Robots Step Up to Machining
Despite some inherent limitations, multiaxis robots are being adapted for milling, drilling, and other hardtool machining processes
By Michael Tolinski
We are still a long way from a sci-fi world in which sleek, articulated-arm robots perform every conceivable manufacturing operation on every material. Given their accuracy requirements and the hardness of their workpiece materials, most metal-cutting applications depend on dedicated, rigid, CNC-based machines. By contrast, robots are flexible—in both good and bad ways: They can be tooled and programmed for various operations, but their construction lacks the stiffness necessary for tight-tolerance machining.
Overall, the importance of robotics continues to grow, especially in the automotive sector. Auto suppliers and manufacturers are the biggest users of robots, and together accounted for 70% of the robots ordered in 2005, according to the Robotic Industries Association (Ann Arbor, MI). Welding and material handling made up over 85% of applications for robots shipped in North America last year, but robots are also being used for higher-force applications as well, including material removal.
In fact, there are signs that more multiaxis robots are being successfully armed to perform accurate machining on harder materials. This is a boon for large shops that want to avoid the expense of dedicated CNC machine tools. And the expanded capabilities of robots are a bonus for smaller shops that perform many operations, but have limited funds for dedicated equipment.
The limitations of material-removal robots have mainly restricted them to the routering, trimming, degating, and deflashing of soft materials like plastics. For machining harder materials, multijointed serial-link robots generally lack rigidity, says Virgil Wilson, material removal product manager for Fanuc Robotics America Inc. (Rochester Hills, MI).
"One of the first rules of machining is: rigidity matters," Wilson says. "This is because of the external forces that are generated while machining." When a robot is used for machining, these external forces are transmitted back to it. A robot's ability to accurately control the tool's position in space depends on how well it withstands those forces, which are higher for harder, less-machinable metallic workpieces. Serial linkage robots have adequate rigidity for cutting plastics, foam, and clay, but "the forces that are generated when milling plastic are much less than when cutting steel," explains Wilson.
Serial robots' rigidity is especially limited because their members are linked joint-by-joint to the end-of-arm tooling. A serial-link robot is an open-structure mechanism (like a cantilever beam), in which the end-effector is connected to the base by a series of links and actuated joints. "With a serial-link robot, the inherent errors and backlash of each joint are compounded at the end-effector," observes Wilson. "Furthermore, each link is unsupported, and must support the load independently." This isn't a problem for cutting soft materials, "but for applications that require a great deal of rigidity, like drilling and machining, this can be a huge factor in the success or failure of the system."
In addition: "Factors like zero mastering, backlash, twist, gravity, and thermal expansion all play a role in the accuracy of a serial link robot." However, Fanuc provides software options that improve robot accuracy through better calibration (under the names RoboCal, TCP Cal, and Cell Cal).
But there's another robot architecture to consider, says Wilson: parallel-link robots. These robots are less affected by most of the above factors that reduce accuracy. Parallel-link robots have a closed, truss-like structure in which the end-effector is connected to the base by at least two independent kinematic chains or legs that operate in unison.
In Fanuc's F-200iB robot, six independent legs distribute the load from the tooling. "The position of the end-effector of the F-200iB is much less sensitive to the inherent error generated by the links and backlash of a serial-link robot," Wilson explains. The low deformation of the legs aids accuracy and repeatability, and inherent errors in the legs' structure are averaged together instead of added or stacked up. Wilson says this makes the F-200iB better for drilling, milling, and holecutting in harder materials like steel, aluminum, and gray iron.
Approximately 50 F-200iB's have been installed for material removal on aluminum castings and other metal components. In one automotive application, the robot replaced an NC machine used for cutting holes in 1.5" (38-mm) diam, 0.065" (1.65-mm) wall steel tubing. When compared to the robotic process, the NC machining process required more handling with less output and flexibility, says Wilson. "The operator would load a single part into the NC machine and mill holes in one end, then manually turn the part around and mill holes in the other end." The F-200iB hole-cutting system, using the same tool, reportedly reduced tool wear and allowed higher volumes, while allowing changes in hole position through robot reprogramming.
The flexibility of programmability is a quality of robots that can benefit even smaller shops. Being able to adapt a robot for material removal as well as more common applications is a bonus. Consider the case of Supreme Corp. (Goshen, IN), a medium-size manufacturer of molded fiberglass panels for commercial truck bodies.
Looking to reduce some excessive manual operations such as part trimming, the company brought in a UP130 serial-link robot from Motoman Inc. (West Carrollton, OH). But Supreme's manufacturing manager Tom Nowak realized the robot would be more productive if it were used to mill contoured mold patterns made from a medium-density fiberboard material. "We originally brought it in for trimming and pattern making, with the trimming being primary. But as we got into it, pattern-making seemed to be more of a fit."
