Consumer technologies are making their way rapidly into industrial manufacturing, and vice-versa. But in this quickly evolving technology environment, manufacturers must maintain focus on solutions for the long term, says Southwest Research Institute's automation engineering manager.
Clay Flannigan is the manager for robotics and automation engineering at Southwest Research Institute, a not-for-profit research and development organization in San Antonio, Texas.
PS: For someone unfamiliar with Southwest Research Institute, how would you describe the work that you do?
CF: That’s always a difficult thing to answer because we are very diverse, but we’re characterized as an applied research and development institute. We’re nonprofit, and we’re independent of any university or government agency, so that makes us pretty unique. Our job is fundamentally to take basic research results that are coming out of universities or maybe government labs and get them into application for a wide range of commercial and government clients.
Our clients come to us usually with some problem they’re trying to solve, and they can’t just go buy a product that solves that for them. We work with our clients and develop technical approaches to solve those problems.
PS: You’re SwRI’s automation point person. What kinds of automation challenges does your team try to tackle for manufacturers?
CF: We do a lot of work in aerospace automation, and aerospace is interesting because you think that building planes and building automobiles is kind of similar in a lot of ways, but there are a lot of key differences. In particular, an automotive plant might build hundreds or thousands of roughly the same product every day, and an aircraft manufacturer’s probably doing pretty well if they’re building one a week.
The challenges from an automation perspective are pretty daunting there because you’ve got these highly variable environments that the automation systems are required to work in. You can imagine if you’ve got a machining center or a paint cell or something like that; those facilities have to accommodate a huge variety of parts and materials just to produce a single aircraft.
In one week, you might have tens of thousands of parts going through a particular piece of automation or machinery. That’s somewhat different from the automotive world, and it requires more adaptive and flexible automation solutions, and that’s really where a lot of the research challenges are today. How do you create these automation solutions that are flexible enough to accommodate those kinds of manufacturing operations?
PS: How do you see new automation applications for the large manufacturers you work with filtering out for broader use?
CF: A lot of these technologies are fundamentally cross-cutting, and we see a lot of common needs. For example, we’ve got a project starting up right now between Boeing and Caterpillar, who are producing very different products, but from an automation perspective they have a lot of the same problems, so these ideas of intelligent sensing and analyzing data to be able to describe the workplace and the work piece, and these ideas of being very adaptive in the way you develop automation. Those kinds of ideas aren’t necessarily specific to any particular process or market segment.
We developed some pretty unique 3D sensing capabilities for a painting application for aerospace where we’re dynamically sensing parts that go by on a conveyor. And we then deployed roughly similar sensing technology for a local manufacturer here in South Texas that produces barbecue wood chips – chips that you put in your smoker. They had an interesting problem in having to measure the volume of this product, because of course with a wood product, if it’s damp, it’s very dense, and if it’s not, it’s very light. You can’t correlate weight to volume. So they had to measure the volume of this product moving by on a conveyor, and it turned out that the solution that we used was roughly the same technology as what we deployed in this aerospace application for sensing aerospace components. We were then basically analyzing wood chips that were going to go into barbecues. It’s kind of cool when you see those technologies trickle down like that.
PS: How did that partnership come about?
CF: In that particular case, it’s through the manufacturing extension partnership, which is a Department of Commerce program through NIST. We run one of the regional centers here in South Texas. The MEP’s mission is to support small and medium manufacturers in technology development and operational improvements. It’s basically a sister organization to mine here at Southwest Research Institute that runs our regional office. When they identify needs of small and medium manufacturers, in some cases we can apply these technologies and basically transition some of the stuff we’re working on for big companies and apply them to small, local manufacturing needs. It’s kind of a success story I think when you think about the MEP and its role, and about how these technologies developed at high-tech manufacturers can flow down to even a small operation producing barbecue wood chips.
PS: With respect to technologies such as 3D scanning and sensing, it seems like both the tools themselves and their applications are evolving at an unprecedented pace. Do you get that sense? How do you help manufacturers address the pressure to keep pace?
CF: I think so. I think there’s been a trend in maybe the last decade or so that manufacturing is driven by the consumer market. In particular, a lot of these 3D sensing technologies are driven at the research level by things like the Microsoft Kinect, which was a gaming device. But they sold so many of them and they’re so cheap that in the research community there’s a huge uptake in using that device as a low-cost but high-performance 3D sensor.
