BY TANYA M. ANANDAN FEBRUARY 5, 2019
Cover Story: Software suppliers are disproving preconceptions about robot offline programming (OLP) being too complicated or not cost-effective. Several companies have developed software programs designed to make robots smarter, faster, and more flexible in a variety of challenging applications.
Courtesy: OCTOPUZ Inc./Robotic Industries Association (RIA)
Even though offline programming (OLP) has come a long way, misconceptions about it still linger. The old-school mindset that OLP for robots is not up to the task is one of skepticism. A lot of robot programming in the welding industry still is done manually, point by tedious point, with a teach pendant. Many still remember the old days of robotics that overpromised and under delivered. However, the dreamers and doers have been working to exceed expectations
Simulation and OLP software has evolved and become smarter, faster, more flexible, and reliable. This is a new era in ease of use and OLP software suppliers are here to demystify the softer side of robotics.
OLP vs. simulation
Most simulation and OLP solution providers will tell you it’s not one or the other. Simulation and OLP go hand in hand. Though the terms oftentimes are used interchangeably, there is a difference. You can have simulation without OLP but you cannot have OLP without simulation.
Robot simulation is the 3-D representation of a robotic cell or production line. It visually demonstrates how a robot moves along a path or trajectory from one XYZ coordinate to another XYZ coordinate. It can include multiple robots mounted on external axes working with multi-axis workpiece positioners, or coordinating on an assembly line. All this movement and planning, however, can get complicated.
Figure 1. Offline programming (OLP) software accurately simulates a robotic welding process using calibrated data for robot kinematics, external axes, and workpiece positioners. Courtesy: CENIT North America, Inc./Robotic Industries Association (RIA)
“A lot of customers purchase a robot thinking it will behave like a computer numerical control (CNC). This is not the case,” said Albert Nubiola, CEO of RoboDK Inc. “CNCs are easy to program. The workspace is properly defined. It’s like a cube. However, robots have a spherical workspace, and because of joint limits and robot singularities (points at which a robot movement is not mathematically predictable), there are certain types of movements you cannot do. OLP helps avoid these errors when programming a robot.
In robotic machining, there could be hundreds to thousands of points,” Nubiola continued. “Nobody would ever be able to program that point by point using a teach pendant. You definitely need software to be able to do that offline.”
Simulation can be used for a proof of concept such as a robot integrator’s sales tool to demonstrate how a robotic system will perform. With simulation, users can detect possible collisions between the robot, tooling, fixtures and any safety fences. Simulation can analyze joint limits, singularities, and reach issues. Plus, it can reveal a host of eye-opening issues that save time and money in the long term.
OLP uses simulation to output robot-specific code that can be loaded onto the physical robot controller and then run the program. Post processors turn programming code into a language the robot can understand. Robot manufacturers have their own proprietary programming languages, which means third-party software solutions must be multilingual.
When OLP makes sense for a manufacturer
The main impetus for OLP is robot downtime, which is the time required to manually program a robot point by point with a teach pendant. There also are costs associated with the machine’s downtime and the programmer’s labor.
“If an end user is manually programming a robot on the teach pendant (online), they have to shut down production in order to program the part,” said Rob House, director of sales at OCTOPUZ Inc. “The benefit of using offline programming is you can be running production and you can program your next two, three, or five parts offline in the software and then once you’re ready to start a new job, you can just switch over the program and then start your production again.”
OLP is best-suited for complex path planning applications that require a lot of points such as welding, trimming, laser cutting, deburring, thermal spraying, painting, laser cladding, and additive manufacturing. OLP isn’t as beneficial for simple pick-and-place applications, assembly, packaging, and palletizing. These applications still can be programmed using offline software solutions, but users may not realize their return on investment (ROI). It’s more cost-effective to program manually if the process has only four or five points.
