The weld seam is at the front

When assembling a fire extinguisher, it is essential to place the wall mount as exactly as possible on the weld seam of the container.
Up to now, the position of the weld seam was determined on the rotating container via indirect robot programming by means of a sensor.
The disadvantage of this method was the long cycle time and its unreliability: It was particularly difficult to detect the weld seams when were very flat or on containers which were distorted during welding. The system integrator Bieling therefore decided to use a vision sensor type FA 45.
The FA 45 is operated in freerun mode, i.e. it is not triggered externally - with the exception of the start signal which is given by a SensoPart FT50 RH sensor as soon as a new container arrives on the conveyor belt. After having received the start signal, the FA 45 „looks“ at the rotating container 15 times per second, until it detects the weld seam on the basis of the taught reference image. After having detected the weld seam, the sensor also determines its exact position. As soon as the targeted position is reached, the FA45 sends the release signal to the handling robot, which then removes the container in correct position and puts it into the holder for welding on the wall mount.

No weld seam is like the other
The problem with this application is that the appearance of the weld seams may vary strongly. The weld seams and also the container surface around them may be metallic bright or covered by weld slag. For this reason it was clear right from the start that an automation by means of a simple contrast sensor was out of the question. In order to still achieve a reliable measurement, the system integrator suggested to use the vision sensor of SensoPart.

Set-up of the FA 45 is very simple thanks to the intuitively operated PC configuration software. All settings are made on one single surface, it is not necessary to scroll menus. Graphic features such as contours, patterns and measuring windows may be easily selected in a live-image of the target object by means of the mouse.
Up to 32 features may thus be determined and consequentially connected. For the localisation of the weld seams, two windows were defined: one for the recognition of the pattern and one for its positioning. In order to detect the various weld seams in a reliable way, three different patterns with OR relation were taught.