Prof. Ghatu Subhash

Ghatu Subhash

Newton C. Ebaugh Professor
Mechanical and Aerospace Engineering
University of Florida, Gainesville, FL, USA

“Machine Learning Methods for Material Discovery, Constitutive Behavior,
and Defect Detection”

Abstract: The presentation will focus on application of machine learning (ML) for (i) discovery of materials with desired properties, (ii) development of interatomic potentials for shock response, and (iii) defect detection in composites.
In the first part, a high-throughput strategy is proposed to design compositionally complex ceramics for extreme environments by establishing design rules that permit accelerated discovery. To illustrate the concept, we first focus on a well-understood design space (Si-C-N system) by identifying new stoichiometries and structures. Evolutionary structure searches coupled with density functional theory (DFT) calculations are applied to predict the ground state and metastable structures. These searches aim to find structures with low energies and maximize the targeted property (e.g., hardness). The data obtained throughout these structure searches is exploited in a machine-learning model that is trained on the fly and can accelerate the structure prediction and provide an accurate and efficient surrogate model of the energy and hardness landscape. Through this framework, we aim to develop chemical design rules on how chemical additions affect hardness and stability.
In the second part, we explore development of machine learned interatomic potentials (MLIPs) to approximate the potential energy surface (PES) of complex ceramics (e.g., boron carbide, B4C) and learn fundamental structure-energy relationships that govern their performance. In our approach, the MLIP development encompasses three main components: (i) training data generated through DFT calculations, involving snapshots of static structures (e.g., atomic species and coordinates of four different polymorphs and their associated total energy, forces, and virial stresses for various strain levels in tension, compression, and shear loading as well as atomic configurations in different material phases), all extracted from ab initio molecular dynamics (AIMD) trajectories, (ii) material structure descriptor that describes the atomic environment in the form of a machine learnable input, and (iii) a learning algorithm based on neural networks (NN). This MLIP was used for simulating the shock response of boron carbide and for investigation of the mechanisms for amorphous bands formation in different crystallographic orientations.
Finally, the physics informed machine learning approach will be illustrated where the wave equation was enforced on the measured ultrasonic wave data to detect the nature and location of defects in isotropic and anisotropic materials.

BIO: Professor Ghatu Subhash obtained his PhD from University of California San Diego in 1991 and conducted his post-doctoral research at California Institute of Technology. He is Newton C Ebaugh Professor in Mechanical and Aerospace Engineering at University of Florida, Gainesville, FL. His research focuses on multiaxial behavior of advanced ceramics, metals, composites, gels and biological materials. He has developed novel experimental methods which have been patented and widely used. He has co-authored 220 peer reviewed journal articles (>11000 citations, h-index=58), 85 conference proceedings, 2-books, and 7 patents. His pioneering contributions in deciphering the complex deformation mechanisms in ceramics have been summarized in Progress in Materials Science and in a monograph Dynamic Response of Advanced ceramics (Wiley, 2021). He has advised 41-PhD students and seven post-doctoral research fellows. For his outstanding scholarship and mentorship, he was awarded the University of Florida Doctoral Dissertation Advisor/Mentoring Award (2021).
Professor Subhash has been conducting innovative experimental and computational research in the broad area of Solid Mechanics with emphasis on deformation and fracture behavior of advanced materials. In recognition of sustained and distinguished technical contributions, he received the 2024 Murray Medal and delivered the Conference Plenary lecture of the Society for Experimental Mechanics (SEM) and also at the International Conference on Experimental Mechanics (ICEM24). For his innovative and impactful contributions to the understanding of engineering ceramics he was awarded 2024 James I. Mueller Memorial Award and delivered the Conference Plenary Lecture of the Engineering Ceramics Division (ECD) of the American Ceramic Society (ACerS). For the development of a rapid processing method for ceramic (UO2) fuel pellets (patented) he received the ‘Significant Contribution Award’ (2014) from the American Nuclear Society. He also received B.J. Lazan Award (2021) and Frocht Award (2018 Experimental Mechanics Educator of the Year) from SEM, ‘Technology Innovator Award’ (2014, 2016) from UF, ASME Student Section Advisor Award, ‘SAE Ralph R. Teetor Educational Award’, and ‘ASEE Outstanding New Mechanics Educator’ award. He is a Fellow of ASME, SES, SEM, and ACerS.
Prof. Subhash is the Editor-in-Chief of Mechanics of Materials and an Associate Editor of Journal of the American Ceramic Society. He has delivered 2-day specialized technical courses for practicing engineers/scientists on ‘Dynamic Response of Materials’ at General Motors, Johns Hopkins University, and ACerS meetings. His research was showcased in a PBS documentary “Secrets of Spanish Florida” which aired nationwide in 2017 for unraveling the mystery behind the surprising impact-resistance of Coquina (the material with which the oldest fort in the USA – the Castillo de San Marcos, St. Augustine, FL, was built) against cannon ball impacts during the seize of the Spanish Fort in 1700s by the British army. He enjoys trekking, bicycling, and spending time outdoors with his family.