Automating with Vision
Latest robotic 3-D vision and other automation advances help spur manufacturing productivity
By Patrick Waurzyniak
Robotic 3-D vision technologies, overhead gantry robots, and other developments in robotics and automation gear are enabling new ways for manufacturers to boost factory efficiencies.
With newer 2-D and 3-D vision-guided robotic applications, more sophisticated techniques for difficult tasks, such as bin-picking and precision assembly, are enabling broader application of robotic vision technology. Similarly, vertical articulated robots mounted on overhead gantry rails can greatly expand the reach of robots used in material-handling capacities.
Vision-guided robots are gaining in popularity as improvements in the technology make robotic vision an inviting choice for manufacturers. "On one of our robot lines, we're currently shipping 25% with 2-D vision," says Dick Johnson, general manager, material handling, Fanuc Robotics America Inc. (Rochester Hills, MI), "and 3-D is a smaller number, but growing."
At the International Robots & Vision show in June, Fanuc demonstrated several vision-capable robots performing complex tasks using 2-D and 3-D vision. In one interactive display, a Fanuc M-710iC/50 robot equipped with Fanuc's iRVision 3-D laser-vision sensor demonstrated an advanced precision assembly operation with attendees attempting to insert a part into an off-angle hole with a keyed shaft, a difficult application requiring high tolerance. The robot used 3-D vision to see randomly located parts, and the system performed the part insertion using Fanuc's FS-60 six-degree-of-freedom force sensor. "The theme was that robots can do precision assembly, with larger parts, and in 3-D space," Johnson says. "That's why we showed an angled, keyed shaft. SCARA robots were designed to do 2-D assembly and we think now, with the proper force control, we can do 3-D. Imagine a piston being inserted into an engine V-block.
"The way we do 3-D vision is more simple than some of our competitors," Johnson adds. "Our 3-D vision is comprised of a 2-D camera and a laser striper. The 2-D camera, first of all, finds it in 2-D space—it gives us the X, Y, and roll. We angle the laser in relation to the camera so when it hits the part, based on where the stripe is located, we can tell yaw, pitch, and Z height.
"Imagine, if you will, if you had a laser stripe on a part and you moved it up, you would see it move across the part. If you tilted it, you would see the X axis modified. This creates a relatively simple way to find a part in 3-D space. It was interesting to me that recently two of the vision companies announced products using a 2-D camera with a laser stripe."
Other vision system suppliers have had 3-D systems that deploy more than one camera, Johnson notes. "Actually, Braintech uses the Cognex underlying software, and then they add software to it to make it 3-D. They're using either two cameras, or they're using one camera with two snaps to find a part in 3-D space. Some companies are using one camera repositioned to take two shots, which is time-consuming, or two cameras. I've actually seen systems with three cameras to find an object in 3-D space. We use a single 2-D camera with structured light to easily find the object in 3-D space."
Manufacturing process improvements from robotic vision are significant, according to Joe Campbell, chief operating officer for Applied Manufacturing Technologies Inc. (Orion, MI), an engineering consulting company which held its Emerging Automation event at Applied Manufacturing Technologies' headquarters in September. "Three-dimensional vision is becoming very real. We're doing a lot of work in sheetmetal—we're doing vision-guided robotics, feeding of presses and press lines, racking and unracking of sheetmetal parts. It's a great place to take cost out, and the technology's getting substantial enough to really stick inside the plant."
Several automation suppliers, including ABB Automation (Norwalk, CT), Fanuc Robotics, Kuka Robotics Corp. (Clinton Township, MI), and Motoman Inc. (West Carrollton, OH) demonstrated robotic vision products at Applied Manufacturing Technologies. "It's becoming reliable," Campbell notes of 3-D vision. "It's truly deployable—you can put it in the factory and not have to be nervous at night."
Today's manufacturers are looking at new ways to wring out process improvements to improve operations. At one of Applied Manufacturing Technologies' customers, an automotive supplier in Michigan making stamped metal and welded assemblies for some of the Big Three OEMs, couldn't keep up with production demands with its installed automation. "They bought a bunch of used robots and equipment, deployed them, and at the end of the day they can't make cycle time," Campbell recalls. "Because of that, they can't produce enough parts and were falling further and further behind. They called us and we went in with what we called a 'cycle-time SWAT team' with methodology to assess the situation. We pick five areas for improvement, and we simulate them off-line because the key part here is they can't stop making parts, not even for an hour."
Simulating the robotic programming helped the customer, who did not wish to be named, gain an initial 7–8% improvement, Campbell notes, with a goal of eventually gaining 18–20% improvement in cycle times. "The bottlenecks were in robot programming," he explains. "The easy ones were motion type accel and decel parameters, and a couple things in process order, what order you do things in the cell. We have a multiphase follow-up longer-term. There were layout issues, which is a real common thing you find in simulation—the fastest cycle time may not be the shortest distance between two points. It just depends on how the cell's laid out, and on what kind of robot and motion you have."
Among the 3-D vision demos shown at Applied Manufacturing Technologies, Fanuc Robotics' full 3-D system picked up a stamped part in a semi-structured environment, Campbell recalls. "They do a couple things. First of all, they grab a 2-D image and extract features, like hole centers and dimensions, things like that, then they fire a laser cross-hair on the part, and they extract another round of now 3-D information from that laser image. They put the two of them together, and it's robust."
More robotic vision systems from other suppliers recently were announced, including Motoman's Motosight3D system, which includes vision software and hardware from robotic software developer Shafi Inc. (Brighton, MI) and components from vision systems supplier Cognex Corp. (Natick, MA). The system uses Shafi's Reliabot PC system for 2-D and true 3-D robotic vision. Aimed at complex vision-guided applications, such as bin-picking with randomly located parts, the Motoman 3-D system is said to save manufacturing costs by eliminating the need for custom dunnage trays or locating devices to repeatably position parts.
