In football, the fastest, strongest runner in the world isn’t worth a hill of beans until you get him the ball. It’s the same thing with an assembly robot. Although they can be incredibly fast, powerful and precise, getting the…er…ball to an assembly robot can often be highly problematic.
Historically, handing off work in process to a robot required that a part, or parts be presented exactly the same way time and time again. Because even the best robots were restricted to following a single, preprogrammed path, the slightest variation could cause serious problems. With the advent of advanced machine vision cameras and recognition software, however, robots are increasingly able to sort out parts of their own.
Ultimately, there are two basic ways of presenting a part, or parts: via some kind of flexible system or hard fixturing. However, today’s technology is such that even in these two areas there can be substantial overlap.
“It’s important to look at each situation individually,” says Nishant Jhaveri, product manager at FANUC Robotics America Inc. (Rochester Hills, MI). “Many issues come into play.”
“This is an area that is always specific to an application,” agrees Phil Baratti, applications engineering manager for EPSON Robots (Carson, CA).
Among the factors that determine what technique, or techniques will work best are cost, part size and shape, and cycle times. According to Baratti, the tolerances required by a process or assembly also play a role.
“The type of process, industry, cycle time, part size, characteristics of the part and number of tasks to be performed are all considerations that should be looked at,” agrees Chad Henry, applications engineering manager at Stäubli Corp. (Duncan, SC).
Another important consideration is the anticipated length and volume of a production run. Traditional hard fixturing is expensive and generally only makes sense if you plan on using it for a while. Machine visions system can also be pricy. But in contrast to hard tooling, they can easily adapt to product changes.
Along these same lines, hard tooling can be problematic when processing multiple product types on a single line.
“If [the product] is always going to be the same, then dedicated fixturing will make sense,” Jhaveri says. “If, on the other hand, you are going to change part styles every two or three months, then things change.”
Bowl in HandTraditionally, vibratory feeder bowls have been one of the most commonly used means of presenting parts in robotic applications, a trend that remains true to this day. With this kind of system, vibrations cause the parts to first line up in and then travel along a spiral track running up the bowl’s inner surface.
The singulated parts then continue on to an escapement and placement mechanism where they can be individually picked up by the robot. At some point along the line, incorrectly oriented parts will either fall back into the bowl or be physically removed, so that only those parts that are correctly oriented are allowed into the queue.
The advantage of this kind of a system is that it is fast, with cycle times of as little as a second. Feeder bowls are also precise. Once a feeder bowl is up and running, it pretty much guarantees the parts will be where you want them correctly oriented.
On the down side, feeder bowls are fairly expensive, complicated pieces of equipment. They are also not very flexible, due to the fact that they are generally equipped with part-specific tooling that has to be either replaced or reworked in the event there is any kind of a product change.
“The advantage of a bowl feeder is that you can rely on having parts presented and ready for picking in a very short time,” Baratti says. “[However], the complexity of a bowl feeder is dependent on the complexity of the part, and the feeder is usually a hard tooled, fixed function device. Changeovers are possible, but costly and raise questions of reliability.”
Similar to feeder bowls in terms of speed and reliability is magazine, or gravity feeding. In this approach, the magazine-fed parts are captured as they are being manufactured and put into tubes for further processing-a task usually performed by the parts supplier. Because magazines are much simpler than feeder bowls, they are substantially less expensive. They are also more flexible in the sense that it costs less to modify or replace an existing magazine than it does a feeder bowl system.
However, there is still the question of getting the parts into the magazine in the first place. The added expense can become especially daunting with larger production runs, due to the fact that the cost per part of an already singulated and oriented part will inevitably be more than that of one supplied in bulk.
Nesting InstinctAnother presentation method requiring hard tooling is pallet-based nesting. This approach can include fixtured pallets with a single nest, a couple of nests or dozens of nests, depending on the application.
Single-nest or double-nest configurations are often found in assembly lines employing pallet-based conveyors. They can also be used on a rotary-dial indexing table, in which the robot is working at one of a number of workstations.
In either setting, work in process is placed in the fixture at the head of the line and remains there until the end. As it enters the robot’s work area, the pallet is stopped, located and even elevated if necessary.
A “smart” conveyor equipped with an RFID or code-reader system can be used to tell the robot what kind of component is being processed in the event multiple product types are being assembled on a single line. When the robot is finished, the pallet continues on to the next workstation.
Pallets with large numbers of nests are often used with freestanding automation systems, like those employing tabletop Cartesian robots. Using this approach, an operator loads the pallet with work in process, places the pallet on the work surface and then pushes a button to activate the robot. As the robot is doing its work, the operator can either load another pallet or perform some other task.
An advantage of this approach is that it ensures each workpiece is precisely located for processing. With this in mind, pallets with multiple nests are often used in dispensing applications in which the robot is tracing out a complex sealing path. Pallets with multiple fixtures are also used with Cartesian systems in which the robot is driving multiple threaded fasteners.
Of course, presentation speed is a non-issue, because the various parts are all within reach of the robot at the same time.
On the down side, the fixturing itself is generally very product specific. A single nest may be able to accommodate more than one type of, say, cell phone housing. However, this will only be possible if the two variants are extremely similar and share identical locating features.
