The first Analysis, Simulation, and Systems Engineering Software Summit (ASSESS) was held in January 2015 but its backstory begins in the spring of 2014 at the colocated Collaboration and Interoperability Congress (www.3dcic.com) and NAFEMS Americas Conference 2014 (www.nafems.org/2014/americas). That’s where intrinSIM’s Joe Walsh talked about how business issues, not technology issues, are going to drive simulation forward.
The Coming Simulation Revolution
More specifically, Walsh talked about how increased competitiveness across all industries has forced business leaders (i.e. CEOs, COOs, CFOs) to focus on innovation, risk management, and cost reduction. As it turns out, simulation is the “key enabler” for each of these components of competitiveness. Not only is it a key enabler, but it’s a driver that can result in a factor of 10-100 growth in the use of simulation, not the 10-15% growth we’ve become used to – a revolution, not an evolution.
This revolution, according to Walsh, will require simulation software to become “fit for purpose, smart, integrated, and transparent.”
The ASSESS Summit
The positive reaction to Walsh’s presentation was the seed from which ASSESS grew with the aid of Cyon Research’s Brad Holtz. Forty “ambassadors” from across the world of simulation (users, software vendors, industry and financial analysts) were invited to this inaugural summit, held in January 2015. After a kickoff presentation and a plenary discussion of simulation’s many issues, the attendees were formed into working groups that collectively identified over 100 issues with the current state of simulation. In a final plenary session, all these issues were voted on to identify those deemed most important.
So, what were the top simulation issues identified by the summit? Presented in no particular order:
- Pre-CAD analysis and optimization
- Combining heterogeneous models in a systems approach
- Model fidelity and the role of unsexy stuff
- Knowledge capture & re-use
- Impact of web, cloud, and mobile
- Design centric workflows
- Ease of use and usability
- “Licensing models need to be revisited”
Unfortunately, that last issue about licensing models was not addressed at COFES. Too bad. It would’ve been an interesting discussion. Revisited why? Revisited how?
Also, the ASSESS organizers admit that their 40 ambassadors may have been too heavily weighted to the software vendor side of the business. (Although you wouldn’t expect #8 from such a group. That tells me the group was operating openly and transparently.) So the organizers plan to compile a list of the top couple dozen issues and present them to a more diverse audience to see whether the top eight change. No timeline was given for when that would be done.
Let’s now delve into the top seven. Please note that what follows is an attempt to summarize the salient points from the presentations and is often quoted word-for-word either from the presenters’ statements or their slides. Any misunderstandings, misinterpretations, or outright mistakes are mine. My thoughts are italicized.
Pre-CAD Analysis and Optimization
Or how do you simulate when there’s no CAD model?
Everyone talks about simulation being most effective very early in the design process when it’s easy to explore changes. Also, the farther upstream you swim in the design process, the more potential users you encounter. Theoretically, you could have every engineer performing simulation in this scenario, maybe by the year 2025.
For comparison’s sake, it is currently estimated that only 1 in 6 engineers use simulation which itself represents a decade’s worth of growth from when only 1 in 22 used it.
So, how do you use simulation when there isn’t a CAD model? To be very clear, there’s a difference between CAD and geometry. But the gist of the idea is to use simulation to evolve the requirements into an idealized design. In essence, this approach would accomplish something everyone says we should be doing anyway: placing function before form.
Topology optimization is an example of how this might work, to drive a shape given a set of requirements, actually mapping out the design space. That’s not to say this is a closed issue because conceptualizing function, choosing the right parameters to optimize, and ensuring manufacturability are open for research. [This sounds very similar to adaptive meshing. If you start with a grid that’s too coarse, you’ll never even start to resolve some features and therefore they won’t be adapted to.]
So what would simulation software look like for use in this pre-CAD stage?
- Lightweight, not heavy like CAD.
- Direct, both modeling and simulation.
- Extreme speed, “good enough” fidelity. [This is tricky. Lowering the fidelity requirement places serious boundaries on the range of applicability of the simulation software, boundaries that have to be communicated and hopefully enforced.]
- Optimization methods that guide design.
Taming System Complexity
We all know that systems are becoming more complex through various factors.
- Many big things
- that may be interconnected
- in a heterogeneous environment
- where the underlying mathematics are difficult and/or not well understood
- all impacted by what we don’t know and even more so by what we don’t know we don’t know.
The way we handle complexity now is simply “divide and conquer” by which we apply domain specific tools and techniques to certain aspects of a problem and connect the results in some sort of hierarchical network. We need to expand on this approach by applying the very well-known principles of systems theory in a way that:
- uses general, formal systems logic [I see this as being a big challenge for things like CFD and FEA]
- is efficient (i.e. parallelizable)
- supports verification and validation practices
- and is made practical through the use of open standards (for example, FMI). [This last issue in particular is one raised by the CFD Vision 2030 Study in the context of multi-disciplinary design optimization. I don’t recall whether that report cites FMI.]
Unsexy Stuff is Key
This presentation is the one I most wanted to hear. After all, nothing is less sexy than mesh generation. But this working group took a different approach to the topic. It’s our “culture” that’s to blame.
- We spend too much time on individual solutions instead of helping management understand the usefulness of simulation. [This statement on its surface seems to conflict with Joe Walsh’s premise that CEOs are now groking simulation. But we are at an inflection point where our upselling of simulation will fall on more receptive ears.]
- We are not good at finding out what it is that users really want to know. [The topic of understanding and managing requirements comes up again later.]
- Our adherence to standards of validation is “disastrous” and “blotchy.” [This is actually kinda sad.]
