[ad_1]
You might not know it from the book’s cover, but according to MIT researchers, you can now make the equivalent for all kinds of materials, from airplane parts to medical implants. Their new approach allows engineers to find out what’s going on inside simply by observing the material’s surface properties.
The team used a type of machine learning known as deep learning to compare large sets of simulated data about the materials’ external force fields and the corresponding internal structure, and used this to create a system that could reliably predict the interior from the surface. data.
The results are published in the journal Advanced materialsIn a paper by PhD student Jenze Yang and Markus Bühler, professor of civil and environmental engineering.
“This is a very common problem in engineering,” Buhler explains. If you have a piece of material—maybe it’s a car door or an airplane part—and you want to know what’s inside that material, you can measure the stress by taking pictures of the surface and calculate how much deformation you have. has. But you can’t really look inside the material. The only way you can do that is to cut it open and then look inside and see if there is any damage.
X-rays and other techniques can also be used, but these are usually expensive and require bulky equipment, he says. “So what we’ve done is basically ask: Can we develop an AI algorithm that looks at what’s going on at a surface that we can easily see either using a microscope or taking a photo of, or maybe just measuring things. surface of the material and then try to figure out what’s really going on inside?” This internal information may include any damage, cracks or stresses in the material, or details of its internal microstructure.
The same questions may apply to biological tissues, he adds. Is there disease, or any kind of growth or change in the tissue? The goal was to develop a system that could answer these types of questions in a completely non-invasive way.
Achieving that goal involved overcoming difficulties, including the fact that “many of these problems have multiple solutions,” says Buehler. For example, many different internal configurations can have the same surface properties. To deal with this uncertainty, “we’ve created methods that can give us all the possibilities, all the options, basically, that could lead to this particular [surface] Script.”
The technique they developed involved training an artificial intelligence model using large amounts of information about surface measurements and related interior properties. This included not only homogeneous materials, but also a combination of different materials. “Some of the new aircraft are made of composites, so they have a purposeful design of different phases,” says Buehler. “And of course in biology as well, any kind of biological material is made up of many components and they have very different properties, for example in bone, where there is very soft protein, and then very hard mineral substances.”
The technique works even on materials whose complexity is not fully understood, he says. “With complex biological tissue, we don’t understand exactly how it behaves, but we can measure the behavior. “We don’t have a theory for this, but if we collect enough data, we can prepare a model.”
Young says the method they developed is widely used. “It’s not just limited to solid mechanics problems, but can be used in different engineering disciplines like fluid dynamics and other types.” Bühler adds that it can be used to determine various properties, not only stress and strain, but also fluid fields or magnetic fields, such as magnetic fields inside a fusion reactor. It is “very universal, not only for different materials, but also for different disciplines.”
Yang says he first started thinking about this approach when he was studying data on a material where part of the image he was using was blurry, and he wondered how the blurry area could be “gapped in” for missing data. “How can we recover this lost information?” he thought. On further reading, he discovered that this was an example of a widespread issue known as the inverse problem, which attempts to recover lost information.
The development of the method involved an iterative process of making preliminary predictions for the model, comparing this with actual data for the material in question, and then fine-tuning the model to match this information. The resulting model was tested on cases where the materials are sufficiently well understood that the true intrinsic properties could be calculated, and the new method’s predictions were in good agreement with these calculated properties.
The training data included images of surfaces, but also measurements of various surface properties, including stresses and electric and magnetic fields. In many cases, researchers used simulated data based on an understanding of the basic structure of the material in question. And even when the new material has many unknown characteristics, the method can still produce an approximation that is good enough to give engineers a general direction on how to make further measurements.
As an example of how this methodology can be used, Buhler points out that aircraft today are often inspected by testing a few representative areas with expensive methods such as X-rays, as it would be impractical to test the entire aircraft. “It’s a different approach where you have a much cheaper way to collect data and make predictions,” says Buehler. “So you can make decisions about where you want it to look, and maybe use more expensive equipment to test it.
To begin with, he expects the method, which is freely available for anyone to use via the website GitHub, will be used primarily in laboratory settings, such as testing materials used for soft robotics applications.
Of such materials, he says, “We can measure things on the surface, but we have no idea what’s going on inside the material many times because it’s made of hydrogels or proteins or biomaterials for actuators and there’s no theory. for that. So this is an area where researchers can use our techniques to make predictions about what’s going on inside and maybe make better grips or better composites,” he adds.
The research was supported by the US Army Research Office, the Air Force Office of Scientific Research, the GoogleCloud Platform, and the MIT Quest for Intelligence.
[ad_2]
Source link