Blog: Artificial Intelligence Comes to CAD – Advanced Manufacturing
The bane of modern engineering is complexity. One promise of artificial intelligence and machine learning is helping engineers to use complex tools and harness vast data sets effectively. While this makes sense for areas like machine vision, I wonder if it applies to other disciplines. Like CAD for instance. As CAD and CAE suppliers give engineers more computer tools to do their jobs more efficiently, the quantity of such tools is itself becoming a problem. Could AI or machine learning help?
Siemens PLM Software, Plano, Texas, thinks so. In a press release dated February 19, 2019, Siemens announced its latest release of NX CAD software with new capabilities that “use machine learning and artificial intelligence technology to create user interfaces.”
Tackling the problem of complexity in the user interface in CAD systems is not unique to Siemens. “The problem CAD has today is that it offers too many commands for users. There is [a] statistic that 90 percent of the time we use 10 percent of the commands. For myself, I find that is true,” explained Ralph Grabowski, editor of upFront.eZine, an online magazine dedicated to CAD and the author of 200 books about computer-aided design. “The idea behind a dynamic UI is that it exposes additional commands at the right time, ones that users probably don’t know exist,” he said.
But why a dynamic user interface, and how is creating one AI? Dynamic interfaces change as the user works day in and day out, as opposed to a static one tailored for a particular use. That is where AI comes in; it collects data and then predicts what UI is best for the context and the user, changing as needed.
The key is in defining workflows, and who defines them. “Our customers have always looked to simplify their processes,” explained Paul Brown, manager, product software for Siemens PLM Software. “We give our customers the ability to [statically] tailor the interface.” Static workflows need an expert predicting what are going to be the best commands in the right sequence for, say, a sheet metal or powertrain part.
“What we are doing with machine learning is constantly monitoring and tracking what users are doing,” Brown continued. “[It’s] software that evaluates and learns. When it finds that a particular set of commands are used in a particular way, it adapts the UI. It revamps the toolbar and brings up the command that is most common after any command is invoked.” He stressed that this is not just in CAD; every application inside the NX environment and in the SimCenter 3D environment will use this capability.
Are there cautions to note in this? Grabowski thinks so. “The drawback is that it changes the UI, which makes users less efficient. Users are most efficient when they don’t have to think about the interface, instinctively executing commands, much like a touch typist does on a keyboard. When the UI morphs, the instinctiveness is lost—until you get used to the changing UI,” he said.
That is always the process with new technology—discovering the drawbacks and finding out if they are worth it when compared to the benefits. AI will be no different as it moves into new areas. Like CAD.