Arc welding seam tracking system based on artificial neural networks

A new approach towards controlling an arc welding seam tracking system based on process-oriented data is presented; an artificial neural network is the central controlling element. Special features of neural networks, such as trainability, abstraction capability and fault-tolerance give this controlling system significant advantages. The adaptation of the system to new welding tasks has been simplified substantially in comparison with conventional approaches: analysing the new process parameters and programming the system are replaced by producing reference data from the new welding process and subsequent automatic learning of an artificial neural network. Results on the basis of a quality function are presented