Democratization of computer-aided engineering (CAE) is happening today, driven by the “appification of simulation” enabled by the rapidly declining cost of creating the underlying automation templates and web front ends for these apps. This new approach is a generational leap forward from the longstanding convention of creating simulation templates through custom scripting and programming.
A key result is automation of the exploration of a product’s potential “design space.” Far beyond just making it easier to apply CAE to narrow point goals in the design process, this app approach automates the exploration of a product’s entire design space across the full product development process.
The diagram at left is our effort to map the technologies and capabilities encompassed by design space exploration software technologies today. Of course, this is still somewhat “aspirational,” as practitioners like to say. But the past decade-plus has seen major advances in design space exploration tools in three areas: functionality, usability and off-the-shelf connectors (“integration nodes”) to most all major geometry modelers, analysis solvers and other widely used engineering calculation tools.
Goal: Automated design space exploration
Even with the many advances in recent years, design space exploration and design optimization have remained comparatively tedious manual processes. Practitioners need considerable expertise in the various tools, and in the methods that they embody. And the tools themselves remain somewhat fragmented—design of experiments, multidisciplinary optimization, multi-objective optimization, stochastic optimization, and then all the different types of structural optimization—topology, shape, size, topography, topometry, others.
The result is that users are still confronted with more or less manual exploration of these various separate aspects of the design space, and from initially uncorrelated viewpoints. And it takes considerable expert interpretation and correlation to assemble all this information into a reasonably holistic view of product performance.
Capturing all that expert knowledge in simulation apps is today making possible the leap forward to a truly automated system that anyone on a project team can use to explore the product’s potential design space not only from their own disciplinary viewpoint, but also with visibility into all other relevant product phenomena, and at any stage in the design process.
Simulation apps that implement simulation process automation, at any level of model fidelity, are the key to making all this work for both experts and non-experts. By enabling robust simulation automation across all levels of model fidelity, these new-generation apps enable everyone on the product development team to explore the design space and refine their designs at all stages of the design process.
Bringing non-experts into the concept design phase
This new approach addresses the old constraints first by allowing everyone entry into the concept engineering phase—not just the experts in systems modeling and 0D/1D simulation, say, but anyone who can contribute their discipline knowledge to help innovate at the point of product architecture definition—not just later, “when it’s their turn,” so to speak, in late preliminary or detail design, when so many key decisions have already been frozen in, and often only marginal improvements are possible.
In the concept design phase, the automation templates allow any user—expert or non-expert—to explore various architectures and to swap out entire components to find a better initial design more effectively. This makes it easy for all relevant engineering team members, not just simulation experts, to contribute their expertise.
Old, script-based automation not robust across realistic range of design changes
To be sure, automation templates for design simulation and exploration have been around a long time. They usually demoed well. But in the real world, they too often froze up, broke down or required the very manual intervention that they were supposed to avoid. They didn’t scale across a realistically wide range of product families, initial design possibilities or likely design changes, either as the design evolved or as requirements changed.
The reason was that, traditionally, those rules focused on geometry; when geometry or topology changed “too much,” the rules would fail. This new approach escapes those limits by capturing the expert rules based on the functional architecture of the product family, not on the geometry or topology of particular designs. This is how the templates are made robust across significant geometry, topology and configuration changes, and across an entire product family.
Case study 1: American Axle & Manufacturing
American Axle & Manufacturing described how simulation apps are making axle design experts’ knowledge available for use by less experienced engineers. Not only is simulation experts’ knowledge now broadly available across engineering teams, but this approach also fosters interdisciplinary understanding and “systems thinking” across project teams.
In addition, simulation is now more readily deployed into the earliest stages of product development, where it can have deep impact on innovation at the product architecture level.
Case study 2: NASA Langley Research Center
NASA Langley Research Center showed how, with traditional approaches to analysis of precision optical systems, data had to flow sequentially from engineer to engineer. In this process, known as structural-thermal-optical-performance (STOP) analysis, multiple data handoffs in initial design were time-consuming. When design changes were necessary, trade studies were time-consuming, and could be less revealing than desired.
STOP a great start
Now, with simulation apps, the knowledge of all the domain experts relevant to a project is captured in one integrated template. An expert in any domain can run multiphysics calculations across all domains, which expands the possible trade space and reduces analysis times. And it fosters multidisciplinary understanding among engineers of how each one’s discipline affects overall system performance.
In short, the traditional STOP analysis process required many expert engineers from many domains, often taking a week or more for a single analysis iteration. By contrast, the new process enabled 140 full STOP analyses in a single overnight run—an extraordinary demonstration of how simulation apps are enabling unprecedented levels of automated design space exploration with complex multiphysics calculations, including design changes.
© 2015 ENGINEERING.com. Originally published here. Reprinted with permission.