So what’s interesting about that is now you’ve got all these academic researchers who are developing advanced algorithms using those techniques, and our job is to transition those kind of fundamental algorithms into application. That’s where we are able to leverage stuff that Microsoft developed for the gaming market, which then went into academic research, and apply the results of that research to manufacturing problems.
It’s definitely a trend, that sort of acceleration, and there’s a lot of drivers, including computing technologies and wireless technologies. The one that we find really interesting are these consumer technologies that are high-volume, low-cost, but in the end they can provide really high-performance solutions even in demanding manufacturing applications.
PS: What are some other examples that you see of that?
CF: When you think just about human interfaces on the manufacturing floor, a lot of the sort of current technologies that you see in maybe graphical user interfaces are straight out of the mid-1990s. So there’s a lot of interest in, “Hey, why can I not have the same kind of interface that I have on my mobile phone or my iPad on the manufacturing floor?” There’s a lot of intuitive interfaces and design idioms that could and should be applied to manufacturing operations.
On the robotics side, there’s almost a consumerism that’s taking hold. When you see products like Rethink Robotics’ Sawyer, these are these kinds of very friendly, human-centric robotic systems that are designed for collaborative operation. In a lot of ways, it’s not the big heavy iron (robotics) that you see coming out of traditional robot manufacturers but products you can almost imagine having in your home. They’re developed more like consumer products.
There are a lot of examples of that, but on the other hand, in the manufacturing world, you have to worry about long-term viability of these things. One of the challenges we’ve seen is as these consumer-driven products get adopted into manufacturing spaces (is that) a manufacturer may put in a production line that might need to run for 15 years. And the consumer industry is moving very fast.
If you saw a mobile phone that was five years old, it would seem ancient. But on the manufacturing floor, there’s an expectation that these technologies would need to be supported a decade later. So there’s conflict there with how viable are these consumer-driven products versus the need for long-term support? That’s one of the things we’re always cognizant of: Are these technologies viable over the long haul in manufacturing environments?
PS: How do you attempt to answer that?
CF: I don’t know that we have the full answer. Our job is to push things forward as quickly as we reasonably can. There’s no magic bullet to solve that problem.
If we’ve deployed something where we have concerns about the long-term sustainability of that technology in the manufacturing space, on the software side, we capture the configuration of that system and ensure that we have a means to support that over the long haul. That means that rather than necessarily having it connected to the Internet and downloading a new update every week, we tend to kind of clamp the development and evolution of that product and say, OK, this is the configuration that’s deployed, and now we just have to maintain that.
On the hardware side, often these consumer products are low-cost, so we end up just delivering lots of spares. We say: “Here, these things are cheap, you might as well just buy a bunch of them, put them on a shelf. It’s maybe not ideal, but at least you’ve got the support there. If there ever is an end-of-life issue, that gives you time to address it.”
Those are kind of just very pragmatic solutions to the problem, but it’s something that industrial manufacturing users have to weigh versus the benefits of adopting new technologies.
PS: What are your challenges in helping manufacturers reach their goals?
CF: Manufacturing processes have to have some level of validation and control on the process to ensure quality, and as you’re thinking about these more dynamic and adaptive systems, that kind of goes in the face of deterministic performance that traditional manufacturers are used to.
That’s a challenge, because you want these levels of flexibility, yet if you’re not careful, that can introduce variability into the manufacturing operation that’s undesirable or even not acceptable. As a manufacturer, if you’ve got a process that’s been kind of validated and proven, and you’ve got data that backs up its quality, how do you move that process forward to continuously improve when you want to ensure that you’re meeting the quality expectations of your downstream customers?
More broadly, this idea of data management is difficult. If you’ve got product configuration or process configuration data, how do you manage that in an effective way so you’re not encumbered by your processes?
It’s always a balance. You want to keep manufacturers moving forward in a broad sense but you have to do it in a controlled way that you know you’ve got consistency and quality in your processes.
I think it’s prudent that the industry and manufacturers take a conservative look at these things and don’t jump on the consumerism bandwagon where you need a new solution every year.