New AI system can predict durability of materials

A new artificial intelligence tool developed at MIT could drastically reduce the amount of time it takes to determine the toughness of a material.

A paper by MIT postdoc Chi-Hua Yu, civil and environmental engineering professor, department head Markus Buehler and Yu-Chuan Hsu at the National Taiwan University describes the system in the journal Matter, according to MIT News. 

The system is focused to predict the way a material would fracture by analyzing the cracks in the material's molecular structure. Normal molecular dynamics simulations take time because they require solving equations for every single atom, according to MIT News. Using a machine-learning system, they let computers observe the relationship between a material's characteristics and its performance. 

They generated hundreds of simulations and fed the data into the AI system to see if it would determine physical properties and predict the performance of a new material, which it did.

“That’s the really exciting thing,” Buehler told MIT News. "Because the computer simulation through AI can do what normally takes a very long time using molecular dynamics, or using finite element simulations, which are another way that engineers solve this problem and it’s very slow as well. So, this is a whole new way of simulating how materials fail.”

Knowing how materials fail is crucial to any engineering project because material failures account for the largest losses in any industry. 

The system could be applied to other processes that unfold over time such as diffusion or corrosion in addition to fracturing.