10-04-2017, 08:13 PM
[attachment=5471]
Abstract
This paper presents a study on an application of genetic algorithm (GA) for edge detection of molten pool in fixed pipe welding. As circumferential buttwelded pipes are frequently used in power stations, offshore structures, and process industries, it is important to investigate the characteristic of the welding process. In pipe welding using constant arc current and welding speed, the bead width becomes wider as the circumferential welding of small-diameter pipes progresses. In order to avoid the errors and to maintain the uniform weld bead over the entire circumference of the pipe, the welding conditions should be controlled as the welding proceeds. This research studies the intelligent welding process of aluminum alloy pipe 6063ST5 in fixed position using the AC welding machine. Themonitoring system used an omnidirectional camera to monitor backside image of molten pool. A method of optimization for image processing algorithm using GA was proposed and has been implemented into a process to recognize the edge ofmolten pool. The result of detection, which is back bead width, was delivered into a fuzzy inference system to control welding speed. The experimental results show the effectiveness of the control system that is confirmed by a sound weld of the experimental results. Introduction
As the many applications of pipe structures for ocean exploration, water and oil pipe lines, etc., are increasing, the requirements for greater productivity, higher weld quality, reduced cost, and accuracy in manufacturing these structures are also increasing. There are many problems in automating arc welding processes such as sensing, monitoring, and line tracking. The difficulty in the welding process related to nonlinear and multivariable-coupled welding involves many uncertainties, such as influences of metallurgy, heat transfer, chemical reaction, arc physics, and magnetization . In fixed-pipe welding, compared to plate welding, if the constant welding conditions are maintained over the full joint length, the bead width becomes wider as the circumferential welding of small-diameter pipes progresses. Therefore, in order to avoid defects during aluminum pipe welding and to obtain uniform weld bead over the entire circumference of the pipe, the welding conditions should be controlled as the welding proceeds. Since the early 1960s, sensing and control systems have been successfully implemented in applications where the sensor could be placed on the backside of the weld and moved in synchronism with the welding torch . The development of intelligent control systems has been conducted for modeling and controlling the welding process as they used neural network and fuzzy techniques . Another difficulty in controlling an arc welding process isdetecting molten pool geometrical features, such as weld bead width and penetration, either from the topside or backside of the molten pool, conveniently and in real time. Various efforts have been made to sense molten pool sizes in real time from the topside, such as ultrasonic detection, infrared sensing, pool image processing, and radiographic sensing, to produce weld quality control . The experiment using the vision sensor to control the TIG weld width for fixed stainless steel and aluminum alloy pipe was conducted using a plain mirror. The next development of the monitoring process was conducted using unidirectional vision-based molten pool monitoring . The problem that occurred in the experiment using an aluminum pipe was the low brightness of the molten pool. Therefore, in previous experiments, the method to detect the edge of the molten pool using brightness range was proposed . However, the method depends on the use of the experience in monitoring. The automatic optimization for constructing the brightness range for searching the edge of the molten pool is needed.