Abstract
During arc welding process, the weld bead occurs through the formation of an electric arc between the filler wire and the specimen. A non-intrusive sensor for the monitoring of the process has been developed. The sensor consists of four electro-optic modules able to detect infrared, visible, and ultraviolet radiation coming from the interaction zone. During the welding process four different signals (IR, UV, visible and plasma electronic temperature) are simultaneously stored.
The sensor was initially designed as a remote, standalone device pointing on the work plane.
This solution is very useful for automatized workstation where the sample moves and the arc is fixed, as typically occurs with X-Y table. In this work the system is splitted in two modules, as shown in the following block scheme, and the optical components for the collection of the radiation coming from the working zone are assembled in a telescope and integrated on the torch used by the operator for manual arc welding processes, in order to allow the sensor application to a wide range of industries.
Process data stored by the system are filtered by applying a continuous Kalman filter, to reduce the S/N ratio. Numerical values are then analysed considering blocks of 1000 values, and for each data set information dealing with intensity, signal noise and behaviour are evaluated. Based on this first analysis, monitoring rules have been pointed out for each signal. These rules are then implemented by a Fuzzy supervisor, able to furnish information regarding the weld bead quality.