Masterclass with Prof Roy
- When:
- Tuesday
18 June 2024 - Time:
- 10am-12 noon
- Where:
- UWA Institute of Advanced Studies
Physics Informed Machine Learning in Engineering Applications
An Institute of Advanced Studies Masterclass with Professor Sitikantha Roy, Professor of Applied Mechanics and School of Artificial Intelligence Indian Institute of Technology Delhi and UWA Gledden Visiting Fellow.
This Masterclass will present a gentle introduction to Physics Informed Machine learning at the intersection of scientific simulation and modern machine-learning approaches for Physical System Modelling. The session will conclude with a discussion of PiMLs applications in some biomechanical, solid mechanics problems.
Agenda
- What is Physics Informed Machine Learning? Its types: physics guided, physics informed, physics encoded
- How physics comes in for model development.
- How data-driven models are framed?
- What is generalizability, interpretability and parsimony?
- How physical bias can come in data
- Data-driven discovery of physical equations
- Basic neural network architecture, its training and testing.
- How data driven discovery can be made for nonlinear dynamical systems, e.g, SINDy
- Some applications of ScML in mechanics problems
Professor Sitikantha Roy is a 2024 UWA Gledden Visiting Fellow. He is a joint faculty member in the School of Artificial Intelligence and the Department of Applied Mechanics at IIT Delhi. He runs a lab named Biomechanics and Soft Robotics at IIT Delhi. His research interest is in Brain Biomechanics, AI in Human-centred dynamical systems, intelligent wearable robotics, artificial muscle (soft actuator), and predictive digital twins for human-in-loop design method. His lab develops tools, numerical algorithms, and generalized computational frameworks to capture the deformational pattern under the applications of external/internal stimuli or forces. Once understood, the next aim is to design and control the morphology (soft robotics in particular) in symbiosis with the human body (wearable robotics), predict pathology (neurodegeneration, etc) and aid in reducing trauma (surgical simulation). Professor Roy uses a combination of differential equations based on scientific "simulation" as well as modern data-driven "artificial intelligence" based physical system modelling and analysis. He also does limited experiments for generating data in a structured/controlled environment.