Engineering Software 2.0

Our laboratory is pioneering a new generation of engineering software that will unify training, solving, and calibrating under one single computational paradigm. The work is a result of a series of papers starting from Hierarchical Deep Learning Neural Network (HiDeNN) by Saha et al. back in 2021. This new AI technology has proved to be orders of magnitude faster and much more memory conserving comapared to many contemporary deep learning frameworks used in engineering.

Real-time Control and Monitoring of AM Systems

One brilliant application of the proposed engineering software 2.0 is the real time online control of complex manufacturing systems. We have implemented our method to achieve real-time control of laser powder-bed additive manufacturing system and currently continuing to enhance the capabilities of the method to handle larger parts.

Multi-scale Models for Designing Additively Manufactured Aerospace Alloys

Design of alloys has been taken to a new dimension by the 3D printing technology. But still a lot of gaps exist in the knowledge required to develop a process-structure-property map for 3D printed alloys. Our research in this field combines experimental characterization data, advanced data science methods, high-fidelity process simulation and crystal plasticity methods to figure out a suitable process map for 3D printing technology to optimize the mechanical properties of the alloys. 

Design of Fatigue-resistant Alloys

Fatigue of metallic alloy components is a key consideration in design of useful parts. Our research combines artificial intelligence, especially, transfer learning, dimensionless learning, with reduced-order Crystal Plasticity-based simulation to design fatigue-resistant alloys.

Modeling Materials at the Nanoscale

"There is plenty of room at the bottom"-R. Feynmann 

Mechanical response of alloys structures at the macroscale is determined by the material science and physical phenomena at the nanoscale. Our lab has performed massively parallel molecular dynamics simulation to understand the mechanical properties of nanoscale structures and design new forcefields for the community to use.

HiDeNN for Biomechanics

Our group has done some interesting work to combine full field finite element simulation data with Hierarchcial Deep learning Neural Network (HiDeNN) by Saha et al 2021 to come up with reduced order models for predicting the shape of adolescent idiopathic scoliocis patients. The work was in collaboration with Ann & Robert H. Lurie Children’s Hospital of Chicago, IL, USA.