“If you’re spending as much time programming in software as you do with a teach pendant every single time you have a new part, you’re not any better off,” said Garen Cakmak, senior director at Hypertherm Robotic Software Inc. “For robots to be utilized in a high-mix, low-volume environment, software needs to be easy.”
Improving ease of use is top priority for these software developers. But simulation and OLP are pointless if they don’t accurately reflect reality.
Calibrate and don’t deviate
For OLP to work, the virtual world must match the real world. This means the simulation must represent the physical robotic cell accurately.
“The virtual environment in OLP software has to be an exact replication of the actual workcell on the shop floor, which is not the case in most situations,” said Helmut Ziewers, vice president of digital factory solutions for CENIT North America, Inc. “The deviations between a computer-aided design (CAD) model and the physical part associated with that CAD model can be minor or significant, especially in conjunction with less than perfect tooling. We still see major issues and people saying we can’t do this offline, because of those deviations.”
However, those deviations are not insurmountable. Calibration is critical.
“If we are off just a few millimeters or centimeters, you can create as many offline programs as you wish,” Ziewers said. “They will never fit. We have to know exactly how that robot was set up on the shop floor, and there must not be any deviations or else the OLP won’t work. The toolpath, the trajectory will always be off. This was the case with Crown.”
Crown Equipment Corporation manufactures powered forklift trucks for companies all over the world. Their Roding, Germany, facility has several complex robotic welding systems with external axes and multi-axis workpiece positioners. Faced with production bottlenecks caused by time-consuming manual robot programming, Crown Roding decided to explore if OLP was feasible. Their journey was not without a few hiccups. Some on the Crown team were skeptical while others were eager to try OLP.
CENIT was one of two suppliers brought into participate in a benchmarking study. Ziewers said they took CAD drawings provided by Crown’s automation integrator and created the virtual robotic workcell in their software. Based on those drawings, they created the robot program and ran it on the physical workcell. But something was off.
Figure 2. A robotic arc welding cell is programmed offline to reduce production bottlenecks caused by time-consuming manual robot programming. Courtesy: CENIT North America, Inc./Robotic Industries Association (RIA)
“The customer said this is exactly what we thought, offline programming is impossible. The last OLP software provider was not able to get it going and it looks like your software can’t do this either,” Ziewers said. “What they didn’t know is that the drawing from the integration company no longer matched how the integrator set up that cell.”
CENIT engineers arrived on site to physically calibrate Crown’s workcell.
“We found out what the differences were, dimensionally,” Ziewers said. “We applied those differences in our software and then adjusted the offline program based on the new setup in the virtual world. This matched exactly the physical setup from the shop floor, and the robot program worked perfectly.”
CENIT’s software coordinates the arc welding robot on an external axis with the movements of a multi-axis workpiece positioner. What used to take several days to manually program a part now takes a few hours. Production interruptions and downtime were reduced significantly and the welding quality consistently met the customers’ high standards.
Ziewers credited the calibration process for contributing to the application’s success. It’s important to consider the CAD model in relation to the physical part being processed.
“The part as it comes out of a press or die never has the same shape it’s supposed to have based on the 3-D CAD model. There’s always a difference between manufacturing and the CAD model. In sheet metal applications, for example, there’s springback.”
CENIT has calibration tools in their software such as three-point transformation, multipoint best-fit calibration, and in-process probing capability that help address issues related to springback and other variables that cause the physical part to differ from the CAD model. For example, the multipoint best-fit calibration tool picks 10 to 15 random points on a part and then feeds them back into the OLP system, which calibrates the part into the robot’s work envelope in the virtual cell.
“The ultimate goal of a 3-D simulation platform is to provide a software environment for the manufacturing automation controls engineer to validate their programmable logic controller (PLC), ladder logic, the human-machine interface (HMI) and OLP, and debug them in the virtual world before the actual workcell is built,” Ziewers said.
Easy to program; no expertise required
Another myth is that OLP is too complicated, difficult to use, and requires specialized expertise. Suppliers of simulation and OLP software are working hard to prove those assumptions wrong.
