Viewpoint

Citation

Madsen, O. (2007), "Viewpoint", Industrial Robot, Vol. 34 No. 6. https://doi.org/10.1108/ir.2007.04934faa.002

Viewpoint

For many years, welding has been one of the main applications for industrial robots. Thousands of new welding robot cells are installed every year in a large range of different industrial plants, and a whole industry has developed around designing, installing and maintaining these robot cells.

With the long history of robotic welding and the high- activity level in the area, it can be tempting to assume that further significant growth in the area is not possible.

However, most of the existing welding robot systems are focused on the manufacturing of large and medium lot size products, whereas the application of robot technology for small lot size manufacturing is sparse. Here, the welding operations require a large degree of human involvement. Hence, a large potential for growth lies in introducing robots into small lot size manufacturing environments.

Already, a lot of work has been made in the area: sensors, such as through-the-arc sensors, touch sensors and laser scanners have for many years been used to control welding robots. They can be used to reduce the requirements for accurate part set-up and fixtures. Furthermore, off-line programming systems are increasingly used to improve efficiency of programming tasks. The off-line programming systems use models of the robot, the environment and the process to generate the necessary robot programs. Presently, most off-line programming applications involve a certain degree of human involvement. Typically, an operator guides a computer model of the robot through its trajectory, and performs simulations to check the generated program. Combining the off-line programming models with automatic collision avoidance and scheduling schemes makes it possible to create robot programs completely automatically.

In spite of the existence of sensors and automatic and semi- automatic off-line programming systems, the application of robots for small batch production is still rather uncommon. Existing solutions are not sufficiently flexible to handle a high number of variants in small lot sizes because programming and set-up/lead time exceed by far the execution time.

To enable application of robots for small batch welding, a range of technological and scientific breakthroughs must be achieved in a number of areas. The technology pillars of the breakthrough can be found in established robotics and automation technology as well as in emerging technologies as; industrial information and communication technologies (ICT), virtual reality, A.I. cognition, and sensing.

In particular breakthrough in two areas could enable efficient use of robots for small batch production. Firstly, I think that a large improvement in the welding robot's ability to perform small batch production can be achieved by using automatic and semi-automatic learning methods to continuously acquire new knowledge about the system. This knowledge can be used for ongoing increase of system efficiency and its ability to adapt to/re-organize according to new tasks. Secondly, I believe that we in future systems to a higher degree must integrate human operators into the system. It must be possible for the operator to carry out new and complex tasks which the system cannot handle. The operator can either solve the problem by himself, or he can show the system how to solve the problem (programming by demonstration).

More interestingly the operator interfaces can be combined with the automatic learning methods so that the welding system can learn how to solve the task the next time a similar task occurs. Hence, potentially a learning curve as observed in human operators can be realized by coupling operator reasoning with machine learning. This will enable the design of a highly flexible manufacturing system which will increase autonomy, efficiency and robustness over time.

The design of a general structure of such a manufacturing system has been the focus of a joint European co-operation (LOTSIZE1). Here, a number of European partners have outlined the concept of a new manufacturing concept incorporating robotics, advanced human machine interfaces and learning methodologies. More details about the LOTSIZE1 vision can be found in www.lotsize1.org.

Ole MadsenDepartment of Production, Aalborg University, DenmarkE-mail: i9om@production.aau.dk; web site: www.production.aau.dk