An Exact Science

Canadian research facility developing precise robotics for aerospace manufacturing

By Treena Hein | Published February 10, 2008

Michel Lambert, team member at the Aerospace Manufacturing Technology Centre (AMTC) in Montreal, is creating robotic systems for aerospace manufacturing.

Of all engineering and manufacturing sectors, aerospace is the most resistant to change—and anyone who’s ever flown on a plane or plans to book a flight in the future should be thankful.

This caution is mainly due to the uncompromising performance and exceptionally high degree of reliability required. All aircraft experience enormous amounts of stress compared to automobiles or other machines. Any weakness in a plane’s components (frame, shell, engine or wiring) is inherently much more dangerous than the same weakness occurring in a vehicle traveling along the ground.

The allowable tolerances for aerospace material performance and assembly must therefore be extremely strict. Rivet alignment, for example, must be in the 1,000ths of an inch. Residual stress and surface finishing of materials must be outstanding.

This need for strict tolerances and a high level of accuracy in the manufacturing process has meant most aerospace assembly is manual; robots have not yet been able to achieve the precision required. At the same time, however, the industry recognizes that using automation and intelligent processes are imperative in the quest to keep costs down and remain competitive.

Claude Perron, for one, believes automation is the only way the Canadian aerospace industry is going to remain productive. Perron is the leader of the Automation, Robotics and Intelligent Manufacturing Systems Group at the Aerospace Manufacturing Technology Centre (AMTC) in Montreal, an initiative supported by the National Research Council and Canada Economic Development (Quebec).

AMTC aims to develop “modern aerospace manufacturing technologies that have the potential for significant cost savings while maintaining high levels of quality, reliability and performance.” AMTC opened in 2000, but moved to its current permanent home and received all current equipment in 2004.

Perron notes that although integrating automation into aircraft manufacture wasn’t an option five or 10 years ago, “Now the technologies and tools required are there,” he says. “They’re flexible enough to combine virtual manufacturing with the real thing—and the time is urgent now in order to reduce costs and compete with emerging countries like China.”

Michel Lambert, one of Perron’s research team members, says that although research into aerospace manufacturing automation is proceeding all over the world, “A lot of people still don’t believe robots can do this type of work.” Tasks such as drilling of fuselage and wing parts for assembly, metallic panel forming and aircraft surface finishing tasks like polishing, grinding and removing metal burrs all require precise and repeatable motion.

This has so far proved difficult to achieve with robot arms, partly due to the inherent elasticity in the motor drives and gearboxes in the arm joints, he says. This elasticity causes unwanted vibrations and drift from the desired path.

For the past 12 months, Lambert has been hard at work quantifying how much various robotic parameters (robot mass, inertia, elasticity) affect the quality of manufacturing operations (such as grinding and polishing) in terms of the tolerances that can be achieved and how much vibration can be minimized, among other considerations.Most of the research has been conducted within the virtual world. The team began the project by considering different approaches and software modeling options.

AMTC used Delmia V5 Robotics to simulate and program its technology demonstrators, such as this robotic system for polishing landing gear assmblies.

Ultimately, they chose DynaFlexPro, an add-on tool for mathematical solving environment Maple, developed by Waterloo, Ont.-based Maplesoft. DynaFlexPro is a mechanical design tool that uses symbolic computation methods to derive the system equations of complex mechanical multibody systems. The team also used Delmia V5 Robotics to design and program the robotic cell.

After choosing his modeling tools, Lambert then spent about a month creating a classical robot arm model with six degrees of freedom. He was confident the model would reflect real-world results because the kinematic components of the model were validated using a real industrial robot while the dynamic components were validated by cross-comparing the same robot modeled using several approaches.

“I used DynaFlexPro because the user works at the topology level,” he says. “Once a given topology is defined, the model’s constitutive equations are derived automatically. If the model is right from a topological point of view, the resulting model will be right from a kinematic and dynamic point of view as well.”  

