In this paper the authors study the applicability of Artificial Neural Networks for the modelling of a widely used particular welding process in automotive industry: pulsed gas metal welding process (GMAW-P). Applying this artificial intelligence technology requires the introduction of input and output data to the network. To achieve this, an experiment was designed and performed. The main functions of the proposed model are: to simulate the process for purposes of training operators; to improve welding process performance by identifying regions that are insensible to variations on input parameters; and finally to increase the flexibility of a robotic welding cell. The concrete benefits obtained from the GMAW-P process model development are: optimization of critical variables of the welding process, support for development of virtual process and prototypes, definition of a robust welding procedure, quick response to product change and support in welding training.