Laser scanning systems in auto assembly plants are often referred to as the "Perceptron," almost in the way that facial tissues are often referred to as "Kleenex." Chances are good in both cases that the products in question actually are those of the names, as they have become so pervasive in their arenas. In the case of Perceptron, Inc. (Plymouth, MI), the company makes on the order of 5,000 sensors per year. So they know a little something about things like gallium arsenide laser systems (eye-safe, 670-nm wavelength lasers, as in the case of the system we'll get to here). Generally, the people who have been using Perceptron systems are those who are dealing with quality control issues of sheet metal parts and assemblies. As in cars and trucks that are making their way through the assembly line. (That product is known as the "AutoGauge." Because it is often set up so that the vehicle travels through a frame with the laser scanners fixed in place, sometimes it is referred to as a "car wash," based on the resemblance.) Perceptron also produces systems that can be used to automatically measure paint appearance (checking for such things as orange peel and the amount of gloss), a system known as AutoSpect, and to measure gaps and flushness of closure panels including hoods, doors, and lift gates, a system known as AutoFit.
Essentially, Perceptron has become common in production operations. But as for product development applications. . .well, the company heretofore has not had much of a play. After all, they're a company that performs non-contact measurement of things that actually exist, not things that are in becoming.
But the people at Perceptron recognized something: In every automotive product development program, there are always physical artifacts produced. Sure, every company may be using CAD to design parts and assemblies, but during the program it is absolutely necessary to produce prototype parts. So the parts are made. And then fixtures are produced to measure those parts. Complicating matters a bit (although this is just the ordinary state of affairs in any development program) is the fact that parts don't exist in isolation. They interface with other parts. So not only does the development program include the design, prototyping, and measuring of discrete parts, but also of assembled parts.
Consider, for example, a body side stamping and the doors. It very well may be that the prototype parts are being produced in different locations. So it is a matter of shipping the parts to a single location, where they are put together on a frame. Then someone pulls out a feeler gauge and measures the gaps and flushness. Results are written on the bare sheet metal with a grease pencil. Unless there is an amazing coincidence, chances are, a decision will have to be made regarding the modification of one or more of the parts. Parts are shipped back, a new die is produced, and then the sequence begins anew. Unquestionably, the influence of math modeling is making this all better, but as Song Yop Chung, director, Technology Components Group, at Perceptron notes, "Not all CAD designs are done with full consideration of manufacturing variations. No designer completely understands the full manufacturing process. If they do, then you could support the argument of going directly from a math model to the manufacturing process." In the world of stamping, for example, there are pesky things like springback that get in the way. Or when parts are clamped for spot welding, there can be fitting problems or distortions resulting.
Enter the laser-based scanning system, which they've named "ScanWorks."
No, Perceptron hasn't invented the idea of scanning parts. There are plenty of laser-based reverse engineering systems out there. But what they are doing that is somewhat different is looking at this from the context of product development. The idea is a simple one: Make the first articles. Instead of (1) shipping them somewhere where they will be (2) placed in an expensive fixture to check the fit with other parts, simply take the scanned and processed data and send it via the Internet to a place where there is a "virtual fixture" upon which the initial checks can be made. Yes, modifications will undoubtedly still be required. But this can be done via something along the lines of a WebX Systems-based meeting, which is a heck of a lot more convenient than a car ride across town (to say nothing of a plane ride anywhere).
The scanning system is itself comparatively simple (from the point of view of use, that is: the sophisticated part—the algorithms—is transparent to the user). In round figures, a system costs about $100,000. Essentially, the ScanWorks probe is mounted on the arm of a portable CMM. Then it is just a matter of the part to be measured being simply set up in place and scanned. The sensor acquires 768 points per line and up to 30 lines per second, which translates into 23,040 points per second. Perceptron programmers have written the program so that there is automatic recognition of features like holes in the three-dimensional space. Song Yop Chung says, "No one in the world, so far as we can tell, has a set of algorithms like this that actually work on—and have been proven on—sheet metal parts. Not just sheet metal parts in an office, but on production parts." Just as they're leveraging their sensor know-how, the people at Perceptron are leveraging their software as well. In fact, the same algorithms are used for the all of the sensor systems the company provides. (It should be noted that the ScanWorks system is not a full-blown CAD system; it works within a CAD system.)
Remember the claim about the familiarity that vehicle manufacturers tend to have with Perceptron's in-line shop-floor products? There are always exceptions to the rule. One is Toyota Motor Corp. in Japan. The company has purchased more than 150 of the systems because they were interested in first product development, then process development, not because they had a lot of factory-based systems on the floor.. One of the things that they are doing with the system is to actually validate production processes. That is, it may be that two qualified parts are welded together and the result is something other than what had been anticipated. Rather than automatically sending the dies back for rework, it may be an adjustment in the welding setup that's called for (e.g., clamping in three places rather than in two).