Using robots for assembly applications has never been easier, thanks to a new breed of software, sensors, controls and machine vision systems. But, even with the most advanced technology, mistakes still occur and problems arise when using robotics. Indeed, there's a right way and a wrong way for every application.
After several years of sluggish growth, demand for industrial robots is on the rise. The Robotic Industries Association (RIA, Ann Arbor, MI) claims that new orders from the North American market increased 13 percent during the first 9 months of 2004.
And, that healthy demand is expected to continue throughout this year as manufacturers seek new ways to boost productivity, improve quality and lower costs. A recent study conducted by the United Nations Economic Commission for Europe (Geneva, Switzerland) predicts that worldwide demand for industrial robots will grow 7 percent annually between now and 2007. Most of that growth will be driven by manufacturers in Germany, Japan and the United States.
However, as more and more companies turn to robotics, new problems are likely to occur. Of course, traditional mistakes, such as having unrealistic expectations, misunderstanding the difference between accuracy and repeatability, or specifying the wrong type of end-of-arm tooling, will continue to cause headaches.
Misapplication will also continue to be a problem. "It is not uncommon to see a six-axis robot being used to perform a task that a two-position pick-and-place would be capable of performing," says Dave Capindale, application engineering manager at ATS Automation Tooling Systems Inc. (Cambridge, ON). "This adds significant cost to the automation, not only for the cost of the robot, but for the control system needed to make the robot both functional and safe."
Fortunately, many of those problems can be avoided if engineers simply do their homework and follow some of the tricks of the trade suggested by various robotic experts.
Michael Perreault, vice president of Midmac Systems Inc. (St. Paul, MN), believes the biggest mistake engineers make with robotics is underestimating the variability of the product being assembled, and any quality issues that may be associated with the application. "Typically, the application being automated is already being accomplished by an operator," he points out. "Often, this operator makes adjustments for inadequate parts by compensating for the deficiencies of a part, and makes that part work within the specific assembly process."
According to Perreault, an operator will make judgment calls if the parts are out of spec or if they have surface texture issues. "Operators will make these assemblies work by adding extra finesse, or by determining if the parts are of acceptable quality," he explains.
Robots can't do that. Parts that were useable for human operators may be unusable for robots. As a result, engineers must clearly define the range of part variances so the assembly system can identify good parts. Engineers may also need to tighten part specifications.
Engineers should consult operators about how they identify good parts and what special techniques they use during assembly. By asking operators to document any areas that may require special attention, engineers can establish parameters for the automation cell. Sometimes, operators are not aware of all the activities that they perform. In that case, it may be worthwhile to videotape assembly activities that appear to require a higher level of attention by the operator.
"When implementing a robotic assembly system, it is generally advantageous to include the personnel who are performing the tasks," explains Perreault. "Encourage them to identify [all the] requirements and special complexities [of the assembly process]. By getting a buy-in from the associated staff, you will be much better informed in creating a total solution for the application."
Don't Forget Support
As the performance of robots has improved over the years, and the experience level of system integrators has grown, many earlier problems have disappeared. "To the credit of integrators, robotic systems are deployed into factories with a very high initial success rate today," notes Keith Bailey, director of product marketing at Adept Technology Inc. (Livermore, CA). "Systems are delivered on time and run-offs often go flawlessly."
Bailey says many problems today start well after the installation. Any part changes or worn tooling forces engineers to re-teach a robot position or change the robot program. "Those can be simple changes, if you are proficient at robot programming," says Bailey. "But, more and more these days, manufacturing engineers are responsible for many production lines, and are not allotted the time to become proficient in the operation of each device on their lines."
"The No. 1 mistake we see today in applying robots is the failure to match the technology with the capability of the factory support personnel," adds Bailey. "Robot technology, as with any technology, only works if it can be supported. In the case of manufacturing equipment that can run around the clock, the support needs to be resident on the plant floor during every hour of production to minimize downtime."
Different plants have different skill sets available on the plant floor. For instance, some support personnel are comfortable with personal computers and PC programming. Others know programmable logic control (PLC) or ladder logic programming. Some prefer only menu-based "point-and-click" systems.
