Classification of Sensor Based Control Systems for Robotic Arc Welding

In this paper a framework for classifying sensor based control systems for robotic arc-welding is presented. It is proposed to classify the sensor based control systems based on the deviations that can occur between models used for programming the robot and the physical world. With such a classification of sensor based control systems, selection of an appropriate sensor based control system for a given manufacturing task can be made by determining a sensor based control system which can compensate for relevant model deviations. In this paper the overall classification principle using the model deviations will be described, together some examples of sensing systems and control strategies related to each class of model deviation. FOR ROBOTIC ARC-WELDING Many different kinds of sensor based control systems using a wide range of different sensor technologies and control strategies have been developed. However, none of these satisfy all purposes. At present the selection of a sensor based control system for a given welding task is to a very large extent based on the intuition and experience of the engineer making the selection. However, as the range of application of sensor based control systems is extended to more complex tasks, and as the systems are growing more complicated, it becomes more and more difficult to select an optimal or even an appropriate sensor based control system. For this purpose a classification of sensor based control systems can be useful. In literature, a number of comprehensive surveys and classifications of the sensing technology used for sensor based control of robotic welding can be found (e.g. in [NOMURA, 1991]). Also classifications of different control strategies can be found (e.g. [NOMURA, 1991], [LAURIDSEN, 1992]). However, a complete classification of sensor based control system with respect to the welding task they can be used for is presently not available. In this paper a framework is proposed inside which such a classification can be made.