Advanced materials and manufacturing are important pillars and drivers of today’s
economy. They have made consumer devices like TVs and computers ubiquitous by enabling
cheap mass production of reliable products, and enabled industrial applications on
a scale not seen before. However, designing new materials is a slow and laborious
process, requiring significant investments in capital and labor.
Recent advances in artificial intelligence allow to improve and speed up this process.
In particular, machine learning techniques can replace expensive and time-consuming
laboratory processes and computational simulations to quickly and reliably predict
how to create materials with desired properties and how materials will behave in specific
circumstances. So-called surrogate models learn how properties of materials and experimental
parameters relate to experimental results.
AIM applies these techniques to materials engineering and manufacturing to reduce
cost and improve performance. The center merges synergistic expertise at the College
of Engineering and Physical Sciences in materials development, chemical engineering,
mechanical engineering, and computer science for the development of powerful methods
to design and model the behavior of advanced materials and manufacture advanced devices.
Graphene oxide can be treated with lasers to create patterns of conducting carbon
within a surrounding environment of insulating materials. This allows to create nano-scale
circuits in thin sheets of materials, enabling next-generation devices. Laser patterning
is a complex process that requires precise real-time control to achieve the desired
results. We apply techniques from artificial intelligence to achieve the necessary
Determining the limits of forces that can be applied to composite materials before
they break is an important problem in engineering and manufacturing. The field has
developed a number of failure criteria, which allow to describe those limits in a
closed functional form, so called failure envelopes. However, these functional forms
are not very flexible, and require significant experimentation with new materials
to be determined. We apply machine learning techniques for more flexible failure envelopes
and to predict them based on very little data for new materials.
Composite microstructures observed in nature serve as templates for the development
of new synthetic materials. However, it has proven difficult to reproduce the properties
found in natural materials due to the interaction of intricate structures at varying
length scales. Rather than attempting to replicate these materials (biomimetics),
the focus of this work is to use a bio-inspired pattern generation algorithm (the
Gray-Scott model), optimized by Bayesian methods from Artificial Intelligence, to
design new materials with high energy absorption properties.
Center PIs Johnson, Kotthoff, and Aidhy have been awarded a $750,000 grant from NASA
EPSCoR to investigate ways of manufacturing flexbile electronics in space. The grant
builds on expertise developed in the center on combining machine learning with advanced
materials and manufacturing and will see the PIs collaborate with researchers at NASA
Ames over three years to develop approaches for manufacturing electronics with low
power and resource requirements that can be used for example aboard the International
More information is available in a UW press release. The grant was also covered in an NPR news item.
The AIM center organized a workshop on AI and advanced manufacturing, co-located with
the 2019 University of Wyoming Materials Science and Engineering Symposium. Invited
speakers included David Estrada (Boise State University) and Kamal Choudhary (National
Institute of Standards and Technology). In addition, AIM PIs and associated researchers
presented the latest results on applying state-of-the-art AI to laser-induced graphene
production, atomistic modeling, and porous media modeling.
The workshop featured a hands-on session where participants were able to use state-of-the-art
machine learning modeling to optimize manufacturing processes themselves, based on
real data and approaches developed by the center PIs. This was the first exposure
to AI techniques for many of the approximately 25 participants, giving them an idea
of the power of the approach and how to apply it in their own research.
Wahab, Hud, Vivek Jain, Alexander Scott Tyrrell, Michael Alan Seas, Lars Kotthoff,
and Patrick Alfred Johnson. "Machine-Learning-Assisted Fabrication: Bayesian Optimization
of Laser-Induced Graphene Patterning Using in-Situ Raman Analysis." Carbon 167 (2020):
609–19. https://doi.org/10.1016/j.carbon.2020.05.087. preprint (PDF)
Wahab, Hud, Alex Tyrrell, Vivek Jain, Lars Kotthoff and Patrick Johnson. "Model-based
Optimization of Laser-Reduced Graphene using in-situ Raman Analysis." Poster presentation in Materials Research Society Fall 2019 Symposium (PDF), December 2019.
Kotthoff, Lars, Vivek Jain, Alexander Tyrrell, Hud Wahab, and Patrick Johnson. "AI
for Materials Science: Tuning Laser-Induced Graphene Production." Poster presentation in COSEAL annual meeting (PDF), August 2019.
Hankins, Sarah, Lars Kotthoff, and Ray S. Fertig. "Bio-like Composite Microstructure Designs for Enhanced Damage Tolerance via Machine
Learning. (PDF)" In American Society for Composites, September 2019.
Kotthoff, Lars, Vivek Jain, Alexander Tyrrell, Hud Wahab, and Patrick Johnson. "AI for Materials Science: Tuning Laser-Induced Graphene Production. (PDF)" In Data Science Meets Optimisation Workshop at IJCAI 2019, August 2019.
Bhuiyan, Faisal H., Lars Kotthoff, and Ray S. Fertig. "Machine Learning Applied to
Biaxial Failure Envelope Prediction of Unidirectional Composites." Poster presentation in University of Wyoming Materials Science and Engineering Symposium
(PDF), April 2019.
Jain, Vivek, Alex Tyrrell, Kaitlyn Vap, Lars Kotthoff, and Patrick Johnson. "Advanced
Manufacturing of Laser-Induced Graphene Electronics" Poster presentation in University of Wyoming Materials Science and Engineering Symposium
(PDF), April 2019.
Bhuiyan, Faisal H., Lars Kotthoff, and Ray S. Fertig. "A Machine Learning Technique
to Predict Static Multi-Axial Failure Envelope of Laminated Composites." In American
Society for Composites 33rd Annual Technical Conference, 2018. preprint (PDF)
Lars Kotthoff, email@example.com