WELDING ROBOT USING A COMPUTATIONAL
MONO-OCULAR VISION SYSTEM

ABSTRACT
In automatic production lines using welding processes, robots are used to move the welding torch. But for the robot to weld it is normally that it has been previously programmed to follow a predetermined trajectory. If the location of the pieces to be welded changes for any reason, all the programming will have to be redone. One way to eliminate the robot programming is to use a vision system attached to the robot, which captures images from the area to be welded, interprets them in a computer, detecting the joint line position and then positions automatically the robot to follow the trajectory to be welded .The system developed here consists basically of a CCD video camera and a computer. Its algorithm analyses images in the simplest way possible, in other words, exclusively through a camera attached to the robot arm, without using any additional devices to enhance them say with the help of laser or filters. The joint line position is detected using a mix of comb scanning and circular searching. Thus we can establish the welding trajectory to be followed by the robot.

INTRODICTION

In industrial automation, the use of sensors is a key for the acquisition of parameters and information, which can be utilized in the control of the manufacturing process. In the case of automation with robots, sensors can measure internal parameters, like joint angles, but also external parameters like torch position. The attachment of sensors to a robot expands its capability, reducing its dependence on other machines or people.


In automated production lines by welding, robots are used to move the welding torch. However, for the robot to weld, it is normally necessary that it has been previously programmed to follow a predetermined trajectory. If the location of the pieces to be welded changes for any reason, all the programming will have to be redone. This happens because robots are usually unable to see where the pieces to be welded are, and are unable to detect the shifting of joint line trajectory.


By the installation of a vision system, the robot can handle this problem, and so becomes more versatile. By doing so, the robot programming is nor necessary. This article describes a mono-ocular vision system for automatic off line guidance of welding robots. The system is able to acquire and process images in the simplest way possible, just using one monochromatic CCD video camera, without use of any extra devices, such as lasers, special illumination and/or filters. It detects the absol8te position of the joint line, identifying the trajectory to be welded. With the trajectory identified, the system acts in the robots controller, moving it to the start point, initiating the welding arc, and controlling its displacement along the full length of the trajectory depositing the weld bead at the same time.

THE SYSTEM

The vision system can be described from the illustration shown in figure below successive images of the work area are acquired by a CCD monochromatic video camera and sent to the frame grabber (coordinating data of the trajectory are sent to the robots controller, using serial communication between pc and controller, which controls the displacement of the robot arm, moving is (and the welding torch attached to it ) along the joint trajectory. The system presented here was implemented in the computer) as a compact video signal. In the PC, with the help of the developed search algorithm, it is possible to detect the joint path configuration, thus identifying the trajectory to be welded.