Isaac Newton may have met his match.
For centuries, engineers have relied on physical laws – developed by Newton and others – to understand the stresses and stresses on the materials they work with. But solving these equations can be a math job, especially for complex materials.
MIT researchers have developed a technique to quickly determine certain properties of a material, such as stress and strain, based on an image of the material showing its internal structure. The approach could one day eliminate the need for arduous physics-based calculations, relying instead on computer vision and machine learning to generate real-time estimates.
The researchers say the breakthrough could allow faster design prototyping and material inspections. “It’s a whole new approach,” Zhenze Yang says, adding that the algorithm “completes the whole process without any knowledge of the field of physics.”
The research appears today in the journal Scientific advances. Yang is the lead author of the article and a doctoral student in the Department of Materials Science and Engineering. Co-authors include former MIT post-doctoral fellow Chi-Hua Yu and Markus Buehler, McAfee professor of engineering, and director of the Atomic and Molecular Mechanics Laboratory.
Engineers spend a lot of time solving equations. They help reveal the internal forces of a material, such as stresses and stresses, that can cause that material to warp or break. Such calculations could suggest how a proposed bridge would withstand the midst of heavy traffic or high winds. Unlike Sir Isaac, engineers today do not need pen and paper for this task. “Many generations of mathematicians and engineers wrote these equations and then figured out how to solve them on the computer,” says Buehler. “But it’s still a difficult problem. It’s very expensive – it can take days, weeks, even months to run simulations. So, we thought: let’s teach an AI to do this problem for you. “
The researchers turned to a machine learning technique called a Generative Adversity Neural Network. They formed the network with thousands of paired images – one representing the internal microstructure of a material subjected to mechanical forces, and the other representing the color-coded stress and strain values of that same material. With these examples, the network uses the principles of game theory to iteratively determine the relationships between the geometry of a material and its resulting stresses.
“Thus, from an image, the computer is able to predict all these forces: deformations, stresses, etc.”, explains Buehler. “This is really the breakthrough – conventionally, you would need to code the equations and have the computer solve partial differential equations. We’re just going frame by frame.”
This image-based approach is particularly advantageous for complex composite materials. Forces on a material can work differently at the atomic scale than at the macroscopic scale. “If you look at an airplane, you might have glue, a metal, and a polymer in between. So you have all these different faces and different scales that determine the solution,” Buehler explains. “If you go the hard – Newton’s way – you have to take a huge detour to get to the answer.”
But the researcher’s network knows how to manage several scales. It processes information through a series of “convolutions”, which analyze images at increasingly large scales. “This is why these neural networks are ideal for describing the properties of materials,” says Buehler.
The fully formed network performed well in testing, successfully rendering stress and strain values from a series of close-up images of the microstructure of various flexible composite materials. The network was even able to capture “singularities”, like cracks developing in a material. In these cases, the forces and fields change rapidly over small distances. “As a materials specialist, you would like to know if the model can recreate these singularities,” says Buehler. “And the answer is yes.”
The breakthrough could “dramatically reduce the iterations needed to design products,” said Suvranu De, a mechanical engineer at Rensselaer Polytechnic who was not involved in the research. “The end-to-end approach proposed in this article will have a significant impact on a variety of engineering applications – from composites used in the automotive and aerospace industries to natural and engineered biomaterials. It will also have important applications in the field of pure scientific research, as force plays a critical role in a surprisingly wide range of applications from micro / nanoelectronics to cell migration and differentiation. “
In addition to saving engineers time and money, the new technique could give non-experts access to advanced material calculations. Architects or product designers, for example, could test the viability of their ideas before passing the project on to a team of engineers. “They can just draw their proposal and find out,” says Buehler. “It’s a big problem.”
Once formed, the network runs almost instantly on consumer computer processors. This could allow mechanics and inspectors to diagnose potential problems with the machines simply by taking a photo.
In the new paper, the researchers mainly worked with composite materials that included both soft and brittle components in a variety of random geometric arrangements. In future work, the team plans to use a wider range of material types. “I really think this method will have a huge impact,” says Buehler. “Empowering engineers with AI is really what we’re trying to do here.”
Funding for this research was provided, in part, by the Army Research Office and the Office of Naval Research.