Lamp holder cement inspection

Special high power lighting requires a high quality standard for safety and endurance purposes. The light cap and the glass are bonded with a special adhesive to garantee a long life connection between these two parts. Damaged cement lamp holder Damaged cement lamp holder-2 inspection3_l Therefore, this adhesive bond must be equally spread out over the entire lamp cap before the glass and cap are bonded. Holes or irregulars in the adhesive ring are not allowed; otherwise, the lamp might break off when unscrewing the lamp from the lamp holder for replacement. Parts of the glass might stay in the lamp holder or wires so that the mains (high) voltage can be touched. This is dangerous for the end users. Here, a few rejected caps with a bad cement ring are shown.
A visual inspection of the cement is required to do this inspection.

Vision inspection setup

In this case, a programmable camera (INCA; INtelligent CAmera 311) is used. The program is developed on a personal computer by means of the program Clicks; a vision software environment of Philips Apptech Industrial Vision. This program contains very accurate measuring tools comparable with the library Promise. The test of this program can be done on the PC and can than be downloaded into the camera. With a startup command or after a power-up, the camera will run the program. Communication with a PLC is possible using the parallel I/O. A start command is given by the PLC and the camera acquires an image. After processing this image an accept or a reject command is given to the PLC. In case of a reject, the PLC shifts the lamp to a special control position so the lamp can manually be removed an reused.

Good cap no pol Reject cement no pol Cement missing no pol_1 The first tests are done with a red ring light from CCS (delivered by Vision Light Tech). Important is to create an image with a very high contrast between the cement and its environment. Before a detection of the cement can be done, the outside of the cap is detected first. When the outside is detected (circle fit) the expected cement ring is known, so the quality detection can be executed.
The first image does have a good cement ring, the second one is a rejected cap. As you can see on the second image, the cement is hanging on the inside and when a bulb is positioned in this cap the cement is pushed down. So a hole between the bulb and the cap is created during assembling.
Slip cap no polaroid inspection1_l
Important is to know the distance of the hanging cement to the center of the cap. Because of reflections in the bottom of the cap a reliable detection was not possible.
The third cap is missing a huge part of the cement. The light is reflected on the inner side of the cap so a missdetection was the result. Even the second image does have a small part of direct reflection of the ring light.
This image (left) shows a slip through cap. In this case the detection missed the missing cement and the ‘hanging’ cement.
Other problems in this set up are troublesome reflections of the outside of the cap. Because of these reflections, a reliable detection of the cap outside was not possible.

The consequences of all these problems are a change in the setup. A significant improvement is realised using Polaroid.
One Polaroid is mounted in front of the ring light the other Polaroid is mounted in front of the camera lens. Of course these two Polaroids have to be adjusted. Small error with polaroid_1 Reject with pol_1
When using these Polaroid‘s the metal (aluminium) cap holder and the outer side of the cap are dark (black) in the image. This means the detection of the outside of the cap causes a problem. In this case, the cap holder is slightly changed. A white Teflon ring is mounted.
The result is a white, diffuse background and a dark cap ring. A reliable detection of the cap position can now be achieved.
The cement is relatively bright and does have a relative constant grey level. There are no longer any reflections on the inner side of the cap. These results leads to a reliable detection of the cement.
The image left shows a good product, the image on the right shows a rejected cap. The (small) difference is the hole between the cement and the outside of the cap. The bright spot in the inner dark area (right down side) is a reflection of the production hall lighting.

software

After detecting the cap position (circle fit) a grey level detection on the cement is performed. Therefore the grey level cement ring is measured step wise. When a significant change is detected, it could be that a part of the cement is missing. Unfortunately this detection is not enough.
Program view Partially cement detection_1 On the left picture a view of the software (development environment) is given.
A relative grey level detection and a special differentiation algorithm is also used to detect changes in the grey level of the cement. After a missing part of cement is detected the inner side is measured so hanging cement will also be detected .

The images show a partial detection of a possible hanging cement ring (image left). Red shows a dark area in the cement ring, green the detection around that dark area .
Of course all relevant information can be shown on the screen and all the relevant parameters can be set. Images from rejected lamps during the test run (that had a duration of one year) are saved on a separate PC. Camera Quality last days
Before this camera test are performed every lamp is visually checked. It is very important to have a slip through of 0!!. Of course this is not as easy as it might seem.
To check the inspection quality, every inspected lamp is visually checked for one year. Also, the rejected lamps are checked, mainly by analyzing the saved images. Results do show a slip through of 0 for more than 1 million lamps tested!. The rejection count was lower than 0,4% for the whole year. To the left the last 25 days are shown in promille.

Nevertheless the interpretation of a wrongful rejected cap is done by different operators. This means the interpretation of the rejection limits is debatable.