What’s the next disruptive technology in CAE?

Great question, Jeff.

I’m thinking about this a lot with Solid Edge Simulation and I hope some CAD folks jump in.

To me FEA/CFD is a lot like CAD and engineering. If its a complex design needing a high level of accuracy, an expert with the right experience is key.

But most problems are not this way. Simple problems where only ballpark answers are needed can easily be done by non-experts.

The barrier, I think, is that non-experts may not always understand the results. If there are disruptive technologies, I can’t help thinking one will be around allowing non-fea experts to properly understand the FEA results, especially for these very common low end FEA and CFD problems.

That is an excellent point, Mark. I think we’ve learned enough to create GUIs that will inspire non-experts to setup and solve simple-but-common CAE problems… but maybe not enough attention is given to “post processing.” That term itself is born from the entrenched CAE expert world.

I’ve often thought that many non-specialists would be far more likely to use a product that is “skinned” to their exact needs. Outputs might be less about creating every wizbang color plot imaginable, and more about a Go-No-Go indicator (on the super simple side of the spectrum).

So, how do you go about creating something generic that can be skinned to that level of specificity without requiring massive amounts of consulting $$’s to customize?

Jeff,

Thank you for another interesting post!
I would have to second Mark – the main barrier is that non-experts don’t always understand CAE results.
To add to the confusion, sometimes experts don’t understand the results either. And many times they have a very hard time communicating to non-experts their understating of CAE results or lack there of.

I also think that the belief in computer-based experiments is not very strong. We are still working on convincing people that simulated experiments in many cases are as good or better then “real” ones.

So I think the answer to your question would be that an “earth shattering” force in CAE would come from the change in perception of CAE by people. And from a drastic change in communication between experts and non-experts int he field.

Not from a specific technology.

Keep on writing!
Best,
Masha

Hello Jeff. Great post! I’m sharing and discussing disruptive ideas on Daily PLM Think Tank. You are welcome to follow disruptive topic http://plmtwine.com/?s=disruptive. Best, Oleg

Good topic, Jeff. I see further migration of technology into the realm of the non-expert, where CAD, FEA, CFD and other disciplines become simply tools. Right now, we can go faster than ever from a physical object to 3D model to parametric CAD to FEA and CFD, but at the very least it takes a degree of CAD modeling skills. It will get more seamless and push-button, eliminating much of the physical-digital divide that currently exists. Ideally, you scan a physical object and automatically generate a 3D model, which is then transformed into a CAD model and meshed accurately for CFD and FEA.

Thanks Bob, I like that thread… something around scanning while prototyping. Moving CAE out to the physical lab to work in unison?

Thanks Oleg! (looks like I’ll be spending some quality time reading about PLM now… yikes!)

Here are 2 other possible idea areas: Many software categories have recently been disrupted by the OpenSource movement and SaaS cloud-based innovations. How might either of those really change the game for CAE?

While I like where this thread is going I have some dissenting views to provide…

Simplifying the geometry manipulation for FEA is not going to provide a leap/revolution. The fact is, direct geometry modelers are not really that new: Rasna Applied Mechanica (and yes, i know the name has changed now!) allowed users to move/rotate faces and have the underlying elements automatically follow (either by remeshing or just morphing the affected elements) more than 15 years ago. Similarly, preprocessors like Abaqus/CAE have provided (for quite a few years now) geometry based tools that allow fillets to be removed.

It is my experience that getting geometry into a preprocessor is really the least of your problems. And furthermore, to really understand *why* you need to or indeed can ignore certain fillets, holes etc requires a pretty deep understanding of FEA – which contradicts the user persona we started with.

While Masha correctly points out that interpreting results is still somewhat of a challenge, the real problem as I see it lies in the step in between: correctly loading/constraining/abstracting the model.

While browsing sites like Eng-tips it doesn’t take long to see how many/most users still struggle with this step. Supposedly simple tasks like constraining and loading a crankshaft leave people tied in knots. What *are* the BC’s? Is there really symmetry to exploit? I follow such threads with interest (there are couple going on right now which are covering the same old ground – bolts, bearings, applied torque, singularities and so on) and it seems we are not much better off today than 15 years ago (with minor, mostly cosmetic exceptions).

It is really disturbing to see how easy it is for users to create “results” that vary by 100%, 200% or more, just because of some different interpretations of how a model is constrained. And if this is the case for even simple models (http://www.ices.utexas.edu/research/reports/2009/0917.pdf) how can you trust the difficult stuff that we can now so easily pump into automeshers!?!

Herein lie the really difficult questions for (us) vendors to ponder: Do we *really* expect designers with no mechanics background to get *credible* solutions and design decision support from FEA? Can BCs/loads/indeed any other abstractions be made robust and easy enough to use without requiring knowledge of singularities, DOF, Saint-Venant’s Principle or other mathematical constructs?

