Lighting the way to detecting tiny defects

A novel process is now pinpointing and classifying miniscule paint defects on large panels. Tom Shelley reports

A new technique allows the detection and classification of micron-sized defects in panels that are metres long – or sometimes the whole length of the car. The process uses LCD screens with moving bars as light sources and an array of three cameras attached to each screen. It is has been developed by a leading instrument maker, in collaboration with a manufacturer of upmarket cars. And though it is not yet considered applicable to mass production, there is no reason why it should not be in the future. Developed by Micro-Epsilon, working with BMW and Austrian company Atensor, the method uses the principle of deflectometry – reflecting structured light into a camera system. Traditionally, such inspections have been undertaken manually, sometimes helped by zebra striped light sources. In this case, bars of light move up a 40 inch diagonal LCD screen, which also carries the cameras. The whole system mounted on the end of a robot arm. The CCD cameras can assess one 600mm x 800mmm area in one recording. Larger areas have to be measured by using the robot to move the system to different positions. One requirement is for the system to ignore slight bulges in body panels that have no effect on customer appreciation of appearance, while showing up defects that have greater local curvatures – picked up as bursts of light by the cameras. This could be achieved by using fixed light sources. However, according to Micro-Epsilon UK managing director Chris Jones: “The pattern is shifted on the screens to increase resolution.” The underlying software is sophisticated enough to be able to null out badges and handles. As well as panels, the technology, dubbed ‘reflectControl’, has also been successfully applied to examining car windscreens, working in transmission mode. Windscreens are normally made up of several layers, which can result in various defects between the laminated sheets. The system has been installed on a robot arm at BMW Dingolfing and inspected around 36% of a car body for defects. BMW is currently evaluating the results and the plan is to examine whole cars, using four robots: two on each side, examining the cars, and two more, one on each side, marking up any detected defects for further examination and remedial work. Jones says this is currently going through the contractual process: “It is aimed at medium to high range cars. It is too slow for fast production lines, taking about 90 seconds per car. But, if you are buying a 5 series BMW, you don’t expect to find a dent or a chip in the paint.” Resolution is apparently 0.2mm square in the XY plane and sub-micron in the Z axis. It requires a highly reflective surface and so would work, for example, on lacquer-finished furniture, but not on textiles. Pointers * The system can detect and classify defects down to 0.2mm square in the XY plane and of less than a micron in height. It works on any highly reflecting surface * It takes about 10s to scan a 1m square panel and about 90s to scan a whole car * It also works in transmission mode, when it may be used to examine car windscreens