Automating the Small
Robotic welding cells evolve to handle shorter and shorter runs
Robotic welding has entered the job shop market with gusto, and here's why.
My son got a haircut recently. He's a kid, so the barber didn't have to cut too much hair, but being 5, my son required some time to get him seated and adjusted to a height suitable for the barber. My son shifted around quite a bit, so I emphasized to him that it was important to sit still. The barber paid special attention to get the haircut right (and this includes his extraordinary efforts to avoid nicking an earlobe).
It may seem a stretch, but try comparing this scenario to robotic welding, specifically those cells devoted to the high-mix, low-volume work of a job shop (seeFigure 1). The lot sizes are small (like my son's head of hair), setup takes up valuable time, and the cost to hold a part to maintain tolerance and repeatability (like my son's fidgeting head) becomes a significant consideration because it cannot be spread over a large volume. The operator, like the barber, has to pay special attention to maintain quality (a good haircut) and avoid scrapped parts (a nicked earlobe).
Historically, small job shops haven't considered robotics. Any purchased equipment had to be sturdy, easy to integrate, and inexpensive. Also, what the shop would make one day wasn't necessarily what it would be making the next.
So, why would job shops actually want a robot in the first place? It's a combination of factors, but more than anything, fierce competition has forced job shops to increase productivity while maintaining or improving quality. With the right fixturing and joint repeatability, robots can increase productivity for a range of part geometries, no matter the lot size.
That said, job shops must know which parts suit robotics and which do not. They must know part fit-up and fixturing requirements; that a robot demands repeatability from upstream operations; and that robotic technologies do exist to account for variability. There are myriad considerations, but most hinge on the following: Keep it simple and keep it flexible.
At its most basic, fixture design involves supporting the part, locating the part, and holding the part down. Supporting the part often is accomplished with a steel fixture plate, and it is common to use quick-changeover principles. Installing a subplate to a machine table enables the installation of receiver bushings that mate to locating shanks installed on modular fixture plates. A couple of turns of a set screw in each of the locating shanks provides positive holding force, accurate and repeatable fixture positioning, and easy changeover for small part runs.
Stick with off-the-shelf pins and mounting blocks to locate the part. These simplify fixture design, and are already machined to tight tolerances, which keeps costs down and requires minimal, if any, alterations out of the box. Off-the-shelf hardware such as hold-down action clamps provide a low-cost method to hold the part down for repeatability (see Figure 2).
Ideally, parts should locate from the most accurate surfaces, and those surfaces usually are machined holes or milled surfaces. As a rule (though there are exceptions), if the part requires a certain tolerance (±0.005 inch, for instance), then the fixture must locate parts to a tenth of that tolerance (0.0005 in.). The weld joint should also be located plus or minus half the diameter of the welding wire the job requires. So if a job uses 0.035-in. wire, the joint-locating tolerance must be 0.017 in.; 0.045-in. wire requires 0.022 in.
A robot lends itself best to lap and fillet welds, and if the weld joints can be positioned slightly downhill, tilted at an angle facing the operator, even better (seeFigure 3). This makes it easier for the operator to load and unload parts, shortening his reach to place something onto the fixture plate. It also allows the robot to use gravity as an ally; welding downhill means the gun can travel between 10 percent and 25 percent faster than when welding flat.
Sometimes the fixture can be blamed for tolerance problems. But more often than not, the weld operation is not repeating because of problems upstream, such as out-of-tolerance bends on a brake. The easiest approach is to program the robot to weave or stitch weld. Alternatively, using sensing devices like probes and low-cost cameras, robots can "see" joint variation and account for it.
As with any technology, joint tracking has positives and negatives. The more gap there is in a joint, the more data these sensing devices must process, which increases cycle time and often causes overwelding. This isn't to say these devices aren't worth the investment; they have introduced many applications to robotic welding that would not have been practical otherwise. Regardless, in job shop applications, such cycle time differences are almost negligible compared to a robotic cell's overall efficiency and quality.
When analyzing cycle time and lot size, make sure part size enters into the equation. For instance, if it takes 2 minutes to weld a part, then 200 parts may be considered a small lot size. If it takes 30 minutes to weld a part because it is a larger fabrication, 200 parts may be a significant lot size. Part size matters, and it helps put the use of joint sensing and tracking options into perspective. An additional 10 seconds to a 30-second cycle is more significant than an additional 10 seconds to a 3-minute cycle.
There was a time when programming was not easy or intuitive. Companies used to hire programmable logic writers and attempted to teach them how to program a robot to weld, an art that requires very different skill sets and experience.
This does not happen as often today. Programming pendants provide easy-to-understand robot commands, in plain English, such as arc start, arc end, and so forth. Robotic cells in a job shop do not necessarily require code writers or someone on staff to support PLCs. If someone knows welding, he can most likely operate the pendant for a robot.
For a technician who will program only certain small runs, training can take as little as a half day. More comprehensive training takes three to five days. By the first day, students can become proficient in the basics of robot motion and start programming their first welds.
By the second day, they can learn how to program more efficiently. A robot may not weld faster than an experienced welder, especially on straight seams; instead, it is in the moves between welds where the robot wins the race. To make sure the robot moves quickly between welds, programmers can define an end point of an air move with a wide tolerance, say, ±3.0 in. Once the robot gets to an approach point, it then can make precise, slower moves toward the arc initiation position, with tolerances down to a few thousandths of an inch.
