Robotic Material Removal and Machining
Building on their ability to polish, deburr, and deflash, articulated arm robots are starting to perform some tasks now reserved for conventional machine tools
By Bruce Morey
Articulated arm robots could be an alternative to CNC-style machining centers in some applications. Engineers are using them for less-demanding tasks now reserved for CNC machines, like machining plastics, wood, or sand-castings. Technical challenges remain before they can machine tougher materials or finer tolerances. Engineers have begun solving some of the problems of using robots for machining, like programming the robot to machine a part. Alternative robot architectures may take them to the next level by providing more rigidity and accuracy.
Robots have carved out a niche in light material removal, such as deburring, grinding, and deflashing, according to Kevin Kozuszek of Kuka Robotics (Clinton Township, MI). "Sensor technology, such as our company's ForceTorque, gives the arm the ability to register the forces it's exerting onto the products, similar to what the human body would experience when performing the same tasks," he says. The next logical advance in robotic material removal would be into machining complex parts. He believes the extended work envelope, flexibility, smaller footprint, and generally lower price of robots should be key advantages over existing CNC machine technology.
"Robotic grinding and machining for material removal is one of the big growth areas in robotics, especially in the nonautomotive general industry market," says Kozuszek. He explains that professionals in this industry who currently use CNC machines are familiar with automation, and understand its benefits. "Their disadvantage is that robots are limited to softer metals and materials, and can't deliver as much accuracy as a CNC machine when cutting harder materials."
For robots to mimic CNC machining, they must mimic their programming. CNC machines are typically programmed offline with a CAM package, using a CAD parts description as input. The package reads in a CAD file and eventually spits out the brand-specific G-code that programs toolcutter paths or workpiece motions. Robots are not programmed with G-code, so they cannot use CAM packages directly.
Some companies have developed translators for robot programs to use CNC-like commands. One such company is Motoman Inc. (West Carrollton, OH), the robot manufacturer and systems integrator.
"We can interface with all of the popular CAD/CAM programs like Delcam, GibbsCAM, SurfCAM, or Mastercam, and turn that information into a set of instructions for the robot," says Greg Garmann, Motoman's development leader for software and controls technology. Motoman's MotoSim EG software simulates the robot path generated from the G-code program in the virtual world on a PC, allowing programs to be modeled offline and then downloaded to the robot controller.
"Creating robot programs from G-Code has worked well for our customers in applications such as cutting molds, and in complex waterjet cutting applications," says Garmann.
Automatic parts programming is the only practical way for robots to achieve CNC-style machining. "I cannot easily program 30,000 points by hand," says Garmann. The effectiveness of robotic off-line programming intimately depends on the accuracy of the robot. Robot programs created offline are often 'touched up' in other applications, like material handling, to compensate for inaccuracies. Motoman filters the initial robot programs created from G-Code by using the robot calibration report, a report delivered with each unit. Using this report, the program adjusts the output to create the most accurate robot program possible. Garmann reports that this application, especially with the robot calibration report as a filter, greatly reduces—and even eliminates—the need for program touch-up. According to Garmann, robots that could be used for machining, such as the Motoman DX1350N, can achieve an absolute accuracy of about ±0.020" (0.51 mm). This model also boasts a repeatability of 0.002" (0.05 mm), and is designed for material-removal applications.
A way to improve a robot's accuracy is to add measuring or machine-vision equipment at the point of use. Termed adaptive control, the technique measures the potentially changing world around the robot, providing feedback to its controller so that it can adjust the robot's motion. An example of adaptive control is to combine robots with a metrology system. That is what Airbus UK (Broughton, England, UK) did, citing that an absolute positional accuracy of ±0.2 mm was required in many application areas if they were to use robots to manufacture aircraft. In 2007, they completed a research project for a vision-based drilling system built by an integrator company, M3, centered around a Metris Inc. (Leuven, Belgium) K-series metrology system connected to the controller of a Kuka robot. Delmia Corp.'s (Auburn Hills, MI) V5 for Robotics and Kuka's VRC software tools provided the needed off-line programming.
The metrology system consists of three linear CCD cameras that measure infrared LEDs attached to both a part-holding jig and a drilling system at the end of the robot. Looking at the LEDs, the vision system measures the position of the tool, and compares it to the CAD nominal drilling position, correcting the position of the tool. The key to achieving this accuracy in real time is closed-loop processing achieved by delivering the results of the comparisons to the robot's controller. Research currently is looking at extending the number of jigs and at applications other than drilling.
The ability to program robots for complex parts programs has opened new niches for some integrators. One such integrator is Robotic CNC Solutions (New Berlin, WI). It has built a market niche around delivering integrated robotic workcells for machining workpieces that are traditionally machined by a five-axis CNC machine. "We are a Kuka integrator that specializes in marrying a high-speed spindle to a robot arm," says Tim Brooks, the company's general manager. "Our robotic integrations are delivered with up to 12.5-kW spindles that can handle tools with a shank as large as 1" [25.4 mm] in diam." They convert G-code created from Delcam, or other five-axis CAM programs, into Kuka robotic language (KRL) through technology obtained from a sister company, Programming Plus Inc. (New Berlin, WI).
