At GH Metal Solutions, fixture expertise drives welding efficiency
February 28, 2012
An Alabama contract fabricator uses innovative robotic weld fixtures and an unusual quick-change system that allows operators to switch out jobs in mere seconds. The innovations have been critical for the shop’s high-mix, low-volume work.
When Eliyahu Goldratt wrote The Goal, the business novel about the theory of constraints, he hit on a hot-button issue: the robot and true manufacturing efficiency. The book tells the story about a fictional plant functioning with what managers believe is a skeleton crew. Everyone is in almost constant motion, including the robots. Still, the company’s finances are in shambles.
A consultant named Jonah (a thinly veiled caricature of Goldratt himself) talks with the plant manager about the robotic welding cell. The plant manager boasts that the cell has sky-high efficiency, and that welding costs had plummeted. The robotic works consistently and continuously. Jonah asks, “Was your plant able to ship even one more product per day as a result of what happened in the department you installed the robots?”
The manager, taken aback, doesn’t immediately know. He ultimately finds (among many other problems) that the robot welds constantly because the floor is flooded with work, much of which isn’t needed. An order may require only 200 parts, but the welding department supervisor assumes the order will come around again someday, right? Why not save the setup time and run several months’ worth of parts all at once?
This means that massive batches move between operations, and many finished products ship to an overstuffed warehouse. Sure, the robot’s efficiency measurements are phenomenal, but they aren’t helping the company sell more products or, ultimately, make more money. The company just keeps buying more metal to feed its machines, which aren’t producing what customers immediately demand. Meanwhile, raw material, work-in-process, and finished goods inventory just sit, eating up cash.
Edwin Stanley would probably agree with Jonah’s perspective. And though it may seem a bit counterintuitive, that’s the very reason his shop relies so heavily on welding robots.
The vice president of marketing and finance at GH Metal Solutions, Fort Payne, Ala., has seen the contract metal fabricator finish a record-breaking year. In 2011 the company had revenue of $44 million. “This year we’re projecting $50 million,” Stanley said. “But I’ll actually be a bit disappointed if we don’t hit at least $54 million.”
If the company relied solely on traditional efficiency measurements, it probably wouldn’t have five welding robots (see Figure 1). The fabricator doesn’t run high volumes. Because customers don’t want to hold inventory anymore, GH now may deliver, say, 20 parts a day instead of 100 a week. This forces the fabricator to run small batch sizes. Historically, that would steer people away from welding robots. Sure, a robot may weld quickly and consistently, but fixtures cost money, as do all those changeovers, right?
GH Metal’s managers thought differently and—judging by recent financial performance—rightly.
The company’s founders wouldn’t recognize GH Metal today. The firm launched in 1958 as a natural gas appliance and heating company, hence its longtime name, The Gas House. Gradually the organization dipped its toe into the stamping business, forming various ductwork and other HVAC components.
In the 1960s the Tennessee Valley Authority initiated a program to reimburse customers who installed vents to keep attics cooler in the summer. This was a bonanza for those who could stamp the mushroom-shaped roof caps, and The Gas House dove in with everyone else. The company evolved into a full-fledged stamping operation, gradually taking on more customers through the 1970s and 1980s. Then, with increased demand for lower-volume work, the company shifted focus to metal fabrication. In 2009 the fabricator officially changed its name to GH Metal Solutions. It still maintains its stamping department, which has both hand- and coil-fed presses from 50 to 1,000 tons, but its niche remains in lower-volume work.
“We run only a handful of parts that require more than 50,000 pieces a year,” Stanley said. “That’s just not our niche anymore.”
Over the years the shop added lasers, punch presses, a punch/laser combination machine, material flatteners (to prepare flat parts for bending), press brakes, and even several tube lasers. Revenue improved, but part flow efficiency didn’t.
Equipment was added piecemeal, and the shop floor looked it. A patchwork of different machines spread throughout several old buildings near downtown Fort Payne. One early 1900s brick building began life as a cotton warehouse; a century later it housed some of the latest in laser cutting technology.
Still, managers knew they weren’t taking full advantage of the technology’s flexibility, because of how inefficiently batches moved between operations. Fork trucks abounded. Parts had to be loaded on shuttle trucks to go to buildings scattered about town. Even if parts had to go to an adjacent building only several feet away, the work-in-process couldn’t be transported on carts, because the other building’s foundation lay on a higher plane, so trucks had to drive up and down ramps. GH Metal processes various sheet and plate thicknesses, much of it between 0.15 and 0.375 inch. Pushing unpowered carts carrying several hundred pounds of metal up and down ramps isn’t safe or practical.
By 2007 part flow improved dramatically when the shop moved most operations to a 200,000-square-feet facility just north of town, right off the interstate. Now with 300 employees, the organization is one of the largest contract fabrication shops in the Southeast. Last year more than 750 truckloads of steel arrived at the loading dock on one end of the shop floor. Raw stock gets nested on one of 14 laser cutting machines (between 3 and 5 kW in power). From there the blanks travel in one direction—to bending, welding, finishing, and then coating. (GH Metal offers both powder and e-coat capabilities.)
