Scanning a Jet Wing in Under Four Minutes
Ian Wright posted on August 11, 2017 | 1975 views
(Image courtesy of NIAR.)
(Image courtesy of NIAR.)
There are lots of pizza places that promise to deliver your order in 30 minutes or it's free, but if they advertised four-minute deliveries, you’d probably be sceptical. Now imagine if someone offered to scan an airplane wing in the same timeframe.

Crazy, right?

Not if you’re working at the National Institute for Aviation Research (NIAR) at Wichita State University. had the opportunity to speak with Brian Brown, director of the robotics and automation lab at NIAR, who presented the results of his team’s efforts to scan a large part quickly at this year’s HxGN LIVE.

Can you explain the project and the challenges you faced?

The story of the wing scanning project started when Hexagon [Manufacturing Intelligence] said to us that they wanted to be able to show customers that you can scan big parts quickly, efficiently and with accuracy and precision, because in aerospace, they can be reluctant to make a big investment into something before they know that it’s going to work.

It’s one thing for a salesperson to tell you how great something is, but it’s another to turn the wheel and take a test drive for yourself, so to speak. So, what we do a lot is give people the ability to take a test drive.

So, we salvaged a wing from our full-scale structural testing facility that had gone through testing and was essentially a scrap part. We didn’t have nominal data and that was part of the challenge, since we couldn’t show the customer the deviation it had been produced to—plus, it had been damaged already.

In any case, we wanted to scan the wing as quickly and accurately as we could so that we could show the customer that it wouldn’t take a hundred hours to program and six weeks to get set up. It took us the better part of a day to get a fixture set up for the wing and that was probably the longest part of the project. After that, we took a scanner and a robot and it was just a matter of getting things put together, documenting as we went.

(Image courtesy of NIAR.)
(Image courtesy of NIAR.)
We treated it as though we were painting it—because scanning is very similar to a painting process—and timed everything and found it took us about two and a half hours to get it programmed. We started off slow, so it took us about 10 or 20 minutes to scan it and then we went to the customer to find out how dense they wanted the data.

Based on what they told us, we started speeding it up. So, it became a matter of, “How fast do you want this to go?” Yes, we were able to do it in three and a half minutes, but we could have done it in one if we really wanted to push it. Keep in mind: these robots can run up to seven meters a second.

So, our challenge was to figure out what it takes to go from nothing to something, and in addition, we didn’t want to use any customized programming. We said, “Let’s just assume that we’re this customer, and we have a robot and tracker. What does it take?” So, we started from a very realistic standpoint in terms of what we had to work with and what we needed to make it work.

It was a very open-ended project, in the sense that they didn’t give us many requirements. It was more about figuring out what can be done. I’m sure they had specific requirements, but we just wanted to be able to scan the entire surface of the wing and get some details for certain features. That was where we started from and it took us about two hours to program.

The other big thing we did was when we asked them what information they wanted to get off the wing and they talked about fastener flushness, for example, we showed them that all we had to do is slow the robot down or change one littler parameter and you can get that data.

So, they walked in looking for one thing and walked away with a whole bunch of ideas of other things they could do with it—all within one setup.


Your organization is industry-driven. How does that affect your approach to research?

NIAR is funded almost exclusively by industry and government research, so it’s not a lot of fundamental or theoretical research; it’s all industry-based. When we look at a problem or situation, we try to think “How can this be applied in the factory tomorrow?” That drives us toward solutions that are already out there on the market and which are being produced by reputable companies.

So, rather than trying to come up with a new scanner or new scanning method, we’d rather source solutions from companies like Hexagon. Then, we’ll help integrate these solutions and try to help think outside the box about how we can do that. That way we can say to our customers, “Look, this is already here, it works and it’s been proven and you’ve got companies like Hexagon or ABB that are supporting it.”

(Image courtesy of NIAR.)
(Image courtesy of NIAR.)
So, it’s not like you have to worry about Wichita State supporting your solution for the next 20 years because there’s a multi-billion-dollar company that’s behind it. When we approach a problem, we look at how we’d be able to use it tomorrow. If we come up with a solution that can’t be implemented for another five years, then that’s a failure to us. These companies have issues and they need solutions today. They might be willing to wait until tomorrow, but they need it today.

My salary, my staff and my lab are entirely funded by these industry projects, so we operate almost more like a business than an academic department. If my lab doesn’t recover costs, then it will close - either because I’m not tending the business properly or because it’s no longer a field of relevant interest.

We’ve had that before, labs in NIAR that we started for something we thought was going to be the next big thing, but it just doesn’t get utilized the way we thought, so we end up converting it into something else.

My lab actually spawned out of one those cases, where they turned to robotics and automation, which is an emerging market that’s just exploding right now.


To what extent are you sacrificing accuracy or precision for speed?

You’re not actually changing accuracy or precision at all, because the accuracy of the scan is based on the accuracy of the Leica scanner and the tracker—it’s the same whether I’m standing there scanning the wing or I have someone running back and forth with the scanner.

