An automated image processing system uses eight cameras to test up to 70 characteristics.
Dipl.-Ing. (FH) Andreas Tarnoki, VMT Vision Machine Technic Bildverarbeitungssysteme GmbH
Safety and quality requirements of automotive dashboards are becoming increasingly complex. As a result, a large number of parts must be tested to ensure that they are present, are fitted correctly, are the correct parts, and are used and positioned accurately before delivery. These parts include rivets (particularly in the airbag area), clips, spring holders, screw fittings, and other add-ons and assemblies.
Johnson Controls of Lüneburg, Germany, manufactures a variety of dashboards and door trim panels for automotive manufacturers, supplying these parts directly to the production lines. In the past, most of the 60 to 70 changing characteristics of safety and quality for the dashboard have been tested manually by the staff of Johnson Controls, working in three shifts.
To save man-hours and to reduce the possibility of human error, the company sought to automate the testing. It needed an imaging system that ensured reliable and continuous testing and that documented flaws on the dashboards.
Obstacles on the line
One challenge is inspecting the color combinations and surfaces. Because of the variety of products and the way they flow on the assembly line, it was not possible to specify the color or surface in advance. Another challenge is transferring faulty dashboards to a reworking point after the inspection is performed. In addition, the flaws must be documented so the parts can be repaired or replaced as soon as possible.
An automated image processing system was designed by VMT Vision Machine Technic GmbH of Weinheim, along with Dieffenbacher Automation GmbH of Seevetal, both in Germany. It includes multiple CV-M50 cameras from JAI of Copenhagen, Denmark, which are 1/2-in. monochrome video CCD cameras. The system uses components and Common Vision Blox image processing software from Stemmer Imaging GmbH of Puchheim, Germany, in conjunction with a robot from Asea Brown Boveri Ltd. of Zurich.
Running on two conveyer belts, the system makes two passes. In the first, the dashboards are placed at the starting point, and the robot presents them in five positions to a total of eight cameras fitted to a steel structure. Sixty-five tests are carried out in five robot positions. In each position the VMT Vision Machine Technic image processing system records and evaluates multiple characteristics simultaneously. The characteristics are located on all sides of the dashboard.
Eight cameras are used to check a variety of characteristics on automotive dashboards in this image processing system.
After the tests in the last position, the image processing system transfers all the measurements to the robot control unit. It takes approximately eight seconds to evaluate the entire dashboard, including all the robot’s measurements.
When inspection is complete, the good dashboards are placed on one conveyer and sent to the production process, and faulty dashboards are sent along the second conveyer to the reworking point, where the positions of the faults are displayed on a digital dashboard on a large screen. Once reworking is complete, the dashboards are tested again. The image processing system remains available for automatic measurements throughout the reworking process.
The image processing software was developed in collaboration with customers from the automotive and automotive supply industry over several years in more than 500 projects. It does not require programming and includes a simple user interface that enables users to carry out tests independently after only a few days. Johnson Controls found that using the robotic vision system decreases manufacturing defects and increases product quality.
Meet the author
Andreas Tarnoki is sales manager of VMT Vision Machine Technic GmbH in Weinheim, Germany, a company of the Pepperl+Fuchs Group; e-mail: [email protected].