By attaching a 7.5-hp (5.6 hp) spindle and standard end-mill tools to the robot arm, Nowak transformed the robot cell into a kind of CNC machining center. The extra work necessary for programming and training could be justified by the robot's efficiency. Some of Supreme's large, highly contoured patterns had required hundreds of hours of handwork to create; now, the robot mills patterns many times faster and mostly untended. Moreover, "the quality has improved," he says, "especially in the areas of symmetry and accuracy for complex curves."
"One real obstacle to Supreme adopting the robot machining solution was the programming of the robot," says Chahe Bakmazjian of software provider Jabez Technologies (Montreal, Quebec, Canada). For true CNC-type machining, the robot couldn't be "taught" its movements, since each part program contains tens of thousands of trajectory points, he says. So Jabez's Robotmaster software is used to convert CNC code from Mastercam (CNC Software Inc., Tolland, CT), based on each part's CAD data, into robot commands. "Robotmaster allowed Supreme to generate the robot programs to enable their robot system to function as a CNC machining center."
Meanwhile, Motoman has developed its own software that translates various CNC machine G-code formats into instructions for the robot. "In the past, robots were typically programmed by teaching individual points in the robot's path using a teach pendant," says Motoman's Greg Garmann, one of the company's G-Code Converter software developers. The software links robot movements to actual part data, without teaching.
"Once a G-code is produced from a CAD/CAM software package and is converted, you can download that program file into a robot controller and run it," says Garmann. "Of course, setup and calibration of the system also have to be considered to allow the real world and virtual world to match each other. This is not much different from what has to be done on a CNC machine." The robot's path can also be edited on the robot or using the Converter software, although the changes aren't retained in the original CAD/CAM files.
Along the same line is a software utility developed for Kuka Robotics Corp. (Clinton Township, MI). Integrated with CNC software from Delcam Inc. (Windsor, Ontario, Canada), the program translates CAM data into robot instructions, says Tim Brooks of Programming Plus (New Berlin, WI), the software integrator. "You take the output code from Delcam and run it through our utility, and it gives you the correct code for the Kuka robot."
The software and a basic spindle end-effector on the robot have allowed some unique robotic machining applications, Brooks reports. "We've got people cutting sand for pattern-less machining," where, rather than cutting a wood pattern and using it to create a sand mold, the robot cuts the sand mold itself. Other, building-related applications include the milling of foam statues, wooden stairway parts (like banisters), and stone facades. Plus, a foundry application is in the works for deflashing, trimming, and grinding sprues and runners from aluminum castings. "We're actually going to use a camera to tell us the location of all the things that need to be trimmed, and then we'll write the code on the fly," he adds.
But for high-accuracy machining, perfect robot calibration is essential. Motoman's Garmann says that G-code conversion software can't by itself make a robot more accurate. As with Fanuc's robots, calibration software like the company's Motocal can improve accuracy. These calibration routines identify the robot's base positions and confirm where a given point is actually located in space.
Calibration helps only so much, and Garmann generally agrees with Fanuc's Virgil Wilson about the structural limitations of robots. "When you're talking about milling applications, the design of the robot limits the amount of force that it can exert on a part," says Garmann. But sturdier and larger serial-link robots have the rigidity to apply the required force to process the part. He points to Motoman's DX1350N robot, for example, as having a more rugged design of the sort needed for material-removal applications. Large robots like the HP165 or HP200 are also better suited to robotic machining. "Just because of their sheer size, they have higher inertia capabilities that make it possible to do some of those tasks."
A Parallel Approach for Machining Centers
An alternative architecture for CNC machining systems has stayed in the background, despite its potential for quicker machining of complicated geometries. Called parallel or "strut" kinematics, it's analogous to parallel-link robotics in the way the tool is positioned.
When used in machine tools, strut kinematics moves the tool with a minimum number of pivot points, offering increased rigidity and accuracy. Only the mass of the tool is accelerated, allowing high dynamic response. "The kinematics allow for high accelerations at the tool tip, resulting in less nonproductive time," explains Karl Rapp, machine tool & automation branch manager for Bosch Rexroth Corp. (Hoffman Estates, IL). The company has developed control systems for these machines.
Machining centers based on parallel kinematics are available, mainly overseas, but generally they haven't caught on for large-scale production (though Rapp says they are used in pilot and prototype testing applications). The reasons may have to do with the complexity of their controls and uniqueness of design. The multiple degrees of freedom of movement offered by strut kinematics makes position verification and calibration anything but straightforward. "Predicting and verifying that the tool point will move to the accurate location within the work envelope is not trivial," says Rapp.
But mostly, these special kinematics machines have failed to catch on for reasons other than performance issues. Due to their unconventional design, would-be purchasers have limited expertise for maintaining the machine on site. Typical end users can't repair the equipment, since it's not based on the relatively familiar slide kinematics of standard machines. These systems are also seen as "high-risk machines," says Rapp, since people "fear that these machine types will not exist in ten years, or that the OEM will no longer exist to support them."
This article was first published in the September 2006 edition of Manufacturing Engineering magazine.