Assembly errors quickly identified

If errors creep in during the assembly of components, costly post-processing is often the consequence. Automatic testing is difficult, especially where individual products are concerned. Now there is a new testing system that is flexible and economical, even for smaller production runs.

Caption: Parts that have not been installed properly are highlighted on the monitor screen. © Fraunhofer IFF

Today's cars are increasingly custom-built. One customer might want electric windows, heated door mirrors and steering-wheel-mounted stereo controls, while another is satisfied with the minimum basic equipment. The situation with aircraft is no different: each airline is looking for different interior finishes - and lighting, ventilation, seating and monitors are different from one company to the next.
 
Yet the customer's freedom is the manufacturer's challenge: because individual parts and mountings have to be installed in different locations along the fuselage, automated assembly is often not an economical alternative. For many assembly steps, manufacturers have to rely on manual labour instead. But if errors creep in - if, for instance, a bracket is mounted backwards or in the wrong place - correcting them can get expensive later on.
 
The fuselage has to be reworked at great expense.

Today, employees use design drawings to determine whether the individual parts have been attached properly, or else manufacturers use rigid and inflexible testing systems to check the part against comparison photos. This calls for an identical part for the template photo - and that can be difficult where one-off parts are concerned.
 
Now, researchers at the Fraunhofer Institute for Factory Operation and Automation IFF in Magdeburg have come up with a testing technology that is reliable and economic even for one-off production runs.
 
''The automated visual testing system generates a digital template and uses it to compare with the assembled components. It reliably identifies any errors,'' points out Steffen Sauer, project manager for measuring and testing technology at IFF.
 
First, an automated camera system takes hundreds of photos of individually assembled holders, load-bearing elements and parts on the inside of the fuselage shell. For every picture taken, the system determines the exact position of the camera relative to the fuselage shell. At the same time, the software generates the same shots again - but this time using a ''virtual'' camera.
 
So essentially it creates ''photos'' using the data of the digital design model. The system compares the photos of real parts with the ''virtual'' images. If the system detects any deviations - if, say, a bracket is backwards - it issues a warning. Parts that have not been installed properly are highlighted on the monitor screen. The interesting thing is these steps are completely automatic.
 
In addition to a two-dimensional check using the photos, the system can also check a completed aircraft fuselage in three dimension: as in the case of the photos, it uses design data to generate three-dimensional data that it then compares with measurements on the real assembly. Here, conventional 3D measurement methods are used to digitalize the components.
 
''What's new about this system is that we convert specifications from the design models into images and 3-D data that the system can then compare with the real images,'' explains Sauer.

The system also automatically draws up the testing plan: first, it identifies the best measuring position for every part to be tested. What is the best location from which to test the component in question? The system forwards the results to the robot, which in turn travels to the position identified, where it shoots the two or three-dimensional images. There is another advantage to this approach: this way, the system can react quickly and flexibly. The result is a continuous process, from design to the finished and assembled part.
 
The main challenge to researchers was to set up the virtual camera that uses design models to ''photograph'' the as yet non-existent component. Another crunch point was to quickly and automatically locate the interesting areas from among the many millions of points in the three-dimensional images - to hunt down, in the mass of points, the tiny components such as brackets and holders, and to check to see whether they are properly fitted.

The fields of application for this testing technology are diverse: the technology can be used wherever flexibility is required and individual parts frequently change. The only condition: design data must be available. Researchers will be presenting the new technology at the Control trade fair, May 8-11 in Stuttgart (Hall 1, stand 1502). The process will be ready to use by summer 2012.