In closing, perhaps CFD has surpassed (mechanical) FEA in ease-of-setup. Indeed I find it fascinating that magazine editorials (and now blogs) have been writing the same editorials since around 1989, promising the dawn of the age of designer FEA, but we are really not than much closer to a solution…

Every few years the old man Ivo comes out with a joke and the academicians run with it waiving at the designers crying out “you’re all wrong….”. Now do these designers care ? Looks like not !
These are the elite group of professors who believe even now that you need a Typist to write a Paper. They scribble on a paper and let the Typist do the writing, formatting, etc. And then would proof read and mark with different colored pens and hand it back to the Typist.
Should we even care to listen to them what Disruptive Technology in CAE would look like..
You need Imagination, Courage and Youthful Exuberance to talk about it.

Refer to some of the comments above – how about enough power that tight meshing without worrying about needing to remove holes or fillets is common? No worrys about symetry, partial models, just mesh it all! Then the term auto mesh could become a reality, things would be simpler and more consistent. Maybe even real-time meshing sort of like the real time photorealism that is approaching. Contraints? Take them from the assembly in context. Then Mr/Mrs designer can focus on loads and achieve perhaps more reasonable results.

That’s an interesting one Kevin. Back when I was running CFD on a 286 PC, you needed to be damn careful and efficient with mesh sizing. I’m sure most engineers today don’t need the level of mesh sizing skill I had in those days. Still need some skill, though. Since computer power grows at Mohr’s law, could be we reach a generation of supercomputers in our pockets that make that possible.

I think that with the modern co-processor, (ie: Nvidea and Matric Porcessors such as in Playstation3) that we can move to adaptive dynamic algorythms such as ACO (Ant Colony Optimization) For real time fluid flow, thermal, and stress analysis. ACO does not solve TSP exactly but it can be 99$ accurate as long as we clear the paths. Combine this is a 3d mesh that is to whatever degree of resolution you need, by using an expert system help set up the parameters, and although not an exact solver, would provide immediate feedback to the design.

The troublesome part in these modern co-prossirs is the bandwidth/memory constraint. If we were to add a solid state disk to the card to store the memory space without having to travel the system bus, we would be well on our way.

@Kimberly: Agreed, the video co-processors look like an interesting trend. I think marginal performance increases on solve time (like 50-100% faster) can't be classified as disruptive. But when you do something revolutionary and leap to 100x faster and do it on a relatively inexpensive machine… that's a big deal.

Price can certainly be disruptive, but I think it follows that same thinking. The only truly disruptive price for a $5-20k software package (in my opinion) is “Free.” (ie, advertising supported or freemium model). But, that's looking at it from the perspective of mid-large size businesses, not smaller shops.

I see the future being dominated by distributed manufacturing except for large parts. I think that it will follow the PC revolution of modularity, in which people become modules as well. Look at the transition the printing, and printing-prepress industry has gone through. The future is the individual or small shop.

Economically we migrate from the centralized big manufacturer to distributed manufacturing. From hand layouts and big printing presses, to computer layout, and print on demand.

If I were focused on this market, I would have a low cost subscription service, with an integrated Cad/CAE product with a tie in to the service houses for custom parts, and I would do something like traceparts where every order I recieved a small commision for every ordered routed through my software.

I would make the system as intuitive to use as spaceclaim, but would add an expert system for CAE with a visual programming paradym, where you can update material properties like you do with materials web, and then connect the dots and be guided by the expert system. For advanced application the expert system would recommend farming it out to expert consultants at a reasonable price for analysis.
Further I would create a social network which tied in common project management schedules and costs.

Such aproach would be making an argument to provide everyone who has ever longed for a custom product to try their hand at it. be it hobbyist or small business.

I think we will get to that point sooner than some care to think. You will have a couple of big winners and the rest of the market will be flushed out.

I think that with the modern co-processor, (ie: Nvidea and Matric Porcessors such as in Playstation3) that we can move to adaptive dynamic algorythms such as ACO (Ant Colony Optimization) For real time fluid flow, thermal, and stress analysis. ACO does not solve TSP exactly but it can be 99$ accurate as long as we clear the paths. Combine this is a 3d mesh that is to whatever degree of resolution you need, by using an expert system help set up the parameters, and although not an exact solver, would provide immediate feedback to the design.

The troublesome part in these modern co-prossirs is the bandwidth/memory constraint. If we were to add a solid state disk to the card to store the memory space without having to travel the system bus, we would be well on our way.