“People think that because a robot is a complex device, that offline programming is a complex tool as well,” Cakmak said. “That’s one of the biggest misconceptions. For people to adopt more robotics, we have to provide tools that make that end-user experience very easy and flexible at the same time.”
Hypertherm’s goal is to make robot programming as easy as possible for the user with their approach to simulation and OLP software.
Figure 3. A collaborative robot is programmed offline, saving operators months of manual programming time for this tedious railway maintenance process requiring hundreds of repetitive movements. Courtesy: Hypertherm Robotic Software Inc./Robotic Industries Association (RIA)
“A welder is a process expert. He’s not necessarily a CAD/CAM programming expert, or a robotician,” Cakmak said. “Task-based programming takes away all that complexity of robot programming, the CAD/CAM and robotics expertise, and really empowers the process expert.”
Cakmak said the software reduces OLP time and effort by optimizing robot trajectories and by automatically resolving robotic errors and collisions. It works with a simple but powerful, intuitive user interface and includes tools to optimize part placement, tool tilting, and effective control of external axes. It allows the end user and integrator to maximize the robot’s capabilities.
In a case study, Hypertherm’s software took on a tedious process by programming hundreds of points for an application involving high-speed trains in northern Germany. If programmed manually, the process would have taken months, which is problematic because keeping up with rail maintenance is an ongoing issue for all worldwide rail and transport services. Added to that is the growing traffic problem due to the rising frequency of trains in the region. Rolling contact fatigue where the trains’ wheels meet the rails is a major issue.
To help remedy the signs of rail wear, German companies NSI CAD/CAM Technik and Mevert Maschinenbau collaboratively developed a rail milling technology designed to smooth and re-profile the rail surfaces. The reversible rail plates have to be turned almost daily or be replaced. Four maintenance workers used to do that manually, but now that process has been automated with a robot.
The robot unfastens and fastens dozens of threaded bolts on each rail plate. Programming the 720 start-up positions was done with OLP software.
“You have hundreds of holes that you have to go into with this robot,” Cakmak said. “Using our software (and a CAD model of the rail plate), you can very quickly program the process. It automatically detects the holes, creates the screwing cycle, takes the bolts and threads them into the correct locations, while always checking for robotic errors and collisions. It validates the process and then outputs an error-free robot program.”
Using OLP software allowed personnel to focus on other maintenance work, while the robot makes sure it doesn’t miss any bolts. Instead of the original four, only two maintenance workers are required for the robotic process, saving Mevert significant costs.
Figure 4. A robot unfastens and fastens hundreds of bolts in railway plates, a process made more efficient with OLP software. Courtesy: Hypertherm Robotic Software Inc./Robotic Industries Association (RIA)
On the job is a collaborative robot. These power and force limiting robots allow operators to work in close proximity without the need for elaborate safety fencing. See below about how collaborative robots benefit from OLP.
High-mix low-volume no longer a constraint
Making OLP easier for users helps facilitate wider adoption of robotic processes. Robots are no longer the exclusive domain of megacorporations with deep pockets and high-volume production. Small and medium-sized enterprises (SMEs) can get in on the action. High-mix, low-volume production is no longer a constraint with simulation and OLP tools at the ready.
“We have some customers that have robot engineers who are running the robot and are very capable of programming it manually. They are also very capable with the software,” House said. “We have other companies that are just starting to automate. They don’t have a robot engineer on their staff. They can be hard to find and they can be expensive. They may be promoting a manual welder to now work with the robot and he may not have much experience.
“We take the approach that anyone should be able to program a robot,” House continued. “We have worked with people that have never touched a robot in their life and do not use software that often. It requires a little bit more training, but we can definitely get them to the point where they are comfortable programming a robot.”
