NDT unlocks die cast flaws in 3D
Researchers in Chile have developed a non-destructive method of detecting and imaging aluminium die casting defects in 3D that they believe could enhance analysis of more complex flaws.
The group, led by the Professor Domingo Mery at the Pontificia Universidad Católica de Chile, Santiago, has simultaneously gathered information from multiple viewpoints by defining a ‘sliding 3D box’ on a test area (see diagram, below) that corresponds to the size of an expected defect, for instance 8x8x8mm. A sequence of X-rays are then taken from different angles and combined. ‘Two or more views of the same object taken from different viewpoints can be used to confirm and improve the diagnostics done by analysing only one image,’ explains Mery.
The concept of the 3D sliding box came from PhD student Christian Pieringer. Mery continues, ‘The key idea is that we move the slide box inside the object, and we can estimate (using a geometric model) portions of each X-ray image and classify them. We measure, using image processing techniques, visual attributes from these views, like contrast and saliency, and we can study the attributes of defects and no defects. The trained classifier [an algorithm] can then automatically discriminate from the mentioned measurements if a 3D box is a defect or not – defects have high contrast, no defects have low contrast’.
Mery claims therefore that, although ‘there are many fully automated [detection] systems with high performance rates’, this technique is more capable of dealing with variables in image intensity and would therefore be suitable for complex parts and defect analysis. ‘We can guarantee that the views we are analysing correspond to a specific 3D space,’ he insists, eliminating false alarms. In laboratory testing, 94% of true positives were detected, with 95% sensitivity in real flaws.
‘Classical methods have several disadvantages, such as the complexity of their configuration and their inflexibility to any changes in the design of the workpiece,’ adds Mery.
John Taggart of technical consulting firm Serco Assurance, comments, ‘The idea of using a few shots at different angles is nothing new, though I have not seen the use of 3D scans with a small region common to each scan. This is a useful way of characterising defects. Any defects in the small region would then be imaged in at least some of the shots so that their position and geometry can be determined much better’.
He believes, however, the approach may be better suited to volumetric defects rather than planar defects such as cracks. ‘With planar defects, the image may not be discernible unless the source-film direction is close to the plane of the defect. ‘[Furthermore] the method potentially involves a large number of shots depending on the number of 3D regions inspected. [This would need] a convenient means of adjusting the source/film positions, and also that of the component under test. In practical terms, it would require short exposures, so it may be better for low attentuation components (such as aluminium of limited size) than for, say, large steel components’.