@Kimberly: Agreed, the video co-processors look like an interesting trend. I think marginal performance increases on solve time (like 50-100% faster) can't be classified as disruptive. But when you do something revolutionary and leap to 100x faster and do it on a relatively inexpensive machine… that's a big deal.

Price can certainly be disruptive, but I think it follows that same thinking. The only truly disruptive price for a $5-20k software package (in my opinion) is “Free.” (ie, advertising supported or freemium model). But, that's looking at it from the perspective of mid-large size businesses, not smaller shops.

I see the future being dominated by distributed manufacturing except for large parts. I think that it will follow the PC revolution of modularity, in which people become modules as well. Look at the transition the printing, and printing-prepress industry has gone through. The future is the individual or small shop.

Economically we migrate from the centralized big manufacturer to distributed manufacturing. From hand layouts and big printing presses, to computer layout, and print on demand.

If I were focused on this market, I would have a low cost subscription service, with an integrated Cad/CAE product with a tie in to the service houses for custom parts, and I would do something like traceparts where every order I recieved a small commision for every ordered routed through my software.

I would make the system as intuitive to use as spaceclaim, but would add an expert system for CAE with a visual programming paradym, where you can update material properties like you do with materials web, and then connect the dots and be guided by the expert system. For advanced application the expert system would recommend farming it out to expert consultants at a reasonable price for analysis.
Further I would create a social network which tied in common project management schedules and costs.

Such aproach would be making an argument to provide everyone who has ever longed for a custom product to try their hand at it. be it hobbyist or small business.

I think we will get to that point sooner than some care to think. You will have a couple of big winners and the rest of the market will be flushed out.

Hi Jeff,

I'm late to comment, but as a twist to the idea of having CAE more accessible to designers, perhaps the disruptive technology of the future will involve CAE products adding database and graphics capabilities that enable a unified platform for both CAD and analysis without having to transfer models between different applications which have their own separate databases. In some industries, preliminary designs are far off the mark without analysis.. analysis is often integral to the design, not some “after the fact” verification of the design.

Analysis packages have come a long way in terms of graphics and modeling user interface.. DirectX offers built-in meshing and advanced surfacing which comes for “free” to users, which could potentially include CAE software vendors. There is of course, more to it than simply adding graphics – generation of dimensioned drawings, BOM's, clash detection and other capabilities would be needed.. But I wouldn't be surprised if CAE vendors at some point go after CAD software “turf” in some industries.

I like your “CAD-ish” adjective. That adjective applies to several large companies that come to mind – Dassault, Autodesk and Bentley. These companies have a long history with and expertise in CAD design, but have decided later on to purchase analysis products as a complimentary add-on to their CAD products. These type of companies would likely squelch any such development coming from their CAE products because it would cannabilize their larger CAD user base in a big way… and besides, they think CAD-centrically. It's what they know. Many CAE products already generate certain geometry very efficiently without the help of a CAD system, but a CAD system doesn't know how to do analysis.. hence the often-problematic linking of CAD with CAE.

Pretty much anytime you transfer models between CAD and CAE programs, applications which were developed on separate platforms with different databases aimed at different users, you're going to experience some problems.. especially given the back-and-forth nature of many design processes. This can vary significantly by industry and depending on what the design objectives are.

For example, it may make sense to analyze a detailed part with bolt assembly as a simplified rigid object in some design situations, but in others not. In some cases it may make sense to analyze an entire interconnected system to get a sense as to what's going on, but in other design situations you may need only to analyze a single part or portion of the system, or check local stresses in one area of a part or piece of equipment. The experienced engineers doing the analysis are going to understand these issues much better than a CAD designer who doesn't know analysis. And again keeping in mind, analysis is often integral to the design. If CAE developers have real-world engineering design experience just as CAD developers have real-world design CAD experience, theoretically this would result in advantages for the CAE companies since they would know how to better manage and classify objects for analysis.

Contrary to what was suggested on this thread, I don't believe that you can mesh your way into better designs or useability for many/most situations, and even if you could, raw output from finite elements is often not a feasible approach for many types of designs.. Furthermore, there are design code requirements associated with many CAE analytical designs (ASME, AISC, AASHTO, etc) which dictate how stresses are to be calculated, and failure modes vary, depending on what's being designed. Fatigue may dominate some designs, buckling may dominate others, and deflection limits may govern in other designs. It depends. There are a lot of these types of issues to consider, and the issues vary significantly between industries that utilize both CAD and CAE. They also vary significantly between engineering disciplines within the same industry – structural and mechanical engineering depts. may both utilize CAE and CAD in different ways working on their own segments of a larger design.

If my prediction comes true regarding CAE companies eventually moving into CAD turf, it would likely occur on an industry-by-industry basis addressing industry specific design issues.. and if it happens, the large CAD companies won't see it coming