Researchers at the Massachusetts Institute of Technology (MIT) presented earlier this month a developed artificial intelligence tool. This deep learning model will determine the stress and strain of a material through image analysis and computer vision. It can be used by engineers, mainly in the context of tests on materials.
The deep learning tool developed by Zhenze Yang (senior author and doctoral student at the Department of Materials Science and Engineering) Chi-Hua Yu (former post-doctoral fellow at MIT) and Markus J. Buehler (director of the atomic and molecular mechanics laboratory and professor of engineering at McAfee), uses computer vision to allow, among other things, to generate estimates of material constraints (fatigue, stress, strain, etc.) in real time.
A generative antagonist network (GAN) was formed using thousands of images matched to each other. Each image pair is made up for the first of the internal microstructure of the material when it is subjected to mechanical forces and for the second of the values of stresses and strains coded by color. The GAN takes into account game theory to determine the relationships between the appearance of the material and the stresses to which it is subjected.
Markus Buehler spoke about the use of this tool for problems related to stresses and strains to which materials are subjected:
“It is still a difficult problem. It is very expensive and it can take days, weeks or even months to run some simulations on the materials. So we thought: let’s teach an AI to solve this problem. […] From an image, the computer can predict all these forces: deformations, stresses, etc. ”
Artificial intelligence can also recontextualize issues like the development of cracks in a material. The neural network, once trained, can run on computer processors of all types. A use of this tool could therefore be envisaged in the field using a photo taken at time T, for example.
The researchers said the breakthrough could allow faster design prototyping and material inspections. Material testing is a fundamental step in understanding the structure, resistance and better anticipating risks. During these tests, engineers can reveal the internal forces of a material that can cause that material to warp or break. Such calculations could make it possible to anticipate the risks and understand how an X bridge would withstand in the midst of heavy traffic or high winds.