The accuracy of the robot also doesn’t make an impact, because it’s the tracker that’s monitoring the scanner. Everything’s independent, so as long as the tracker and scanner are calibrated properly, you’re good. What changes is the density of the data that you’re getting.

(Image courtesy of NIAR.)
(Image courtesy of NIAR.)
So, if the scanner’s operating at 1,000 Hz, how far I move over that second affects how many points I collect within a region. Your limitations are how fast you want to move and how much data you want to collect.

The nice thing with robotics is that you can speed the scan up or slow it down at any time, so you can go really fast over areas of less concern, slow down for the areas of high concern or go back and scan it again.

Can your setup be applied to industries other than aerospace?

The wind energy industry is very interested in this technology, because when you’re dealing with wind turbines that are very long, a robot on a rail is a very attractive option because it can inspect these things very quickly.

Automotive is another good area and their tolerances are little bit looser than aerospace. Although, with the mandate for higher efficiency vehicles they are focusing more on tolerances and aerodynamics these days.

But really, any large components, whether it’s marine, aerospace or wind energy, can definitely see an advantage here.

What about smaller parts? Would it be advantageous to use a desktop-sized robot with this approach?

Definitely. One of the projects I keep wanting to take on with Hexagon is an inspection table. So, with collaborative robots these days you can teach them just by guiding them. My idea is to take one of these collaborative robots, mount it to a table and then hand it a scanner.

The idea is that I, as the metrology expert, could come in and teach the robot how to properly scan a part and just let it scan the parts over and over again. You could even change out the end effectors and have it go from a gripper to a scanner and have it tend its own parts: place the part on the fixture, scan it and then put it on the good pile or the bad pile and then go to the next one.

It’s similar to what you can do with CNC machines, where you can turn one operator into a multi-operator because he can program five different robots to scan parts and then just tend to them.

You’ve suggested that the aerospace industry is resistant to automation, whereas it’s very much the opposite case in the automotive industry. Why do you think that is?

In my opinion, there are a couple of reasons. One is that we don’t have the necessary throughput. In aerospace, if you produce 50 aircraft in a month, that’s a really good month. One of the most popular airplanes in the world is produced at a rate of about 40 per month right now, and that’s actually a very fast rate.

In automotive, 50 cars in a month puts you out of business pretty fast. They’re talking about cycle times in seconds or minutes whereas we’re talking about them in terms of hours or days. So, one of the big challenges is that the aerospace industry is afraid of making a big investment to have a robot do just one thing.

(Image courtesy of NIAR.)
(Image courtesy of NIAR.)
That’s where we try to help by showing the robots can do multiple things; they can tend to the machines, handle the parts and scan them, and they can do all these things at one time. But if you think a robot can only do one thing, and it only does that thing for ten minutes a day, then it’s not really worth the investment.

The other trouble you run into is the cost of the items these robots are working on. If you’re in automotive, you might be working on a $75,000 Corvette, but in aerospace, you can be working on a $100 million commercial airliner or a $3 billion fighter jet. The scale of the “oops” factor is much greater when you’re working on aircraft.

One of the things I’m always hearing is, “Well, I wouldn’t mind putting a robot in front of it, but you have to guarantee me that the robot will never, ever hurt my airplane.” Granted, we don’t want them to hurt people either—that’s a given—but if I have a robot painting the plane at the end of the line and it puts a big gouge in it, now I have an airplane that was all but done and ready to ship that I have to send it back for repair and explain to the customer why it was late.

From what I’ve seen, the contrast case with automation is additive manufacturing, where the aerospace industry is jumping in with both feet while automotive is waiting for the technology to improve. Is that a fair assessment?

I’m glad you asked about that, because we do a lot of work with additive manufacturing as well. Aerospace is very interested in additive because it allows manufacturers to make unique parts, and it doesn’t matter if its slow. If you can produce 45 parts in a month with additive, you’re okay. Whereas in automotive, they might need to produce 45 parts in an hour. Additive is at the point now where it’s good for interior parts and non-structural parts, but they just can’t make them fast enough to work in an automotive environment.

In my opinion, metal additive still has a ways to go, but plastic parts can be used in a lot of places already, and they have been.

Do you have any upcoming projects that you can tell us about?

Basically, we have a lot of different companies coming to us to explore what’s possible: whether it’s possible to handle textile-type materials—aircraft use a lot of composites—because a lot of people are allergic to carbon fiber composites and a lot of the resins aren’t exactly good for your health.

There’s also a lot interest in dealing with kitting, putting pieces together into kits and handling those kits so they can have the right parts at the right time. In aerospace, it’s very important to have no extra parts as well as having all the parts that you need. It’s basically a modular approach, where I’m only getting the parts and tools I need when I need them, so that in the end I have nothing left in the kit and no missing tools—if I do, that’s a quality control problem.

For more information on the National Institute for Aviation Research, visit the NIAR website.

Recommended For You