Vision sensing, modeling, and control of laser weld pool geometry

A study was undertaken to develop a vision-based laser welding control system for on-line monitoring of the weld pool image, for studying the dynamic response of the weld pool, and for controlling the bead width. A number of sensing techniques have been developed for detecting various welding features. Among them, the vision-based sensing system which mimics the behavior of the experienced welding technician by directly inspecting the topside weld pool shape, has been considered as a prominent technique that reveals the complex weld pool dynamic response. The first goal of this research was to develop an on-line vision sensing system which processes the weld pool image in real time. A novel fiber optic cable based vision sensing system was set up to capture and digitize the image. An adaptive edge detection algorithm was developed to search the weld pool boundary in 32 directions and use the information from the previous image frame to adjust the searching position (Chapter 2). Experiments were performed to calibrate the image resolution and threshold value for the image processing program. The actual weld pool dimensions, i.e. width, length, and area, extracted by the edge detection program, were used to investigate modeling and control of weld pool dynamics.;The second goal was to model the laser welding process by considering both the weld pool and laser machine, using a two-dimensional heat flow analysis and system identification methods. The weld pool was modeled as an elliptical or tear-drop shape and the thermal properties were assumed lumped and temperature-independent (Chapter 3). The experimental results showed that the weld pool dynamic response is very fast (of the order of 10;The third goal of this research was to investigate control of pool width by proportional-integral-differential (PID) and adaptive self-tuning control algorithms (Chapter 5). The results showed that both control methods are satisfactory in regulating the weld pool width. However, the performance of the PID controller was influenced by the welding conditions. Adaptive control showed no advantage over PID control because of the limited experimental conditions.