- Uncertainty quantification is non-existent which may not be that big a deal because we also don’t know the degree of fidelity required by our users. [This concern was also cited in NASA’s CFD Vision 2030 Study.]
What are the next steps toward resolving these issues?
- Create a standard for conducting simulations and ensure its adoption by simulation users and software providers. [I wonder well how NAFEMS’ Professional Simulation Engineer certification fits this need.]
- Define simulation cost and uncertainty over time to track improvement.
Simulation for Discovery and Knowledge Creation
It’s time to stop running simulations and then losing or just deleting the results. Instead, simulation can and should be used to build databases of knowledge about system performance from which we can discover engineering insights.
Web, Cloud & Mobile
There’s probably no better COFES attendee than Onshape’s Jon Hirschtick to present ASSESS’ opinions on web, cloud, and mobile (hereinafter “the cloud”) as they pertain to simulation software.
As Jon said, CAE may be the last group of people on earth debating whether the cloud is a relevant topic that we as an industry need to consider.
The cloud is not new to us in general, but is in the sense of using CAE tools that are hosted on a remote server, run in a web browser, and are accessed via mobile platforms.
The benefits – critical to the future of CAE – of the cloud are:
- Accessibility: If we are going to expand use of simulation software by 10x or more, offering a platform that’s highly accessible (no download, no install) is key. There is also a cost benefit here, because users don’t have to worry too much about what computer they have to use for simulation. Which leads to the next benefit.
- Performance: Because CPU clock speeds have flatlined, performance gains will come from massive parallelism and these performance gains are critical if people are going to be running 10x the simulations. Cloud computing means everyone doesn’t need to own their own private HPC system.
- Simplicity: This is an IT topic and pertains to downloading, installing, maintaining, updating, license managing – all of the overhead that no one likes.
- Design Exploration: To fully integrate simulation into the design process, and to do so very early in the design process to make significant impact, simulation will be used to populate and explore the design space. The compute load alone makes this task well suited for the cloud.
- Cost: Because the cloud would be used only when needed, you’d pay only when you need it rather than 24×7 like we do for installed software. Other cost savings arise from not having to maintain your own HPC system and the IT staff to maintain it and the software.
- Collaboration: Because the cloud is central, common, and easily accessible to all users it naturally enables group collaboration on projects.
Of course, the cloud does come with some baggage.
- Security: While there are legitimate security concerns about use of the cloud (e.g. military aerospace companies who simply cannot remotely host their data) many of security concerns are simply (and likely mistaken) perceptions that data on a cloud server is less secure than data on an employee’s laptop. The same laptop they’re using to browse random websites, the same laptop they take home and on business trips.
- Legacy code: Programming for cloud is an entirely new paradigm that many (most?) CAE programmers don’t currently have expertise in. Do we port our current codes or do we reprogram from scratch?
- Territoriality: Sometimes international government regulations restrict where data can be put. Plus, governments will want their cut of the action through tariffs and taxes.
- User acceptance: We will have to take into account the user community’s level of comfort with cloud-based simulation applications and the software-as-a-service (SaaS) business model. [For example, I have heard user complaints about upgrades being thrust upon them in the SaaS model whether they like them or not. And then there’s the “big brother” aspect of SaaS in which the software vendor can monitor which features are being used and at what frequency.]
Design Centered Workflow
The essential essence of this issue is one that has been around for at least a decade: geometric design should be an output of functional design rather than the reverse which is how it is now. In other words, right now we draw a shape in CAD and then simulate to see how it performs. What we should be doing is exploring the design space with simulation and seeing what geometry is produced.
The phrase “tyranny of CAD ” was used to describe the situation where manufacturing lead times dominate the schedule for product development which resequenced the flow of engineering so that simulation is done after the fact where it is less effective in achieving design goals.
How do we go about keeping design at the center of simulation?
- We must have traceable and verifiable requirements that are linked to simulation and to CAD. [See above where this is deemed one of the unsexy things.]
- We must take a hierarchical approach to having models of varying levels of fidelity.
- These models must be the cornerstones of collaboration.
- Simulation and optimization must be ubiquitous to the design process. embedded and optimizable simulation throughout meaning that the user (a designer) won’t need specialized simulation knowledge.
The rhetorical question was asked: “Why don’t we have geometry for X where X ≠ manufacturing?”
Usability goes well beyond the user interface. The pinnacle of usability may be the apps we all have on our smartphones.
- They are useful with an obvious or almost implicit ROI.
- They have a simple, intuitive interface.
- They don’t require a user manual.
- They are focused on solving a single problem.
On the other hand, the nadir of usability may be general purpose CAE software. Where does that lack of usability come from?
- Extensive functionality with more added each day.
- Significant expertise and experience are required to get reliable and accurate results.
- Users must have expertise in both the relevant physics and the simulation software.
We all appreciate the fact that physics simulations can be complex. But do they have to be complicated too?
One method for solving this problem is simulation apps (aka sim apps) in which expert knowledge is captured in a narrowly focused tool that non-experts can use via the web for easy access.
The Way Forward for Us
The general intent of ASSESS’ organizers and participants is to maintain this effort as a long-term initiative, to advocate, shepherd, monitor, cajole, support, and promote the analysis and simulation market through its anticipated and desired growth.
The manner in which this will be accomplished is currently a matter of discussion.
In the last 12 months or so I’ve seen the future of analysis and simulation addressed from both the very technical side (in the form of NASA’s CFD Vision 2030 Study) and the business side (in the case of ASSESS). Both these efforts were led by very smart people.
Like you, I have spent a lot of time thinking “What does all this mean for us? Are these issues already influencing our users? Should they influence us? And when? To what degree? And how.”
I should have some answers for you in the coming weeks.