In partnership with ABB Robotics , Spookfish has deployed vision-based inspection software in August 2014 at the Wheels India factory in Chennai, for determining the angular position of the weld bead on wheel rims. This article provides a high level overview of the manufacturing process for wheel rims, the importance of finding the weld bead, and how Spookfish have met this requirement with a computation time of less than 40ms.
Wheels used in automotives – anything from bicycles to fighter jets – go through a rigorous and complex manufacturing procedure consisting of several steps, to bring together what is probably the single most vital part that keeps us in motion. Although often overlooked by the end users, the manufacturing of wheels and their parts also use the latest in engineering to ensure maximum safety and efficiency – the industry is indeed constantly “reinventing the wheel!”
Wheels usually consist of two important parts – the disk and the rim – and the work presented in this post focuses on the wheel rim. Although there are several different ways of manufacturing the wheel, this nice little video by STARCO provides a quick overview of spun wheel manufacturing:
Two steps in the manufacturing process of the wheel rim are of interest to the work covered in this article – (1) a weld that turns the sheet metal into a cylinder, and (2) a drilling procedure to make a hole for the valve that keeps the tyre in good health.
One of the first steps in manufacturing the rim is to weld the sheet of metal into a single cylinder. In terms of lateral force, the welded portion is the weakest and most vulnerable part of the rim because it is a joint in the sheet. The snapshots below have been taken from the STARCO video.
In order to fit the valve, a hole needs to be drilled into the wall of the wheel rim. The following snapshot, again taken from the STARCO video, illustrates this step.
Like the weld, the hole is also essentially weakening the wheel rim because of the break in structural continuity, and hence the position of this hole must be chosen strategically. Because these elements form two weak portions on the rim, the obvious strategy would be to place them exactly opposite each other (as far apart as possible) to reduce vulnerability to damage.
Spookfish, in partnership with ABB Robotics, have developed and deployed a solution to automatically find the weld bead on a wheel rim, so that the hole for the valve is drilled automatically and accurately. The setup involves the use of an ABB Robot to handle the wheel rim, a circular light to provide uniform and consistent illumination and a Basler camera to capture images for processing. The camera is placed within a custom-built enclosure to ensure that external lighting has little or no influence on the captured images. For processing, the camera is connected via a GigE port to a customised computer with a dual core CPU. Communication between the ABB robot and the software application is carried out via TCP/IP communication through a serial port.
Once a wheel rim is ready for inspection, the robot picks it up with a unique gripper that holds the wheel rim from the inside. The robot then rotates and translates its arm based on pre-calibrated settings and places the wheel rim in front of the camera housing, with highly accurate positioning (tolerance within a few microns). The robot then communicates to the software application to indicate that the camera can now capture a frame for processing. At this point, the light is switched on and the inspection software is triggered. The software then captures a frame to process.
Spookfish’s algorithm for detecting the weld bead analyses intensity profiles of the wheel rim along its circumference, and uses a combination of non-linear aggregation combined with a wavelet-based low pass filter to generate a probability distribution of the location of the weld. The angle with the highest probability, calculated to an accuracy of 0.3 degrees, is then indicated as the position of the weld bead. This angle is communicated back to the robot, which then readies the wheel rim for drilling.
The total time required by the inspection software for computing the position of the weld bead and reporting the result is on average 40 milliseconds, thanks to proprietary technology that employs a fast multi-threaded approach. After running the system live for a period of 4 months, and testing with 15+ variants, the system has achieved an accuracy rate of 99.93%.