A mid-plane extraction for thin-walled parts for a structural analysis is required. You can’t perform a structural analysis without it.

Multi-Disciplinary Divergent Abstractions

Abstractions and simplifications don’t just apply to structural simulation and analysis. In fact, there are different reasons for abstracting or simplifying models for just about every type of analysis.
  • Kinematic and dynamic motion studies often involve the removal of some parts and subassemblies that don’t affect the results.
  • Dynamics (natural frequencies, harmonics, etc.) involve abstractions that relate to the stiffness of the part or assembly
  • Fluid dynamics often need to be modified such that the external or internal flow areas are representative of real world scenarios.

The list goes on and on, which I won’t exhaustively detail here. But each type of simulation or analysis does have abstractions and simplifications. Some are optional and can be ignored at the penalty of performance. Others are mandatory and can’t be skipped. But none of that is at the crux of the problem, which is this:

Different types of simulations require different abstractions.

And when you are running multi-disciplinary simulations where those abstractions are tied together and affect one another, you run into some issues.

Juggling Multiple Abstractions

What issues? Well, think about it this way.

Let’s say you have a single sub-assembly that you want to run in a multi-disciplinary simulation that is composed of three different analyses. The different abstractions required by those three different analyses result in what is essentially three different model abstractions. In that scenario, a couple of questions rise:

  • How do you get the simulation software to recognize that these three models (configurations) should be used for the three different simulations? That of course is a CAE software issue.
  • Let’s say a design change is made. How do you get that change propagated to those three different models (configurations, abstractions, whatever you call them)? That is a CAD software issue.
  • What about centrally managing these three different models (four if you count the original design model)? Does PDM (or SDM/SLM) software recognize that these are functionally equivalent abstractions created purely for the purpose of simulation?

So as you can see, there are some issues even if multi-disciplinary simulations are more possible than ever before. I am sure some software providers have good answers to these questions. As I engage them in briefings, I’ll be sure to dig in and provide these answers as part of software reviews.

Conclusions and Questions

Different types of simulations have abstractions or simplifications, some optional and some required. In multi-disciplinary simulations, these abstractions are used by different analysis types and must somehow be related and connected to one another. There are outstanding questions related to how CAE, CAD and PDM software handle these interrelationships and how change is propagated.

Obviously this sort of issue isn’t new. Analysts have been managing something close to this issue for years, albeit manually and probably painfully. Feel free to sound off on the ‘old school’ way of dealing with these issues. I’d love to hear your perspective.

Take care. Talk soon. And thanks for reading.