Next, Lambert linked the robot model with a suitable simulation model of surface finishing processes (grinding, polishing, metal cutting) he developed from models published in scientific journal articles. He says that it wasn’t an easy task to find a useful process simulation program.

“Grinding, for example, is hard to model from pure theoretical principles,” he says. “Many models that exist are based on experiments, so I had to find a model that was analytical enough to be useful in a simulation context.”

Then came assigning parameters such as elasticity and vibration to the virtual robot, and measuring the effect of these on the virtual manufacturing processes. Achieving a realistic representation of elasticity, Lambert says, was difficult.

“You have to fiddle around with the program at the parameter level and the component level,” he explains. “For example, you have to determine whether to place the elastic component before or after the gear box in four of the robotic arm joints.”

The next phase of the project involved building a real hardware system to test how best to optimize surface finishing. The choice of a particular industrial robot brand to use in the next phase was carefully addressed.

“One needs a robot that is commonly used in the industry—aerospace in particular,” Lambert explains. “At the same time, the robot has to have a good degree of openness in its system architecture.”

This 500-kg-payload industrial robot, equipped with a drilling and riveting effector for assembly of aircraft panels, is one of the prototype robots being developed at AMTC.

Open architecture allowed the researchers to interact with the robot at levels not commonly accessed by an end user, such as performing sensor-based control and real-time modification of the robot motion.

“KUKA robots were found to have the best balance between industrial ruggedness and openness of architecture,” he adds.

With the virtual and physical robotic systems in place, Lambert says the team is in a position to find an approach that allows a robotic system to perform up to the exacting specifications required in aerospace manufacturing.

“We used Delmia to perform the simulation, determine precise robot placement and program trajectories,” he says. “Then, when the physical cell does the execution, you can compare the real-world trajectory with what you wanted originally. You can then look at the deviations and use DynaFlexPro and Maple to analyze the causes of these deviations, which are related to elasticity.”

As of yet, Lambert says the team hasn’t achieved its goal—and reaching it will take the better part of the next two years.  

“Achieving the right conditions should be challenging,” he says. “We’re going to be pushing the system in terms of flexibility and tolerances. We will have to deal with the limitations of the commercial components.”

When asked whether his research will drive improvements in the future design of robots, Lambert answers that he believes it will. He says that some robotic companies, particularly KUKA, are very active in designing robots specifically for the aerospace industry and are interested in all research pertaining to that.  

Other projects currently underway at AMTC related to the development of low-cost reconfigurable robotized cells for aircraft component assembly include the design of a fully integrated vision system for drilling sequence and panel inspection, and design of robotized cell auxiliary hardware components, in co-operation with Bombardier Aerospace.

“Special calibration procedures were developed, and a real-time simulation system,” Perron says, “which optimizes the position of the camera.”

In another project, also with Bombardier, a hardware demonstrator to position panels for riveting operations has been developed over the last two years and is now ready to be used on the production line.

Lambert says that “usually you have someone on the line positioning the part by hand, but now two robots will be positioning the part [for a third robot that does the riveting].”

The impetus for this project was to prevent health problems with employees (since moving heavy parts can cause tendonitis and other issues) and to achieve better accuracy. This project’s main challenge was co-ordinating the movement of the positioning robots.

“The first one would deviate with what the second one was doing, but we fixed that through software changes,” Lambert says.

After 12 years with Bombardier Defence (working on F-18 bulkheads), Perron was attracted to work at the AMTC because of the opportunity to do research using equipment not available in the industry. He also believes the cross-displinary work performed at the centre is helping bring these next-generation techniques to market faster. Besides robotics, other AMTC research groups include metal removal and machining, metal product joining and forming and composite products.

“We are within walking distance of one another, but also work with groups in Ottawa,” Lambert says. “We define joint projects and combine our expertise. This collaboration makes our strength.”
iar-ira.nrc-cnrc.gc.ca
www.maplesoft.com