Scientific and physics-aware machine learning, and data assimilation
  • Luca Magri
    • Group
    • Collaborations
  • Publications
  • Research
    • Overview
    • Scientific machine learning >
      • Physics-aware machine learning
      • Chaotic time series forecasting
      • Nonlinear model reduction
      • Super-resolution and reconstruction
    • Real-time digital twins and data assimilation >
      • Inferring unknown unknowns: Bias-aware data assimilation
    • Optimization >
      • Bayesian optimisation
      • Chaotic systems
    • Mathematical modelling of multi-physics fluids >
      • Reacting flows and sound
    • Quantum computing and machine learning >
      • Solving nonlinear equations with quantum algorithms
      • Linear methods from quantum mechanics
    • Data and codes
  • Jobs/grants
  • Outreach
    • Research Centre in Data-Driven Engineering
    • Data-driven methods, machine learning and optimization
    • Data-driven Dynamical Systems Analysis
  • Consultancy
  • Teaching
    • University modules
    • Artificial intelligence for engineering
    • Mathematical methods
    • Misc
  • Contact

Overview

The three big U's in thermo-fluids
Bluff-body wake video.
The three big Us are (some of) the reasons for which it is difficult to time accurately predict the evolution of unsteady thermo-fluid dynamics. Example of a wake past a bluff body.
Simulation limits and capabality schematic
The chaotic nature of many flows, such as turbulent flows, makes the time-accurate prediction of the dynamics very difficult. This is because after a time horizon, which scales with the inverse of the Lyapunov exponent, a little perturbation to the flow equations is exponentially amplified in time. The perturbation can be caused by floating-point arithmetic, numerical schemes, number of processors, parallelization of the code, initial conditions, boundary conditions, parameters, etc.
Philosophy in learning.
It is important to understand human learning before doing machine learning!
Math modelling
1. Reduced-order models are practical tools but they have parameter/model uncertainties. 2. High-fidelity simulations are accurate tools but they cannot time-accurately capture rare/extreme events. 3. Data driven methods help combine data with models but we ought not to forget the physical constraints.
Adjoint methods schematic
Adjoint methods enable the fast and accurate calculation of the sensitivity of a cost functional when the design parameters are many. We use adjoint methods for stability calculations, data assimilation, uncertainty quantification, multiple-scale methods, optimization, and other applications.
© 2024 Luca Magri
  • Luca Magri
    • Group
    • Collaborations
  • Publications
  • Research
    • Overview
    • Scientific machine learning >
      • Physics-aware machine learning
      • Chaotic time series forecasting
      • Nonlinear model reduction
      • Super-resolution and reconstruction
    • Real-time digital twins and data assimilation >
      • Inferring unknown unknowns: Bias-aware data assimilation
    • Optimization >
      • Bayesian optimisation
      • Chaotic systems
    • Mathematical modelling of multi-physics fluids >
      • Reacting flows and sound
    • Quantum computing and machine learning >
      • Solving nonlinear equations with quantum algorithms
      • Linear methods from quantum mechanics
    • Data and codes
  • Jobs/grants
  • Outreach
    • Research Centre in Data-Driven Engineering
    • Data-driven methods, machine learning and optimization
    • Data-driven Dynamical Systems Analysis
  • Consultancy
  • Teaching
    • University modules
    • Artificial intelligence for engineering
    • Mathematical methods
    • Misc
  • Contact