By the third day, students are welding parts and learning to program simple logic for data collection, including how to insert a program timer to measure cycle time or a counter to measure the number of particular welds or moves within a period of time.
Consider an application involving 200-part lots for a wheel assembly fixtured on a plate with off-the-shelf toggle clamps. The robot is programmed to weld a circle, something it can make very efficiently, with just one arc start and stop. Humans have difficulty making such a weld, because it is a challenge to hold a gun steady at a consistent gun angle and travel speed around a tight circle. Instead, the welder would be forced to start and stop two to four times along the circle circumference. Hand welding the joint, an experienced welder might achieve 15 inches per minute (IPM), while a robot could weld the same joint at 45 IPM—a threefold speed increase with only one arc start and stop.
Of course, some part geometries should not enter a robot cell. Repeatability is the name of the game, and significant gaps in fit-up can cause problems. A robot can be programmed to weld those gaps, but if those gaps are inconsistent, problems arise. Here, a hand welder would be better suited to stitch those gaps closed than programming a robot for the worst case on every part.
Part geometry considerations are not always so clear-cut, and often they must be weighed against other factors, such as the availability of skilled labor. A shop may bend a sheet metal box in a brake operation. The seams generally are easy to weld because they can be located on a fixture. This is not necessarily true at the box's radius corner, where the seams come together. Depending on what type of shearing is used, that corner can have an odd contour, and the gap can be very difficult to control. In some applications, it is better for the operator to tack and weld those manually. Note that robots can weld such joints, and they commonly do, often because skilled labor simply is not available. However, if a shop has a skilled hand welder on staff, he may be able to join these parts more efficiently than his robotic counterpart.
The job shop's goal is to maximize throughput, and often that means not sending every weld through the robot cell. Consider a part for which most welds can be made in the flat position on one side, while two basic stitch welds are required on the other side. A company might look into building a positioner to flip the workpiece under the robot. But in a short-run situation, why spend the time building such a complex fixture or deal with the added cycle time, particularly with a welder on staff? A robot could make most of the welds, after which a welder could simply remove the part and take a few seconds to make those two stitch welds and then send the part downstream.
A robot (depending on the model) can save about 15,000 program points into memory, with optional memory expansion capability. Each small-part program might have about 40 or 50 program points, including air moves and welding moves. So a robot has a nearly unlimited capacity for programs, and the programmer should use this capacity and flexibility to maximize throughput.
To do this, the robot should never wait for an operator to load parts. Using a simple spreadsheet showing cycle times for various parts and part sizes, a shop can see what arrangement of small and large parts ensures that the robot arm is always on the move.
Consider two jobs that must flow through a robot cell: a 32-piece part run of 5-in.-wide parts, and a four-piece run of 40-in.-wide parts. An operator could load one large part and initiate the program, after which he could nest eight smaller parts to a fixture plate and have it ready before the welding cycle finishes. After loading the eight parts, he moves on to the next large part. Staggering these two batches maximizes throughput, and the robot is never waiting for parts.
Time for Maintenance
In job shops, people wear many hats. The person programming may be the same person designing the fixture, operating the robot, and maintaining it. In these cases, the maintenance hat should not be thrown away.
Unlike humans, robots cannot anticipate when problems arise. A hand welder can feel the vibration in the gun when debris begins to hang up in the contact tip. A robot, of course, cannot, so it is better to replace contact tips on a regularly scheduled basis.
Though schedules may change because of uptime differences, manual welding PM practices generally mirror PM for robotic welding. This is where working with experienced welders helps tremendously. They already know that contact tips need to be changed regularly, often once a shift. Nozzles might be replaced anywhere between once a week and once a month, while gun liners and drive rolls might be on a six-month schedule.
Some maintenance items are unique to robots. They should be wiped down periodically, and they should be oiled or greased, depending on the robot type. Backup batteries, which allow systems to maintain motor positioning integrity during a power outage, should be replaced twice a year or so. Also, look for repetitive rub points and adjust robot motion as necessary.
Opportunities for Short Runs
In the U.S. many job shops are investing in their first robotic weld cell. As small-footprint systems and adaptive technologies become more affordable, more robots are sure to make it to the small shop floor, and most will tackle short part runs—those thought too small for automation just a few years ago.
The Importance of Skilled Labor
In an ideal world, shops should consider robotic welding not just because they can't find skilled welders, but because they want better efficiency from their welding operations. Skilled welders at the robot can account for variability. What if a robot program needs to be tweaked because of a minor programming error? The more errors an operator catches before loading parts, the more uptime a robot cell will have. This is especially true in a job shop, where the operator has ownership over the process. He should be able to recognize these changes without having to ask for help, to keep things running smoothly.
However, while it is desirable to have a skilled person program a robot, operators need not be welding engineers. They should have some welding experience, at least enough to understand the effects of gun angle, welding parameters, and wire placement to tweak a program on-the-fly if necessary.
It is about getting the most out of the skill set available. A semiskilled operator might load and unload a part while the robot welds. Meanwhile highly experienced welders can spend most of their day welding complex or prototype joints that would not suit the robotic welding cell.
The FABRICATOR is North America's leading magazine for the metal forming and fabricating industry. The magazine delivers the news, technical articles, and case histories that enable fabricators to do their jobs more efficiently. The FABRICATOR has served the industry since 1971.