Today many of its largest accounts place a repeat order (not for a new part) and expect it to ship within days, not weeks. Advanced cutting and bending machines manufacture parts within ever narrower tolerance windows. This not only increased overall quality, but also gave GH Metal additional options for welding.
Precise cutting and bending opened the door to joining automation.
Welding remains a principal bottleneck for many fabricators. It takes seconds to laser-cut a part, minutes to bend it, and even longer to weld it.
The way GH Metal saw it, consistency is everything. For instance, weld draw (or shrinkage) can become a problem for some parts. If a manual welder moves slightly differently from one weld to the next, the heat input changes, and so does the weld draw. In these cases, a robot could consistently weld a part the same way time after time, using precisely the same current settings, weld speeds, and standoff distances.
Still, would such consistency really increase throughput, especially considering the company’s small batch sizes? And would all the fixture development costs be worth it? A robotic fixture takes longer to develop than a manual welding fixture, because the robot can’t manipulate a part like a person can. If a fixture doesn’t provide access to all joints, a manual welder can simply tack pieces together and then manipulate the part as necessary to access the hidden seam. Not so with the robot.
The answer to these concerns lies with GH Metal’s product mix. The floor is a high-product-mix operation, but a good portion of the work is repeat orders. Most customers order a little bit at a time, but at frequent intervals. A worker may process a batch of 20 parts, but chances are he will see the same order many times over the course of a year.
GH Metal has a robust tooling department, with a high-speed vertical machining center dedicated to milling fixture and tooling components for the company’s welding and stamping operations. For most repeat orders, the company develops a fixture to weld a part robotically. When the order comes up again, the company has the fixtures on-hand, ready to go. Most fixture development costs occur with the first order; the company makes up for it and then some with all subsequent orders.
Engineers also have developed ways to save time and money when designing and building fixtures. Software plays an integral role. When the company creates a new 3-D model of a new, never-before fabricated part, engineers import the model into CAD and design a fixture around it on the screen.
Still, no matter how quickly personnel can develop a robotic welding fixture, would robotic welding increase overall throughput—especially considering an operator may weld 10 of this, 15 of that, and 25 of something else? Having the fixtures at the ready is fine, but when operators process batches of dozens instead of hundreds, they must perform changeovers frequently. All that time adds up.
High-volume operations use hydraulic or pneumatic clamping, but GH Metal’s small runs don’t warrant such investment. Besides, designing and building all that plumbing for hundreds of different fixtures would complicate an already highly complex situation.
The company may have hundreds of different fixtures, but every one of them uses the same basic base plate design—and it is here where engineers found a solution. Traditionally, an operator switching out a fixture would need to bolt the base plate in various areas to locate the part for the robot. But what if those base plates could be attached and located quickly using pneumatics?
GH Metal found a way, and to explain it, Ron Kruckenberg, tool and die team leader, held up a billet of S7 tool steel. His department machines the billet into a cylindrical component with holes for ball bearings around the circumference. The cylinder slides into a machined hole with a raceway, and pneumatics force those balls into the raceway, locking the two components together. This locking mechanism is a central component of a quick-change base-plate system that allows operators to change out a robotic welding fixture in seconds (see Figure 2 and Figure 3).
To set up a new job, the operator wheels a cart fabricated to the exact height of the fixture base plate in the robot weld cell (see Figure 4). The base plate unclamps, and the operator slides the plate onto one cart. He then slides the new fixture plate off another cart on guides, which lower the plate onto the cylindrical quick-lock system that locks and locates the fixture assembly in the robot cell.
On his computer screen, Phillip Finch, robotic welding programmer, brought up the last piece of the puzzle: offline programming. For many parts the company still uses a teach pendant to program a first-run part. But for others, Finch has been able to simulate the welding process offline.
“We’ve gotten a program within 70 to 80 percent offline and touch it up on the actual machine with the teach pendant,” Finch said. “We hope, as we continue down this path, to do everything in the virtual world before we ever make a chip or weld a component.”
The simulation also helps perfect weld fixture design. Today fixtures for repeat orders are continually tweaked. The more a part is robotically welded, the better fixture solutions engineers devise. These improvement efforts will continue regardless of what software technology brings; some ideas just don’t come to light until a piece is welded in the real world. But offline programming may help engineers identify more areas for improvement before the robot ever strikes an arc on the part.
Stanley doesn’t lack in humility. Without hesitating, he showed where he thought the operation could improve. At some point, he said, the company would like to reduce the company’s significant raw stock inventory.
Of course, having stock on hand does provide insurance. A case in point was when a string of tornados skirted the Fort Payne valley last year, carving a path of devastation up the Appalachian foothills, just several miles from the shop. Luckily, the shop lost power for only two days—but those were two days Stanley will never forget. “We had no power and no land lines, only cell phones,” he said. “It was a crazy two days, needless to say.”