The Motosight3D system supports true 3-D (X, Y, Z, yaw, pitch, and roll) with one, two or three cameras, without the use of range sensors or lasers, according to Motoman. The single-camera option requires a robot-mounted camera and multiple inspections. With the multiple-camera option, the cameras can be mounted on the robot or at a fixed location, allowing for greater flexibility and accommodating irregular part shapes. A customizable HMI allows the user to change system parameters, calibrate the system, and see real-time inspection results. Including an intuitive, easy-to-use operator interface and customizable menus, the system supports fixed-mounted or arm-mounted off-the-shelf cameras. The software supports multiple coordinate systems and provides statistical production data.
Improving robot accuracy has also helped with vision system applications that demand precision. "Robots have always been repeatable, and that's great if you're doing teaching and repeating," notes Fanuc's Johnson. "Accuracy is the future, and we have a variety of tools that make the robot more accurate. Repeatability allows you to go back to a taught point. Accuracy allows you to command the robot to go to a point in space. That's becoming more important when you do vision guidance, when you do CAD-to-path programming, and when you do part cloning. When you're teaching [robots] one of six identical parts, then you provide the offset information and have it repeat that same path on the other five parts—some people call that cloning."
Application-specific software helps enable robot builders to develop very customized systems to suit the specific needs of different industries, notes Kevin Kozuszek of Kuka Robotics. "The robotic world is a cyclical kind of industry. You have advancements in the actual hardware, let's say robot arms for the packaging world, where you're developing very packaging-specific arms, and the counterpart to that is the software," he says. "You get very advanced robotic arm technology that almost outstrips the capabilities of the software, and then you can have a reversal, where you start developing their software capabilities and soon your software is not maximized to the robot arm.
"What we're seeing is end users are looking for software that's easier to use," Kozuszek says. "You're seeing more very intuitive graphical user interfaces, and the end users really don't care so much about the zeros and the ones that make it work; what they're truly looking for is 'what button do I push?' That's really become the trickle-down of robotics and automation in general that we see. In the automotive world, companies that started with large engineering groups of 100 engineers, and it's down to where a manufacturer might have only 10 engineers."
For material handling, Kuka also earlier this year announced its 1000-kg payload Titan robot, which is said to be the industry's highest available payload for six-axis robots. Aimed at automotive applications requiring large payloads, the Titan can lift materials weighing more than a ton, making it highly suitable for automotive powertrain or auto body and chassis assembly tasks.
"It's a pretty big hardware advance, because of what it allows end users to do with a robot," Kozuszek says of Kuka's Titan robots, which have been installed by integrator Grenzebach Corp. (Newnan, GA) at a commercial glass-handling facility. "In the past, to get that kind of payload capacity, you would need to use multiple robots or technology where two robots are synched together moving something jointly. This robot reduces your footprint. It can pick up car frames and do Bodyin-White, which is a big step because floor space is not cheap—if you maximize, it's that much better."
Overhead gantry-based robots like Fanuc's top-loading six-axis R-2000iT robot also are making headway in material-handling applications due to lowered cost and improved capabilities, notes Johnson. "I would say the powertrain area has been dominated by linear gantry devices for 20-plus years. But robots have increased in speed and payload, and we've decreased cost. In this particular case, for powertrain including automotive engines, a big advantage is the fact that our robots have full articulation. If you want to pick up a block and rotate it upside down to dump out the coolant and chips, we can do that easily, and actually it's our standard product. A linear gantry must add additional hardware to do the flip."
With vertical articulated robots mounted on gantries, manufacturers can greatly expand the reach of robots used in material-handling capacities, Johnson adds. "Because our robots have full six-axis articulation, it gives us X, Y, Z and yaw, pitch, and roll. A linear gantry has what you give it—certainly, X and Z and roll, but not full articulation. Typically, they're missing two or three axes.
"Something else that we take for granted, but apparently is an issue with the linear gantry: If you have a series of machine tools, and you're doing operation 10, op 20, op 30, oftentimes a robot will be transferring a product from op 10 to op 20, and then a different one from op 20 to op 30, etc.," he says. "What you need to do with the linear gantry is laser-sight all the machine tools so they're perfectly in line, because without the additional articulation, they need to be perfectly aligned so they can pick the component and carry it to the next station."
Depending on the size and payload, an articulated robot can be competitive with linear gantries, says Johnson, noting that Fanuc Robotics makes both top-loading articulated robots, and also underslung robots that are mounted below the gantry rails. "On a case-by-case basis, we are close to parity," Johnson says. "When we first brought out the top loaders, I remember one of our engineers mentioned "It's amazing how far we've come with the robot's feet nailed to the floor." Now, with the linear overhead rail, we not only free up the floor space, but we also can extend the reach of the robot almost indefinitely. We have made rails that were 70-m long."
More space-saving automation is available with the new integrated drive and motor units from Bosch Rexroth Corp. (Hoffman Estates, IL). With the Bosch Rexroth IndraDrive Mi compact drive and motor combination, manufacturers can reduce space used by 50% over comparable conventional servo solutions, and as much as 30% less space than other integrated solutions.
With the IndraDrive MI, electronics for the drives and amplifiers are brought down to the machine level. "This was tried before and there were a lot of failures, especially with bringing the electronics down to the temperature of the machine, and with vibration," notes Rami Al-Ashqar, product manager, Electric Drives & Controls, Bosch Rexroth. "With this product, it's proven to be very successful. You get rid of a lot of those issues with this technology because it goes straight to the first motor, and you're daisy-chaining the motors together."
This article was first published in the November 2007 edition of Manufacturing Engineering magazine.