Bailey says the widely varying skill sets on the plant floor have driven his company to take "a unique approach to help solve this challenge." Its robots can be ordered with any one of several different software environments, thereby allowing the system integrator to select the environment that best matches the capabilities of the plant floor support personnel.
The robots can be programmed in Allen-Bradley PLC ladder logic, Adept's V+ programming language, or a point-and-click environment called AIM. "Each of these software environments have their own strengths," says Bailey. "For example, in the automotive industry, where PLCs are common, CobraPLC is an excellent choice. The CobraPLC robot is extremely easy to learn if you are an electrician or a PLC technician. Since all programming and setup is accomplished through the PLC, technicians can proficiently re-teach points or change programs with very little training."
In other industries, such as semiconductors, Bailey says it is common for engineers to provide plant floor support. They may prefer a PC-based programming environment.
"The key to a successful robot installation lies in the ability to support the robot software with onsite personnel," says Bailey. "Selecting the software environment that matches the skill set of the plant floor support personnel will help ensure the long-term success of the system."
Don't Overlook Fixturing
Fixtures are an important, but sometimes overlooked, aspect of robotics. Without fixturing to maintain repeatable part positions, a robot cannot operate effectively.
Kent Walker, applications engineer at Rixan Associates Inc. (Dayton, OH), says the biggest mistake made by manufacturing engineers is to implement a robot with fixturing that does not accurately locate parts.
"Typically, people do not realize [the importance of fixturing] because the current process may be a manual operation," explains Walker. "Parts and components need not be accurately located if the operator is manually assembling or locating parts."
To avoid this problem, Walker suggests approaching robotic applications from the component point of view. "Realize that a robot is essentially a dumb device which will go where you instruct it every time," he explains. "This means the items you are processing must be in the same place every time." Well-designed fixturing maximizes system performance and reduces the potential for problems.
It's also important for engineers to understand the tolerance of the components. "Try to visualize and understand every step that an operator performs and try to implement the same steps with robot motion," says Walker. "Understand the robot will perform every task as you indicate, the same way, the same sequence."
Idle Time Is Evil
Many manufacturing engineers fail to consider the real cost of idle time. "Idle time is any time that a robot is idle and not performing a value-added maneuver," explains Aaron Odham, applications engineer at ATI Industrial Automation (Apex, NC). "This time must not be confused with downtime, which is the direct result of the failure of a component within the assembly system."
Many factors add to robot idle time. Some are preventable and some are not. "This mistake is made because of a lack of experience in implanting robotics in an application, as well as manufacturing engineers being pushed to make assembly systems as cost-effective as possible," says Odham.
Preventable contributors to robot idle time include inflexibile end-effectors, downtime from crashes, and lack of spare parts or components. "If these preventable items are addressed during design and implementation, robotic assembly applications will be more profitable and more successful," claims Odham.
One way to achieve the basic goals of lean manufacturing, such as making higher quality products with less waste and for less cost, is with the assistance of robots. "Robots are quicker, work longer hours and are more consistent than humans," says Odham. "As long as a robot has parts to assemble, it will assemble them quickly and correctly. However, a major problem arises when there are no parts to assemble-idle time. Replace this robot with a human and the human very quickly finds another value-added task to perform until these parts are replenished.
"The reason humans have the advantage over robots is because they are flexible and adapt to their surroundings without being told," adds Odham. "This, of course, is not the case for robots, but they can overcome the small detail of not being able to think for themselves and become very flexible, productive machines."
Odham defines "flexibility" as a ready capability to adapt to new, different or changing requirements. Although robots cannot react without external input, it is easy to add sensors and other external inputs to enable them to react to changing conditions.
Tool changers allow robots to adapt to changing conditions. They allow robots to quickly change from one tool to another tool and from one process to another. "By increasing the number of processes a single robot can perform, the robot becomes more productive and can continue to perform value-added functions in the event one process is prevented," notes Odham.