But the company was lining up contingency plans with area shops. If the storms had reduced their metal suppliers’ ability to deliver quickly, GH Metal’s managers knew they had some stock to get them through the emergency.
The company attempts to limit its finished goods inventory, though Stanley conceded the shop does overproduce on occasion. Customers may change order quantities for parts already in production, so some finished goods sit until the purchase order comes up again. The shop is working to reduce that finished-goods inventory, and one way is to make product flow as quick and predictable as possible—and this includes part flow in the welding area.
Robotic welding moves more experimentation and process development away from the shop floor and into the toolroom and computer desktop. Designing and machining a new tool takes time, but it doesn’t hinder the flow of current production. And in the robotic welding area, production happens fast. As just one recent example, manual welders took about 12 minutes on one component. Welding that same part on the robot took only 3 minutes.
As sources explained, GH Metal’s welding strategy takes advantage of the best attributes of both robotic and manual processes. A robot takes on repetitive tasks, while manual welders perform what robots can’t do effectively, including one-offs and non-repeat orders (see Figure 5). Manual welders excel at adapting to new parts and different challenges. A standard welding robot doesn’t adapt well at all. Parts must be consistent, fixtures precise. That requires a lot of prep work, and for one-offs and non-repeat orders, all that just doesn’t make business sense.
On occasion a robot may weld 90 percent of the assembly, while a person may weld the rest. This seems a bit counterintuitive on the surface. After all, what about all the extra handling? Workers must remove the part and take it to the manual welding station for a few final weld seams.
Stanley explained that the reduction in weld time has been so dramatic, automation still saves time. Robotically welding a component and moving the workpiece to the manual welding station can be more efficient than hand welding an entire assembly.
Some parts welded entirely by the robot may also need to be repositioned once or twice, so the robot can access hidden joints. This also seems somewhat counterintuitive. If you’re welding with a robot, why not strive for that “done-in-one” ideal? Besides, repositioning components adds more variability in positioning accuracy—which is exactly what robotic welding attempts to avoid.
In these cases, process simplicity plays a role. Designing one fixture with umpteen components may allow the robot to access all welds in one setup, but that one setup may be incredibly complex. If one complicated setup is split into two less complicated setups, the operation becomes much simpler and more reliable. Yes, repositioning a part does add variability, but engineers minimize this by using the same datum points for both fixturing steps, when possible. Having both steps locate the part off of, say, the same face of a plate minimizes the variability that comes with part repositioning, ensuring the final part measurements remain well within the tolerance window.
In 2008 the company shipped $34 million. Then in 2009 revenues plummeted to $21 million. In 2010 GH clawed its way back to $29 million, and last year revenue began climbing to the current record levels.
When revenues plunged so deeply during the recession, managers knew why: They had too many eggs in too few baskets. “Before the recession, the bulk of our business was with just two accounts,” Stanley said. “2009 taught us a lesson. We’ve added some new customers. And we’ve been lucky. The economy kicked some of our competition so hard, they weren’t able to get back up. That’s given us some opportunities. Coming out of the recession, there’s more low-hanging fruit out there.”
Even low-hanging fruit demands quick response. More customers bring more parts, which require more welding fixtures. During recent years, though, Kruckenberg, Finch, and others on the tooling team have reduced fixture development and machining time.
One team member, Tooling Engineer David Came, pointed to two seemingly insignificant, fingertip-sized spring stops. One spring stop had a cylindrical shaft that pushed a square pad against the workpiece. Came then pointed to an improved version with a triangular shaft that can’t rotate, so the pad pushes against the workpiece the same way every time (see Figure 6). It’s a small but critical element of consistent fixturing.
Came then pointed on his computer screen to a 3-D model of a fixture design, including several magnetic components that eliminate certain clamps. He described several plate components cut to hold a tubular workpiece (see Figure 7). “We use laser-cut components when we can,” he said. “This really reduces our costs, versus machining these components out of a solid block.”
The company carries over its tooling competencies in its stamping area. Using a similar rationale as they use in welding, engineers may design a tool to stamp all or at least some formed elements for certain repeatedly ordered parts. Stanley pointed to a part in a staging area for press brakes. “This part has 14 bends in it,” he said. “So we ran the part for a year [through the brake department only], and then our tool and die people got together and said, ‘Hey, I think we can build a die to form nine of those 14 bends in one hit.’” The remaining bends are made on the press brake.
He conceded that the company’s emphasis on tooling and fixtures didn’t come from some grand plan or comprehensive analysis. Managers just knew that customers wanted orders quickly, and to speed manufacturing the fabricator needed to make those highly variable manufacturing steps, including bending and welding, much simpler. According to sources, smart tool and fixture design helps the company do just that.
In all likelihood, Goldratt’s Jonah would agree.