A versatile tool, digital tomography can be automated and brought closer to the production line to better control product quality. Its software is expanding to meet as many applications as possible.
In 2005, the German company Werth presented the Tomoscope, the first digital X-ray (RX) tomograph which deduced dimensional measurements from the scan of a manufactured part. Coming from the medical world and used in the 1980s across the Rhine by Eurocopter for the health-material inspection of rotor blades, tomography was breaking into the art of metrology.
Thanks to X-rays’ ability to precisely probe the interior of matter, a single examination provides external and internal ratings, while revealing invisible defects, such as porosities and inclusions. Non-destructive testing, dimensional and geometric analysis for comparison with the nominal CAD file: this unique combination is appreciated by manufacturers manufacturing complex parts, whose internal structures are inaccessible to 3D scanners in particular.
The applications are varied, including prototyping, the review of the first part in the event of modification of the production tool, reverse engineering and of course production control. The automotive and aeronautical sectors are looking at this technology, in particular for the control of plastic parts with complex geometries. The plastics industry is getting more and more equipment. “Plastic components are becoming very technical, with many functional qualities,” explains François Lambert, the manager of GOM France. Additive manufacturing is also demanding, because the parts are made from a single block and sometimes structural, combining mechanical resistance and lightness. »The power of the X-ray source of a tomograph is suitable for most densities of materials used in the manufacturing industry: 100 kV for plastics, 150 kV for light alloys, 200 kV for aluminum and titanium, and 300, 450 and 600 kV for steels.
But a digital tomograph costs at least 200,000 euros and its implementation is difficult, much more than a thermographic control device or ultrasonic. Few of the suppliers, such as Nikon Metrology, Zeiss, GOM (Zeiss), Werth, Waygate Technologies (formerly GE Inspection Technologies), therefore seek to improve their productivity, linked to their speed, precision and progressive automation. The 4000 x 4000 pixel sensors have been available since 2015, but are currently reserved for expertise and R & D.
A scan per minute
“The reduction in the duration of the acquisition / reconstruction cycles of 3D data and improving the resolution is the main trend of recent years, confirms Sylvain Genot, sales engineer for the RX part at Nikon Metrology France. The power of the X-ray source and the sensitivity of the sensor are the two factors to shorten the scan time. »
A more powerful X-ray source means in fact a higher photon flux: the rotation of the part on the tomograph manipulator, which is used to project the 2D sections onto the detector, can be faster. The speed gain is all the more important the more sensitive the detector is. The laws of physics however impose inevitable compromises: the more the X-ray source increases in power, the more the measurement loses in precision, because of the diffusion of X-rays.
Appeared over the years 2000, the microfocus source, with a power of around 300 kV, nevertheless made it possible to reduce the focal spot and obtain precision of the order of a micron on small objects. Innovations on the part of manufacturers, like the rotating target introduced by Nikon in 2011, also help to accelerate the acquisition phase.
“In 2016, a scan took two minutes, explains Sylvain Genot. Today, in the automotive industry, the same operation on a plastic or aluminum part is of the order of a minute. The reconstruction cycle of the 3D volume induces an additional one or two minutes. And the health-material and dimensional information is already available. »
The evolution of the reconstruction time decreases in parallel. But the computer resources required, sized by the need and the size of the detector, are important: the files are heavy and the mathematical algorithms to reconstruct a 3D mesh from the cloud of points (or voxels, the equivalent of the pixel for 3D ) are demanding. “Based on a flat panel collector