Another major source of robot idle time is downtime from a crash. "Robot crashes are inevitable and it is not a matter of ‘if' the robot crashes but ‘when,'" says Odham. "Due to the extremely high forces of impact during a crash, the tooling is often damaged and requires replacement. If the tooling is not attached to an automatic tool changer, this repair can take hours and be very costly.
"Mechanical collision sensors are a great tool for sensing a crash and absorbing the impact energy from the crash," adds Odham. "A good collision sensor should be pneumatically variable and have repeatability of no more than +0.001 inch. It should offer an automatic reset option so as the robot is removed from the crash, the end-of-arm tooling automatically returns to its working position without a human entering the robot cell."
Most robot manufacturers offer "soft crash" protection, which is software that monitors the drive servos and can detect if the robot collides with an object. "This system is great, but it cannot absorb the impact energy created during rapid deceleration of the tool once it contacts a foreign object," warns Odham.
Mechanical collision sensors work in conjunction with software collision detection systems offered by most robot manufacturers. The mechanical collision sensor absorbs the impact energy created by the crash and sends a signal to the robot informing it of the crash. According to Odham, a mechanical collision sensor is "an airbag for your end effector." By allowing the robot and end-of-arm tooling to withstand a crash without severe damage, collision sensors prevent idle time.
Spare parts are another often-overlooked part of the productivity equation. "As with any mechanical application, there will come a time when a failure occurs from fatigue, damage from a crash or other factors," Odham points out. "The ability to replace the damaged part and resume operation with little delay is extremely important in minimizing idle time."
However, managing the stock of spare parts is equally important. "Spare parts should be stored in a safe location, should be readily accessible and in working condition," explains Odham. "Lost spare parts are not uncommon. A well-stocked spare parts inventory gets the robot and the application back to work as soon as possible."
Have Clear Expectations
Phil Baratti, manager of applications engineering at Epson Robots (Carson, CA), says the two biggest mistakes he sees are unclear specifications and unrealistic expectations.
When writing an RFQ, engineers should include specific data on throughput, precision and potential causes of downtime with the application. "[Integrators should not be] expected to account for every variation that may come along," notes Baratti. "Nor are they expected to account for every fault condition that's possible."
The throughput requirement is directly related to the type of system that will be built. Throughput determines the product mixes, manufacturing variations, repeatability, tolerances, and changeovers that can be accommodated by a single piece of equipment, Baratti explains. Engineers need to consider these factors up front to determine the scope of the machine that will be built.
Similarly, the precision specifications will determine the tolerances for each part of the assembly and the accuracy requirements for each machine in the assembly system.
Ideally, machine specifications should cover five basic requirements: Safety, precision, throughput, repetitive motion and quality. "If the design spec addresses one or more of these topics, you can usually begin to narrow down what you need to build for, and what the expectations are," says Baratti.
"For instance, if precision is an issue in building a certain product, there's a good chance that you'll be looking for some type of vision guidance," he points out. "This is because the parts, assemblies or tolerances are so small that people have difficulty doing the assembly on a continuing basis. If the assembly also requires high throughput, you may be looking at a high-speed assembly device, but use the vision for postassembly inspections instead of inspecting during assembly. Of course, this also assumes that the yield rate is acceptable, and not excessively costly to complete the assembly."
According to Baratti, many first-time robot users simply don't have enough answers to their assembly questions. "It's very difficult to identify the ‘unknowns' of automation," warns Baratti. "You don't know what to ask, so you don't know what you're missing."
To avoid this problem, Baratti urges manufacturing engineers to "do your research and ask questions until you understand what and why." And, he says, "Don't bite off more than you can handle. Start with a clear production problem that is ready to be automated."
Don't Overestimate Throughput
"The biggest mistake in deploying a robot depends on the nature of the application," says Tom Jerney, program manager at Remmele Engineering Inc. (St. Paul, MN). For example, engineers often overestimate the throughput of high-speed assembly applications. They don't consider how the inertia of the payload and gripper will affect settling time, or they don't account for nonsynchronous feed rates, which make the robot wait for eligible parts or a place to put them.
