题目： Control of Welding Process Based On Modeling of Human Welder Response
摘要：A real-time vision system and image processing algorithm has been developed to monitor and measure the specular three-dimensional weld pool surface under arc radiation in 40 milliseconds. A set of parameters are extracted from the three-dimensional surface to characterize the weld pool surface. A model has been established to use these characteristic parameters to predict the width and reinforcement of the root-face weld bead that quantify the weld joint penetration. Experiments have been conducted where a novice welder adjusts the welding current to overcome the effect from the fluctuations in other welding parameters to maintain the weld penetration consistent at desired levels. The weld pool surface being observed and used to adjust the welding current is also real-time measured by the developed vision system and recorded together with the welding current. A model has been established to correlate the recorded current adjustments made by the human welder as function of the characteristic parameters the human welder observed and the vision system recorded as the human welder response model. This response model has been used to develop an intelligent welding process control system where the real-time vision system measures the weld pool surface and a control algorithm uses the response model to determine how to adjust the welding current based on the measured characteristic parameters. Because the model was from a novice welder who may have overreacted, a filter has been added to filter out the high-frequency overreaction dynamics. The model together with the filter provides an improved intelligent control system functioning as a more skillful human welder.????