Robots: More Capable, Still Flexible
Today's robots possess a range of capabilities that can solve shop-floor manufacturing problems
By Erik Nieves
Senior Manager of Technology Advancement
West Carrollton, OH
Recent advances in technology now allow industrial robots to perform more, and different, applications than ever before. Some of these changes--like the advent of multiple robot control and application-specific robots, plus advances in vision-guided robot technology, connectivity enhancements, and improved laser seam-tracking and weld inspection--are available now for application in the production line.
Manufacturers looking for ways to streamline their processes and become lean invest in flexible automation that allows them to do more with the same equipment--while also using minimum floorspace. Multiple robot control is a trend that fits in well with these lean-manufacturing concepts.
Multiple robot control can be used in workcells dedicated to a single process, or in cells where the robots perform multiple processes. For example, due to built-in collision-avoidance features, multiple robot control enables up to four robots to simultaneously weld on a part, thereby reducing the number of stations or cells required for production.
Multiple robot control is also useful in applications where one handling robot serves as a positioner to manipulate parts for as many as three processing robots. Any of the robots can also automatically exchange end-of-arm tooling to allow additional flexibility within the same workcell. For example, a handling robot can switch jigs or a processing robot can switch from gas metal arc welding (GMAW) to gas tungsten arc welding (GTAW) or from welding to grinding. The result is a truly flexible workcell capable of processing completely disparate parts with different manufacturing requirements.
Multiple robot control can lower overall fixture costs associated with a cell. To weld large automotive components, for example, a handling robot can present a pre-tacked assembly to two arc-welding robots for finish welding. The handling robot can serve as a highly flexible positioner, locating the part in the optimum position and orientation for the different welds. Instead of a fixed two-station cell with two sets of expensive, complex tooling to locate all the components, the multiple robot control cell requires only one set of complex tooling (for pre-tacking) and can use a simple holding fixture for the handling robot.
When manufacturing automotive exhaust systems, multiple robot control can leverage the coordination between a handling robot and a welding robot to process highly contoured parts that require continuous welds around compound angles. Such parts require a complex tacking fixture tended by a second arc-welding robot. Although this approach can also lower total fixture costs, the main benefit is that the handling robot can present the final welding robot with a pre-tacked part unencumbered by clamps, sensors, or brackets.
Benefits of multiple robot control with a single pendant include built-in automatic collision avoidance, greatly simplified programming structure, reduced integration cost, and ANSI/RIA R15.06-1999 safety-standard compliance. Besides lowering capital expense by potentially processing the parts with fewer workcells, additional cost reductions are realized by eliminating redundancy in I/O controls. The multiple-robot controller can control all of the I/O for the workcell--a single fieldbus connection rather than a connection for every robot manipulator.
Today's robot controllers support established fieldbus standards, such as DeviceNet and ProfiBus, and many also support emerging standards, such as EtherNet/IP. This off-the-shelf connectability improves reliability, promotes uptime, and permits a more seamless information infrastructure throughout the manufacturing facility. High-level supervisory controls, such as SCADA systems, can have ready access to any of the robots on the factory floor via TCP/IP protocol.
In many cases, the controls engineer has a single connection to the robot controller, and through it can access all I/O, logic, data structures, and files. The single connection means less hard-wired infrastructure for the line. In fact, the same is true for all services to a workcell that employs multiple robot control: for example, a single power drop for all the manipulators, or a single Ethernet drop for the workcell. This ease of installation is a side benefit to multiple robot control that should not be underestimated, because it represents a significant cost savings.
The advent of multiple robot control highlighted the problem of cable interference in arc-welding applications. With manufacturers now controlling up to four arc-welding robots--sometimes working in a space no larger than a mini-fridge--cables created major interference, and also added complexity to the robot programming. Using general-purpose robots in arc-welding applications presented shortcomings that needed to be addressed through a new approach.
Some robot manufacturers allowed the process to drive the manipulator design, and developed application-specific or purpose-driven robots with built-in attributes that make them well-suited for arc welding. Changes include through-hole reducers, and an open-yoke design for the upper arm that allows the entire torch cable to be integrated and contained within the robot structure. Cable flop is eliminated, wire flip is greatly reduced, and interference between robots, torches, and fixtures is mitigated. New in-line torches allow access to welds that were previously not possible with offset torch mounts, and long welds and circumferential welding are no longer limited by wrist rotation restrictions.
The idea of an application-specific or "purpose-driven" robot changes the paradigm for robot design that has prevailed for more than 30 years. To achieve economy of scale in production, manufacturers created general-purpose designs, and robots were sold as multipurpose tools. As a result, robots were generic in that they were general-purpose manipulators equipped with different process equipment to allow them to perform various applications. The same robot model might be equipped with a mechanical gripper for material handling/machine tending, or a plasma-torch package for use in another workcell. Ideally, this approach provides manufacturing companies with flexibility if and when they might decide to redeploy a particular robot to perform a different application.
In reality, though, this sort of redeployment from one process to another just doesn't happen often. In the metalworking industry, large numbers of robots will always be used exclusively for arc welding. The products being welded might change, but the process will stay the same over the robot's service life. Despite advances in alternative metal-joining technologies such as laser welding and friction stir welding, arc welding continues to be dominant, and that fact is not likely to change. So, having a robot that is designed specifically for arc welding is an advantage.
