Neurofuzzy model based non-linear modeling and control of weld fusion zone geometry

The state of fusion is one of the primary determinants of weld quality. Because the fusion state is specified by the top-side and back-side widths of the weld pool, a direct feedback measurement of the fusion state requires both top-side and back-side sensors. However, it is often difficult for a sensor to access the back-side of certain constructions, and a back-side sensor cannot be attached to and moved with the torch. A neurofuzzy system was developed to precisely estimate the back-side bead width using the measured top-side geometry of the weld pool. The feedback could, therefore, be used for the closed-loop control of the fusion state. A closed-loop control system was designed to regulate the welding process to achieve the desired fusion state.