This dissertation documents the investigations into on-line monitoring the CO2 laser welding process and optimising the process parameters for achieving high quality welds. The requirements for realisation of an on-line control system are, first of all, a clear understanding of the dynamic phenomena of the laser welding process including the behaviour of the keyhole and plume, and the correlation between the adjustable process parameters: laser power, welding speed, focal point position, gas parameters etc. and the characteristics describing the quality of the weld: seam depth and width, porosity etc. Secondly, a reliable monitoring system for sensing the laser-induced plasma and plume emission and detecting weld defects and process parameter deviations from the optimum conditions. Finally, an efficient control system with a fast signal processor and a precise feed-back controller.With the purpose of optimising welding process parameters automatically or semi-automatically, the fundamental principle of the laser welding process and the correlation between process parameters: i.e. the gas parameters and the focal point position, weld quality characteristics: i.e. the seam width and depth, and monitoring signals are systematically studied and investigated. For gas parameter optimisation, there are 5 gas variables optimised when applying the Design of Experiment (DOE). One photo diode is set to monitor the welding process in order to characterise the consistence of the discrimination features of the welding process by applying FFT analysis of the signals. DOE is proven to be a useful tool for manual parameter optimisations in laser welding. The frequency of the photo diode signal could be related to weld quality measures, however, the application possibility of the signal in a control system has not yet been discernible.For focal point position optimisation, a focus control system is designed and built up with the controller, sensor, and signal processor. The optimisation procedure starts with the welding process being carried out by continuously moving the focus from above a welding plate to below the plate, thus the process is ensured to be shifted from initially surface welding to deep/full penetration welding and back to surface welding again. A clear change on plasma brightness from the process is monitored by the photo diode. Thereafter the optimum focal point position (OFPP) is generated by the self-learning mechanism - the neural network. The optimisation procedure is completed with the welding process being carried out by adjusting the focus of the laser beam to the OFPP. The self-learning focus control system, employing the off-line trained neural network to generate the OFPPs from on-line monitored photo diode signals, has proved to be capable of optimising the focal point position automatically with good accuracy in CW CO2 laser welding.