The same process-focused design applies to spot-welding robots with equal success. Using a general-purpose robot for spot welding involves routing unwieldy power cables, water lines, and other services out to the wrist, resulting in bulky brackets, interferences with parts/fixtures/other equipment, and poor reliability. Moreover, for users programming their robots off-line, the inability to model the external robot dress was a serious detriment to the fidelity of the final program. Today, new robots are available that have been designed especially for spot welding. These new manipulators eliminate the external dress issues by routing all the necessary services along the upper arm in a controlled and reproducible manner. Users enjoy increases in the life of the dress package, improved uptime, and ease of programming--on and off line.
In general terms, robots are a sharp contrast to the hard-automation tools that preceded them. Hard tooling is designed for a single, specific application, and does it fast and very well. However, hard tooling makes design changes difficult and expensive. In contrast, robots were designed to be completely flexible in not only the specific task (for example, weld here versus weld there), but also in their application (weld today, glue tomorrow). The tradeoff, though, often was fitness to the task at hand. Now, with the development of purpose-driven robots, we have reached the intersection of hard-automation tools and flexible automation. We get the benefits of a machine designed for a specific application process, married with the flexibility of reprogrammability. So the application stays the same, but the tasks are infinitely variable to accommodate design changes and new products. This intersection might very well be the lowest total cost of ownership (TCO) for capital equipment.
Multipurpose robots still fill a broad spectrum of needs, and they aren't going away. The market for industrial robots has, however, grown to the point where differentiation in the products is both desired and possible.
Advances in machine vision have opened new applications for robots in material handling. Early vision systems required imaging of the parts to be identified and/or located to take place in an enclosed environment with controlled, consistent lighting to provide the necessary contrast and detail. Today's improved vision systems are no longer as constrained by lighting conditions. It's now often possible to have the system image properly by merely using a ring-light mounted around the camera.
This new portability is key for robotics. The robot manipulator can now carry the camera as part of the end-of-arm tool, and image features on a part anywhere within its work envelope. Robots use this scheme to locate stampings on conveyors, and parts in dunnage. Often the robot will take images of two different features of a large part to determine its skew. In this way, it's no longer necessary to trap or locate the part mechanically, eliminating much of the tooling cost associated with part-transfer machines. The robot then handles the part for stacking, processing, or racking. In racking applications, the rack itself is often not repeatable. Therefore, vision guidance is also required to locate the part-holding features.
In another significant advance, robot controllers can now integrate vision and line (conveyor) tracking to allow on-the-fly part picking. The part no longer need be stationary to enable the robot to pick it up. A high-speed camera snaps an image of the desired feature, and the vision system software calculates part offset and rotation. The robot controller takes the data, correlates the data to an initial encoder value for the conveyor, and continually updates the pick point for the moving part. Cycle-time reduction and improved work flow can be significant.
Improvements in vision-system processing--along with the high level of integration with the robot controller and adept gripper design--have been brought to bear on one of the most bedeviling of material handling applications: 3-D bin-picking.
In 3-D bin-picking applications, parts are layered in bins (or totes) without the benefit of dunnage. Some loose constraints still apply--parts in the bin must all be the same, or at least of the same family. Parts are generally layered as opposed to being completely random in the bin.
Blind robots require that parts be in a repeatable location and orientation in order to pick them up. Vision systems enable part pickup when other methods are costly and/or impractical.
However, 3-D picking is qualitatively different from standard vision applications in that single-camera systems usually only provide planar (X, Y, Q) information. Because the parts are on multiple layers in the bin, a second camera (or at least a second image), or some additional sensor must be integrated to measure the Z component. Much of the development effort today seeks to create easy-to-use tools to merge, or co-register, these multiple images and output the 3-D offset data. These efforts are bearing fruit daily as more and more robots are being deployed to load parts into machine tools from bins. The value proposition is clear. Dunnage costs money and it's designed for only one part, or at best, a family of parts, while bins are generic and can be used for almost any part.
Arc welding is among the most robotized applications in industry. Robots have been MIG-welding parts since the 1970s, but even so, until recently many parts still could not be welded robotically due to joint variation. Thin-gage parts with long lap joints (kitchen appliances, for example) were difficult because the joint would not repeat in space. Over the length of the part, the thin-gage material would not lay straight, causing deviations to develop along the joint. Traditional through-the-arc seam-tracking methods would not always follow the joint, because the lap configuration would not provide sufficient material for the weave to discriminate a change in welding current.
However, the last few years have seen an increase in the use of laser-based tracking for these types of applications. These laser sensors can track thin-gage lap joints without requiring a weaving motion. What's more, tighter integration to the sensor means the robot controller can modulate the speed and other weld parameters based on feedback from the laser system. Now, the same laser-based technologies are being applied to weld inspection. The robot can pass a laser over the welded joints, and the system can report on such variables as weld length, size, and undercut.
Emerging technology will increasingly allow robot systems to be used for many applications that were almost exclusively performed by CNC machines, such as complex cutting and material removal, grinding, mold creation, surface finishing, and drilling and tapping applications. Certain applications will always be better suited to CNC machines, due to accuracy and repeatability issues beyond the capability of robots. But many applications don't require the precision of CNC machines, and robots now have the positioning accuracy, repeatability, and rigidity (mechanical stiffness under load) required to process these parts.
PC-based off-line software programming tools now exist to translate large, complex programs written for CNC machine tools, including I/O and other non-motion commands, into robot programs that are ready to run. As a result, manufacturers can take advantage of the lower overall equipment cost and increased flexibility that six-axis robots can offer, as opposed to more expensive CNC machines with only 3 - 5 axes of motion. Robots can perform many of the same tasks more efficiently, with a faster and cleaner process that provides high throughput rates and great flexibility.
This article was first published in the May 2005 edition of Manufacturing Engineering magazine.