Plan rada Seminara Mehanika mašina i mehanizama - modeli i matematičke metode za JUN 2021.
UTORAK, 21.09.2021. u 17:00, Live stream
Chatterjee Tanmoy, College of Engineering, Swansea University, Swansea, Wales, UK
FROM DATA-DRIVEN TO PHYSICS-INFORMED MACHINE LEARNING FOR APPLIED SOLID MECHANICS RESEARCH
With the exponential growth of available data and high-end computing resources, recent advances in machine learning and predictive data analytics have revolutionized the scientific outlook across diverse disciplines. However, it is often observed while dealing with complex physical, biological, or engineering systems, data acquisition is either difficult or the associated cost is prohibitive. Hence, the prediction based on partial information makes the decision-making process less credible for real-life applications. In this regard, two classes of problems arising in mechanics of structures and materials, one of which is the availability of few data and the other is when there is known physics but no data, will be discussed. The first topic will comprise the application of adaptive supervised machine learning in the small data regime. The applicability of the data-driven framework will be demonstrated in stochastic modelling and design optimization. The second topic will cover the recently developed concept of physics-informed unsupervised (simulation-free) machine learning in capturing the dynamic system behaviour without the use of any data by utilizing prior information in the form of governing physical laws (ordinary or partial differential equations) that govern the space and time-dependent dynamical response. The encoded physics act as a regularizer that constrains the space of admissible solutions and conserves the physical laws. Finally, some interesting future research directions will be highlighted.
Seminar Mehanika mašina i mehanizama - modeli i matematičke metode započeo je sa radom u junu 2018.god. Seminar se održava do dva puta mesečno, utorkom u periodu od 17.00 - 19.00 u Matematičkom institutu SANU.