Scientific Machine Learning - Where Physics-based Modeling Meets Data-driven Learning

Explore the emerging field of scientific machine learning in this lecture from the Santa Fe Institute. Delve into the synergy between physics-based modeling and data-driven learning, examining their application in complex scientific, engineering, and medical problems. Discover how this approach combines the predictive power and interpretability of physics-based models with the flexibility and computational scalability of machine learning. Learn about specific methods that integrate reduced-order modeling with machine learning, illustrated through examples such as modeling combustion in rocket engines and tubular reactors. Gain insights into the challenges and opportunities of harnessing data for knowledge extraction and decision-making in scientific contexts, and understand the importance of combining domain knowledge with modern data-driven techniques.

Scientific Machine Learning Where Physics-based Modeling Meets Data-driven Learning
Scientific Machine Learning What are the opportunities and challenges of machine learning in complex applications across science, engineering, and medicine?
How do we harness the explosion of data to extract knowledge, insight and decisions?
Example: modeling combustion in a rocket engine Conservation of mass (p), momentum (w), energy (E)
There are multiple ways to write the Euler equations
Introducing auxiliary variables can expose structure - lifting
Lifting example: Tubular reactor
Modeling a single injector of a rocket engine combustor
Performance of learned quadratic ROM
Data-driven decisions