Brooks agrees that practical direct robotic machining is currently limited to softer materials, such as wood, fiber-glass, plastics, various densities of engineering foam, and some limited applications for aluminum alloys like Al 6061. "Most direct metalworking is not appropriate for robots—yet," says Brooks.
By working indirectly, however, robotic machining has advanced metalworking processes. An application his company delivers is a robotic workcell for cutting green-sand molds for a patternless casting process. Normally, green sand packed around wooden patterns creates a casting mold set. "The process of creating a wood pattern can be expensive, depending on the part," explains Brooks. "Delivery can take weeks or months. Depending on the volume of production, our robotic patternless process can save considerable time and money. Another plus is that there is no wooden pattern to store, only the instructions for creating the mold." The typical ±0.020" accuracy that calibrated robots deliver serves patternless casting well enough, according to Brooks.
A limitation he notes with using robots is the difficulty in cutting straight lines and circles. "The robot has to move all six axes at once to make a straight line," says Brooks, "as a result, you do not get completely straight lines or flat surfaces. Robots have difficulty machining circles for the same reasons. They are within tolerance, but not perfectly straight or perfectly round."
"In practical terms at present, the best use of robots for machining is working softer materials like plastics, RIM composites and hand lay-up fiberglass," agrees David Voves (vice president general manager) of Rimrock Automation (New Berlin, WI), an automation and robotic integrator for ABB, Kuka, and Güdel robots. Rimrock is comfortable with delivering to an application that needs about ±1/32" (0.79-mm) accuracy. Rimrock delivers six-axis robotic routers for the materials mentioned above. In such applications, they compete with a special class of five-axis CNC-controlled routers. "Companies that have come to us have been frustrated by the reliability and speed of these CNCs, and their inability to reach areas that are difficult to access or located underneath the workpiece," says Voves. He notes that robots are highly reliable, with a minimum of 80,000 hr MTBF, more reliable than the competing CNC router machines.
Successful applications of robotic metalworking, he says, include metal finishing, deburring, and deflashing applications. "Where we have seen a lot of activity lately is in metal polishing for tool companies. The robots are automating hand labor for metal polishing," says Voves.
"To adapt to harder materials, robots must become more rigid. The problem with robots with link arms is that, by design, they are not very rigid compared to a machine tool," says Voves. "They have a series of linkages, coupled by pivots with gear motors in the pivot, to provide the torque to move each one. There is an accumulation of tolerance through each pivot, and precision is difficult when you have high cutting forces to deflect."
The future of robots in machining metals may very well be a middle ground in which robots enhance rather than replace CNC machines. Some people in the robotic supplier community fully appreciate the accumulation of tolerance through the pivots and joints of articulated arms.
"We see a big opportunity for robots in the foundry area," says Doug Niebruegge, segment manager for foundry and plastics for ABB Robotics Inc. (Auburn Hills, MI) "We do not think a robot is ever going to be able to completely replace a CNC machine in terms of accuracy, but we think we can do premachining with a robot," says Niebruegge. He describes a hybrid vision of robots performing premachining on parts that are about one cubic meter square with a dynamic path accuracy of less than ±0.020". Once pre-machined, these parts could then be finish-machined on CNC equipment. "If you did everything with a CNC machine, you might need 10 CNC machines. If you used robots to do premachining, you might be able to use two CNC machines and perhaps six robots. Robots continue to drop in price, and the cost savings would be substantial in that scenario," says Niebruegge.
"There are certain design techniques that can make a robot stiff enough to cut metal," says Niebruegge. As he describes it, one technique is a good servo system, another is a stiff arm made of cast steel or iron, and finally the architecture of the arm itself.
ABB unveiled a prototype with an additional link-rod in the upper arm at the International Foundry Trade Fair (GIFA) in May. The additional rod increases the stiffness of the robot arm to the point where ABB believes it can perform pre-machining with a consistent path accuracy of less than ±0.020" while supporting a metalcutting tool. Although not yet released as a product, it's this kind of robotic architecture that ABB will take into the new hybrid world of robotic premachining and CNC machining centers.
Niebruegge also sees force-feedback in robots as an important means of optimizing the cutting speeds of the tools. "You can measure the forces on the tool because they are the forces on the robot. If you are getting too much force in the direction of your cut, you could change the speed so that you can optimize the cutting speed for the material," he says.
Another group of robot-like machines, parallel kinematic machines (PKMs), provide better rigidity and accuracy than articulated-arm robots. They could bridge the gap between articulated arm robotics and conventional machine tools.
With several patents pending, Milwaukee MachineTool Corp. (Milwaukee) believes in the concept, and is developing an affordable PKM workcell, integrating existing off-the-shelf technology to deliver a complete machining and multipurpose workcell. "Our concept is based on a PKM robot, offering greater flexibility and economy than CNC machines, with a larger work envelope. It will hold better tolerances than articulated-arm robots," says Aaron Russick, vice president, Milwaukee MachineTool. "An example of the workcell we could deliver will hold up to ±0.025-mm accuracy."
"PKMs can be configured to match the reach and flexibility of traditional six-axis articulated-arm robots, with up to six or more degrees of freedom," says Russick. He goes on to say that modular configurations will allow multiple PKM robots and spindle combinations, in different horizontal and vertical orientations. Target applications include finishing castings, cutting blanks and grinding.
This article was first published in the November 2007 edition of Manufacturing Engineering magazine.