"This mistake can be avoided by doing a realistic mock-up of the cell and having the robot vendor demonstrate actual throughput before building the production system," says Jerney. "It may be necessary to derate the throughput to provide some margin for unknowns."
For complex applications where the robot is integrated with other intelligent systems, such as machine vision or external servo positioners, a common mistake is to select a robot whose controller is not easily integrated with other controls, creating a need for engineering custom interfaces and software. This can be avoided by selecting a robot system-robot, controller and software-that can directly control other parts of the cell or easily interface with them. Jerney recommends asking robot vendors to demonstrate the connectivity of their systems.
According to Jerney, the biggest mistake new robot users make is deploying a system that is too complex for them to operate and maintain. "The system provider should make a realistic assessment of the technology that the customer can support, and simplify the system to a level that is consistent with the local resources," explains Jerney. "That may mean staying with air cylinders, relay controls and other simple devices."
Don't Overpay for Speed
"Different applications have different challenges and thus different mistakes," says David Fiedor, sales coordinator at Janome Industrial Equipment (Elk Grove Village, IL). "For example, on dispensing applications, one of the largest mistakes made by engineers is overpaying for speed on a robot.
"With a dispensing application, the more critical specification may be [accuracy and repeatability]," adds Fiedor. "Does it really matter that the robot you purchased has a standard cycle time of 0.39 second or that it can move at 4,200 millimeters per second? You typically don't need the speed when dispensing."
Fiedor believes that accuracy is more important than speed or cycle time. "A lot of robot manufacturers try to sell speed when that is not necessarily relevant to the application," he warns. "The way this can be avoided is to not be sold on speed and focus on what is most critical for the application you are trying to complete."
According to Fiedor, another common misconception is that a PLC is needed for all robotic applications involving external equipment. "This is not the case in all instances," he points out. Some robots feature built-in sequencers that act like a PLC.
"It is a separate function that is constantly running in the background, independent of the individual programs," says Fiedor. "One-line logic commands can be written such as IF, THEN, ELSE statements."
Through the sequencer function, a few part present sensors, a light curtain, and a camera for vision inspection can send and receive commands from the robot. However, PLCs are needed for complex applications or applications that require large amounts of external equipment.
Remember: Safety First
Manufacturing engineers should never overlook or underestimate safety when implementing robotics. Unfortunately, the latest robot safety standards have not been fully implemented in many assembly applications, says Mike Hancock, controls systems hardware design manager at ATS.
"For example, a start-stop circuit would traditionally be designed with a single channel circuit," notes Hancock. "In this type of circuit, a single component failure can cause the loss of the safety function. Because of this, manual or automatic checks must be ensured at specified intervals and whenever the state of the system changes to ensure a degree of safety."
According to Hancock, today's robot safety standards emphasize a "control reliable" approach for safeguarding control circuits based on a dual-channel input and output design. With this type of design, the safety interlock on a safeguard has two contacts from the keyed interlock switch that are positively actuated (physically forced open) when the safeguard is opened. Redundant inputs (contacts) are monitored by the safety relay or controller and the safe state is driven by two output channels for redundancy.
The output channels individually drive contactors or force-guided relays whose contacts are wired in series to the hazardous devices, while their normally closed contacts are monitored for reclosing in the reset or on circuit.
"This methodology provides redundancy and monitoring in the safeguard circuit and is only now common through many standards [covering] situations of a moderate to high degree of risk," claims Hancock. "The latest robot safety standards [require] control reliable methods that have gained roots from EN 954-1 Category 3 and higher performance.
"Most robot integration mistakes are the result of inexperience, poor planning, and a lack of commitment by [end users] to keep informed of the latest innovations in robot technology and changes to national safety standards," adds Hancock. This can be avoided by thoroughly evaluating the robot application early in the design stage of the project and re-evaluating it at key points as the project progresses. "Incorrect implementation of safety standards can only be avoided if companies make a commitment to ongoing education of all employees involved in robot integration," Hancock points out.