Accumetal Manufacturing Inc. produces high-mix fabricated components for the off-road equipment industry, which includes subassemblies for mining equipment. The Canadian company has less than 100 employees and specializes in welding proficiency.
After many years of manually programming their robot, the volume of work increased to the level where Accumetal needed to take some pressure off their welders. They acquired a robotic welding cell and sought out a simulation and OLP software solution.
The cell consists of a six-axis welding robot and a single-axis headstock/tailstock workpiece positioner. House said this type of cell is very popular in the welding industry.
Figure 5. An arc welding cell uses OLP to help meet production demand by cutting programming time in half and reducing robot downtime. Courtesy: OCTOPUZ Inc./Robotic Industries Association (RIA)
“Our software handles all seven axes easily,” House said. “We can either index the part in position and weld it, or we can support coordinated motion where the robot and positioner are moving at the same time.”
OCTOPUZ added support for Accumetal’s robots, which were developed by Panasonic. The software easily converts the programming code into Panasonic robot-specific code.
House said that was one of the benefits of their software. “You could be programming three different robots to do the same process and the software would convert it to each robot code.”
House said it also allows the user to grow with automation and become flexible. A company may be using the software for welding right now. If for example they want to do a trimming application in the future, they can use their same suite of software to program that application.
“It also allows them to work on more jobs,” House said. “Many companies will have a welding robot, but they are only comfortable running one or two parts on that robot. Everything else might be manually welded because they’re not sure the robot can handle it. But with offline programming, you can do things like R&D and testing to make sure you can weld a new part. You’re doing all this from the comfort of your office and your computer. This allows you to start opening new jobs for your robot.”
In addition to helping Accumetal reduce programming time, OCTOPUZ was able to help their customer reduce robot downtime and meet production demand. The customer reported their initial programming time was cut in half.
Figure 6. OLP software simulates all seven axes of this arc welding cell and optimizes the process with real-time seam tracking and one-click weld recipes. Courtesy: OCTOPUZ Inc./Robotic Industries Association (RIA)
Accumetal’s cell also used through-arc seam tracking (TAST) for real-time tracking of the weld joint. This is used in case the robot needs to adjust its trajectory on the fly like when the CAD model and the actual part don’t match.
“We do a lot in the welding industry,” House said. “Welding is notorious for having parts that are not exact, especially for the CAD. There can be quite a bit of discrepancy. Lasers are very popular. You create a path in our software and then using a laser seam tracker, if the part is not perfect, it will modify our path to follow that seam. Same thing with touch sensing. The robot will touch the welding wire to the part at different positions to locate that part. Then, based on any discrepancies between our program and the physical part, it will modify that welding path to match the part.”
House said they have in-depth support for welding, with over 200 different variables for welding applications alone. Users can modify variables like torch tilt, torch twist, push/pull angle, and touch sensing with their program.
Also common in the latest OLP and simulation tools is drag-and-drop functionality. Everything is very visual and designed to be intuitive to make it easier to use.
“Rather than plugging in numbers to position components, you can simply drag from the catalog and snap it onto something like a robot pedestal,” House said. “You can drag which tool you want to use and it will automatically snap to the end of the robot. You can even hover over a part and it will highlight an entire toolpath around that part for something like trimming. Then you can click once and it develops that entire path.”
One OLP tool, multiple robot brands
A key advantage of third-party simulation and OLP software is its universality. Most can manage multiple robot brands. This is in sharp contrast to the OLP solutions offered by robot manufacturers. Robot OEM software is proprietary and specific only to one brand. OCTOPUZ has an online catalog of components included in their software. House says there are thousands of components in their library, including all the major robot brands and their various models.
“Many integrators we work with have multiple robot brands,” House said. “Rather than learning a number of different tools to program each one, they find it beneficial to have one tool to program all of them.”
Collaborative robots and OLP
Collaborative robots can often be “taught” an intended path via a lead-through teach feature. A user pushes the arm to the desired position and records the points. However, complex edge-following tasks often call for a more sophisticated programming solution than the standard platform that comes with some of these collaborative robots.
