A hierarchically structured environment that integrates a knowledge- based expert system, adaptive process control and pattern recognition techniques for controlling a laser cutting process is described. Knowledge of the laser cutting process for different materials is organised and encoded into a rule-based system. An adaptive control algorithm based on on-line recursive parameter estimation and on-line control law synthesis was adopted for the highly non-linear cutting process control. Cutting speed was selected as the major control variable. Irradiance emitted from the cut front is used for the feedback signal to this adaptive controller. The irradiance signal feeds the recursive parameter estimator for system identification. Techniques of pattern recognition, which have been well developed in coherent optics, were applied to assess cut quality by characterising the exit spark cone images of the gas assisted laser cutting process. Images from the cutting processes were grabbed, edge enhanced and correlated with a synthetic discriminant function filter which was synthesised from reference images to give good cut quality. Results from digital simulations based on these pattern recognition algorithms are also presented.
Read full abstract