Material science is very important in the aviation, construction, automotive and many other industries. Determining the best material for the task is crucial in terms of structural integrity, safety and resource management. However, sometimes determining integrity of metals can be hard and expensive. But scientists from the University of Waterloo have found a better way.
People think that metal is very uniform, unlike wood or other natural materials. However, metal has grains, which may determine its strength. Different alloys have different properties that can only be determined by looking into molecular structure of the piece. Scientists have now developed a powerful AI model that can accurately detect different atomic structures in metallic materials. This technology allows detecting imperfections that previously would have slipped by unnoticed. Industry wants to use metal for consistency and strength, but each individual piece can have different properties which are difficult to detect – current methods fail to identify the symmetry in imperfect conditions. Scientists believe that their new AI-based method could lead to better material design and less manufacturing defects.
Scientists created an algorithm, which can create data which relates to the real world. Now they used this technology to generate about 80,000 images of the different kind of defects and displacements for an AI model. It can analyse different types of crystal structures in practical scenarios. This AI system could detect asymmetry in metals, which could result in catastrophic defects. Every metal is made up from different crystalline structures, which may not align perfectly and have weaker areas.
Metals are consistent only in theory – in reality you cannot trust them blindly in extremely demanding situations. One of the reasons why defect can appear in metal is cheap manufacturing, but they also just happen – metallurgy is a very demanding field. Devinder Kumar, one of the authors of the paper, said: “All these current methods fail when they try to match actual ideal structures, most of them fail when there is even one per cent defect. We have made an AI-based algorithm or model that can classify these kinds of symmetries even up to 40 per cent of defect”.
It will take some time until AI systems are used in factories to determine the quality of metal stock. However, it is pretty much inevitable, because such systems could improve the quality of the finished product making it safer and cheaper at the same time. This kind of quality control could benefit a variety of industries from aeronautics to car manufacturing and electronics.
Source: University of Waterloo