Fundamental research was carried out to demonstrate the feasibility of employing neural networks to analyse data, providing high performance monitoring and adaptive control of the metal active gas (MAG) arc welding processes. Data of three generic types are received (video, acoustic and electrical). Each type of data is acquired from multiple sensors. The designed neural network system implemented in digital signal processing (DSP) hardware produces an accurate, fast classification of the state of the welding process. The system accepts data from the data acquisition module in high speed serial form utilzing a local area network system (LAN). It outputs a classification of the welding process via a T800 transputer serial link to the process controller which generates a control strategy from a fuzzy model. The final networks comprise around 700 weights, including bias inputs for the nodes. All of these networks are executed once in every system cycle. Using a minimal run-time support host program and the full complement of DPRAM input and output operations, the DSP device could execute the 700 weight neural code in around 11.5 ms. The update rate of the entire system reached a maximum of 3.14 Hz. During evaluation, the update rate was deliberately reduced to the 1-2 Hz scale, to eliminate lag induced hunting. Closed loop welding evaluation was performed with a control strategy that avoids defects, controls bead dimensions and controls transfer mode, stability and spatter. The controller was very good at reaching and maintaining a required bead geometry. The system performed well with root gaps of up to 2 mm, after which weld quality rapidly deteriorated. In post-weld assessments, bead geometry control proved to be the system's best feature. The rule based controller actively drives the process into operating areas where the network outputs are stable and accurate.