Another disadvantage is that the pallets need to be manually loaded prior to processing and then unloaded when the robot is finished. There will inevitably be some system down time as the operator is changing out pallets.
Having VisionCommon to all the techniques discussed thus far is the fact that the robot follows a preprogrammed path every time it picks up a part. This is possible because work in process is always located in the same spot. By employing a machine vision system, on the other hand, an assembler no longer has to worry about the time and expense required by things like singulation, because the robot can start doing the sorting on its own.
Currently, truly random picking, in which dozens of complex parts are simply jumbled together in a pile, remains both problematic and slow. The algorithms required are highly complex due to the variables involved.
However, the sorting of “semi-random” parts is a now a proven, mature and highly robust technology that is being used in plants worldwide in everything from the automotive to the consumer electronics and medical device industries.
Relatively flat parts randomly distributed across a belt conveyor or vibratory tray are especially easy to identify and pick, even at high speeds. A prime example of this kind of application is one in which a high-speed SCARA or delta-style robot is used to identify and pick thin solar wafers for further processing.
Along these same lines, Daniel McGillis, global business development manager for robotics at ABB Inc. (Auburn Hills, MI) say vision-equipped robot are also increasingly being use to unload multiple layers of flat parts directly from a bin. The classic example of this kind of system is one in which the robot unloads a box of gears, with each layer of gears separated by a thin partition.
As is the case with the photovoltaic wafers travelling on a conveyor, the gears all lie flat, but their specific rotational orientation is completely random. The robot can even be programmed to remove the partitions as each layer is emptied.
Again, the advantage of this kind of approach is that it saves the assembler or supplier the time and trouble of having to precisely position the parts in space. Another advantage is the system’s flexibility. In contrast to those techniques requiring hard tooling, an existing robot and vision system can be quickly adapted to a new product through reprogramming.
On the downside, machine vision systems represent an additional cost. They can also be tricky to implement effectively, with lighting, in particular, often causing headaches. Dirty or smoky environments, like those in a foundry can be especially problematic.
That having been said, today’s robots are often preconfigured to accommodate machine vision, making it easier than ever to install. Cameras, lighting systems and identification software are also much more affordable than in years past, not to mention far more robust.
FANUC’s Jhaveri adds that assemblers can further justify the added cost of a vision system by using it to perform more than one task. For example, a single system can be used both to locate parts and perform inspections, in essence killing two birds with one stone.
In terms of speed, vision-guided robots are not quite as fast as their hard-wired counterparts picking parts from hard-tooled fixtures. There is also the question of how many parts a vision-guided robot has to pick from at any given time-largely a function of its field of view. However, the two remain comparable.
“Using a vision-guided robot for bulk singulation can yield a cycle of one second per part, but it’s rare,” says EPSON’s Baratti. “ A two-second cycle is far more reasonable.… The most important cycle time factor of bulk singulation with vision-guided robots is the number of pickable parts available to the robot at any given time. With a large field of vision, you have more parts to pick from, but you lose precision. With a small field of vision, you have less parts to choose from, which has a direct impact on cycle time. ”
Baratti notes that in many cases, assemblers will solve these kinds of problems by equipping a single robot with two cameras: a fixed overhead camera with a large field of view to initially locate the part; and a second camera with a narrower field of view on the robot itself, which allows it to home in precisely on the part immediately prior to actually gripping it.
Mixing Things UpNo matter what the specific technology or approach an assembler uses, the main thing is that it fit the needs of the application in question. If speed and precision are absolutely paramount, then a traditional feeder bowl may be the way to go, especially in the context of a long production run. If, on the other hand, a company anticipates more moderate throughputs, it may be better to go with a hand-loaded pallet with multiple nests.
“All parts presentation methods, regardless of the task, need to be constant,” emphasizes Stäubli’s Henry. “The specific application will determine the best presentation method.”
Jhaveri adds that one of the best things about today’s technology is the way it can be mixed and matched to create a hybrid systems that takes advantage of the better aspects of multiple parts presentation techniques.
Just recently, for example, FANUC created a system employing multiple robots to both gather up and then assemble the parts making up a crank and connecting rod assembly.
As a first step, an F-100iA robot picks up a tray of cranks and rods from a storage rack and then presents it to the six-axis LR Mate 200iC robot that will perform the actual assembly. At the same time, a second vision-equipped LR Mate robot picks a number of crank pins and bushings from a nearby vibratory feeder and also presents them to the assembly robot. When the first LR Mate has finished assembling the crank, a second F-100iA system removes the completed assembly to a storage rack.
Along these same lines, Baratti says that in recent years a number of assemblers have begun using machine vision in conjunction with traditional feeder bowls. The idea here is to use the feeder bowl to take care of singulation and let the robot do the rest.
“Instead of investing 80 percent of cost into the last 20 percent of orienting the part, companies are now making the feeders general singulation devices and using vision to determine the exact position and orientation of the part,” Baratti says. “This allows for lower costs, higher reliability, and some amount of flexibility. ”
Finally, in some instances, an assembler may be able to forego parts “presentation” entirely. For example, Stäubli’s Henry says an increasing number of manufacturers are employing machine-tending robots both to unload newly stamped, molded or machined parts directly from the equipment that created them and then do some additional processing of their own.