In the automotive industry, the industrial robot is a common solution to achieve high-speed, repeatable processes. But due to its lack of accuracy and its compliant nature, use of robots has not been as successful in aircraft manufacturing.
One commonly used robot with a 60-kg payload has a repetitive accuracy of ±0.05 mm, though the worst-case absolute accuracy within the whole workspace may generally be about ±3 mm without extra calibration. Also, since the robot is of a serial type, it is a compliant structure and the accuracy is only guaranteed if no external force is applied.
Even in the automotive industry, the industrial robot is seldom used for complex assembly processes. Its position-driven nature (the fact that it is programmed to achieve a certain pose in space) makes it unsuitable to handle complex fittings, paths, and geometry variations.
Automation within aerospace has been solved by using dedicated fixtures to achieve high accuracy paired with specially built machines. The results are very capable installations, but they are often also very expensive and take up large amounts of floor space.
Research efforts using industrial robots have been focused mainly on drilling and riveting, and the robot has been paired with metrology or vision systems to improve accuracy. In one approach, predrilled locator holes are used to align parts. In another approach, off-the-shelf metrology systems are used to both locate parts prior to pickup, and calculate misalignment and deformation.
Force control is a newer approach to handle part variation and part alignment. With force control, the robot can switch between being position-driven to being force-driven, working to achieve a desired force rather than position. This approach makes the robot suitable for many assembly processes in which part alignment and part-to-part relations are key issues.
Force control can therefore be used to place parts and minimize gaps in assemblies, reducing the need for shimming. The technology requires a force sensor to be mounted on the robot; a good location would be in the joints or on the tip/nose. It also requires additional hardware and software. Commercial systems already are available on the market, with already packaged functionality as in the ABB S5 with functions aimed at machining and assembly processes.
Researchers at Airbus, Linköping University, and Lund University wanted to see if the force-control capabilities offered in today's systems could be taken further. To that end, they conducted a trial using an ABB IRB4400 with the S4C+ controller, predecessor to the current S5 generation. It is a research system, based on the commercial one. The controller was opened up by ABB to allow changes of reference for the low-level joint control loops. To enable this, a Motorola G4 processor was added to communicate with the S4C+ controller through a common bus. Also added was an extra card to handle force data in x, y, and z axes.
A six-degree-of-freedom force sensor was mounted on the robot manipulator between the flange/nose and the attached gripper, allowing force data collection in x, y, and z and for corresponding torques. An external master computer used for programming was connected through TCP/IP to the controller cabinet. This also acted as the cell controller.
Controller code was generated in Matlab/Simulink.
The trial assembly involved mating a rib to a jig containing multiple brackets, with the robot using a vacuum gripper for placement. In operation, the rib would come in contact with difference surfaces on the jig and adjust movement accordingly until the final position was achieved.
The process took 140 s. With improvements to the system, the researchers believe the time can be reduced. Challenges for the trial, and for real implementation, include vibration. In the trial, vibrations from people walking on the floor was amplified through the robot structure and interpreted by the system as force readings. To avoid this, the force sensor readings had to be filtered. Overall, the sequence was run successfully multiple times, with the setup proving to be stable and robust.
Factors that affect process time include the type of materials being mated. In the case of metal to metal, robot motion would have to be relatively slow to avoid springback issues. Fast feedback also would be required in metal-to-metal conditions, because small motions can cause large forces. A solution could be to use a vision system or some other type of sensor to detect when contact is close.
The trial showed that a stable, repeatable assembly process consisting of several iterative steps can be realized. Force-controlled assembly using an industrial robot reduces the need for bespoke tooling, and the industrial robot system is relatively cheap in comparison to specialized systems.
The article is based on SAE technical paper 2011-01-2734 by Marie Jonsson of Linköping University; Andreas Stolt, Anders Robertsson, and Klas Nilsson of Lund University; and Thomas Murray of Airbus.