In the general construction industry, many different parts are cut from sheet metal and then welded together on a large machine or vehicle. The trouble arises when the sheets are placed on the cutting table; the larger and heavier they are, the more difficult they are to position on the table. Without proper alignment, the sheet is likely to be cut outside acceptable tolerances, which results in waste and has a negative impact on the welding process. When sheets vary along the X, Y, and Z axes, the issue of exact positioning becomes even more problematic.
Sensor Control (Västerås, Sweden) was contacted by Severt Maschinenbau, a German robotics company, to help develop a vision system that automatically cuts sheet metal accurately. According to Sensor Control, their PathAdjuster works together with a robot to measure the actual part and adjust the preprogrammed robot path for the cutting process based on the position and size of the sheet.
The PathAdjuster uses 12 Sony ST50CE cameras connected to two Matrox Meteor-II/Multi-Channel frame grabbers. The system is powered by a 2.1-MHz Celeron processor housed in a 19" industrial enclosure. Fourteen double fluorescent tubes, plus a single laser line generator, illuminate the work area. Image acquisition and processing are performed by the Matrox Imaging Library (MIL). When cutting the sheet metal, an ABB robot follows a track developed by Severt Robotertechnik, Severt Maschinenbau’s sister company.
Ten cameras are installed in the ceiling over the cutting table; two additional cameras are attached to the robot’s end-effector. Using a composite image generated from all ten camera feeds, MIL first determines the position of the sheet with its Geometric Model Finder (GMF) module and then measures it. Sensor Control can measure critical dimensions such as distances between holes with an accuracy of 2 mm for 1200 x 9500-mm sheets. If the need arises, even greater accuracy can be achieved with the close-range robot-based camera for smaller sheets, or sheets with particular details or other critical dimensions. The new path is essentially the offset from the expected position of the sheet. With the new path fed to the robot, the camera and the laser line generator can then direct the robot to cut the sheet according to its actual position.
By using the PathAdjuster, the robot can cut the material with a high level of precision, within 0.1 mm of the required path. The sheets typically vary in size, ranging from 200 x 300-mm to 1200 x 9500-mm, and can be 30 mm to 300 mm thick. Furthermore, the robots frequently cut sheets of different shapes, including square, round, elliptical, and nonsymmetrical shapes. Since the sheets vary in size, shape, and position, the application is ideally suited to MIL’s geometric module.
According to Severt Maschinenbau, the company can produce edge-cut sheet metal with a very high degree of precision, which reduces costs for the subsequent welding processes. The customer can perform some jobs twice as fast as before using the PathAdjuster.
Challenges in Development
Generating a composite image of the entire sheet from the ten individual cameras was one of the more difficult aspects of the application. In order to align all the images without losing image data and while maintaining accuracy, calibrating the system became a very crucial task in the PathAdjuster’s development.
Controlling the robot also presented challenges for Sensor Control, but since the company has a great deal of experience with ABB and their robots, they could prevent certain problems before they occurred. For example, the sheets are not in pristine condition when they are placed on the worktable.
The robot program is rather complicated. Sensor Control made good use of being a former ABB company, as the recalculation of the robot frames had to be made in the correct order. When setting up a plate for the first time, it has to be sandblasted to receive as little unwanted data as possible. The MIL GMF function then takes care of virtually all disturbances in the production plates. The plates are fed into the system dirty, full of rust, and some of them have a lot of markings. The GMF functions locate the sheet without any difficulties. It is no longer necessary to clean up (sandblast) the sheets since the GMF function will only search for the information from the clean (taught) plate.
Users of the PathAdjuster benefit from improved accuracy, less waste, and faster job completion. Implementing geometric pattern matching means the system is flexible in placing the sheets on the table. Regardless of the position, the system will find the sheet and let the robot run its pre-programmed path.
This article was contributed by Matrox Imaging, Dorval, QE, Canada. For more information, visit http://info.hotims.com/3445-321.