House said they have a number of customers using collaborative robots. Glue dispensing is a popular application for collaborative robots, which are designed to work in tight spaces near their human coworkers without safety cages.
“It all comes down to the application, not the robot,” House said. “For a pick-and-place application with only four or five points, you can drag the robot arm manually from point to point. But if you have a complex edge-following dispensing program you want to do with a robot, you can program it manually, but it’s going to be very time-consuming. That’s where software will help.”
Figure 7. NASA researchers use collaborative robots and OLP software to automate an otherwise tedious thermal inspection process for composite aircraft. Courtesy: RoboDK Inc./Robotic Industries Association (RIA)
Simulation and OLP software helped researchers at NASA Langley Research Center automate a novel aircraft inspection method using collaborative robots. The collaborative robot is equipped with infrared inspection cameras, which can detect material or structural defects in composite aircraft fuselages without damaging them by analyzing the flow of heat through the structure.
The inspection system is bulky and heavy, and must be moved across the entire exterior and interior surfaces of the fuselage. While the collaborative robots handle the heavy lifting, RoboDK software simulates and programs the inspection pattern to help ensure the robots don’t miss any areas.
But why use OLP when collaborative robots are supposed to be easy to program? Nubiola said that while the user interface is designed to be simple and is made for people who have never programmed a robot, for more sophisticated operations, collaborative robots may need to learn something beyond their initial programming, which is intended for more basic operations.
Consider a complex process like NASA’s inspection system, where the thermal sensors are analyzing hundreds of points. In this case, OLP software is a practical solution. Manual operation or even point-to-point robotic programming would be tedious. It would take three or four people to do the same thing one robot and one operator can achieve.
Using calibrated 3-D CAD models of the robot, fuselage, and inspection tool, NASA uses RoboDK software to not only generate the robot paths, but also test for the most efficient path. Automating the process with a robot also creates digital data of the inspection that can remain with the vehicle record. Subsequent test results on the exact same areas can reveal structural and material changes over time.
Figure 8. OLP software simulates robot path planning for a novel inspection system under development at NASA. The system uses collaborative robots equipped with infrared cameras to test for defects in composite aircraft structures. Courtesy: RoboDK Inc./Robotic Industries Association (RIA)
An inspection system using one collaborative robot on a wheeled base has been in testing at NASA for about a year. Nubiola said a new version of the system, which uses two collaborative robots working together on the same linear rail, has been in testing for a few months. This speeds inspection time and saves an operator from having to repeatedly wheel the robot from one section of the fuselage to another. The movements of the two collaborative robot are coordinated in the simulation and subsequent program. The setup brings this system one step closer to a production scenario.
Less time programming, more time producing
In a perfect world, a virtual solution should replicate the physical world. However, part and process variances, or deviations, can throw a monkey wrench into a company’s simulation and OLP setup. That’s where calibration is critical. Complex robotic processes and sophisticated software are no longer the exclusive domain of megacorporations. High-mix, low-volume job shops can join the OLP party. So can collaborative robots.
Tanya M. Anandan is contributing editor for the Robotic Industries Association (RIA) and Robotics Online. RIA is a not-for-profit trade association dedicated to improving the regional, national, and global competitiveness of the North American manufacturing and service sectors through robotics and related automation. This article originally appeared on the RIA website. The RIA is a part of the Association for Advancing Automation (A3), a CFE Media content partner. Edited by Chris Vavra, production editor, Control Engineering, CFE Media, firstname.lastname@example.org.
Keywords: robots, robot programming, collaborative robots
Robot offline programming (OLP) is believed to be too complicated and not cost-effective for companies.
Companies have been working to make their programs easy-to-use and flexible for many different brands and applications.
Collaborative robots can be programmed to learn more sophisticated and challenging applications that go beyond their initial design and purpose.