Fluid Mechanics

The Section for Mechanics works on various problems within fluid and continuum mechanics.  Typical topics for investigation include water suface waves and internal waves, tsunamis and freak waves, marine hydrodynamics, renewable energy including wave energy converters, floating solar panels and offshore wind, computational fluid dynamics (CFD), mechanics of soft materials, viscosity and elasticity, biomechanics and medical mechanics.  The Section has several laboratories, and count on some large installations including a wave basin for investigation of directional waves, two long and narrow wave flumes, and instrumentation for high-speed and high-resolution imaging.  The Section maintains close ties with several research institutions, including Simula Research Laboratory, University Hospital of Oslo, Institute for Energy Technology, Det Norske Veritas, etc.

Four sub-themes are offered by the Section for Mechanics.  All topics have their motivation from Fluid Mechanics.

NB! Please indicate clearly which of the sub-themes you have chosen for your proposal by using one of the codes FM1, FM2, FM3, FM4.

Mentoring and internship will be offered by a relevant external partner.

 

Theme FM1. Foundations of physics informed neural networks

  • Contact person: Kent-Andre Mardal
  • Keywords: Physics informed neural networks, constrained optimization

Physics informed neural networks (PINNs) are algorithms where neural networks learn from data while at the same time are constrained by physics, most commonly described in terms of partial differential equations (PDEs). As such, PINNs gives raise to non-linear optimization problems constrained by PDEs. The mathematical foundation of PINNs is at precent under-developed as compared with traditional modeling methods such as finite element methods, but recent efforts hints towards a general framework also for PINNs. Applications in focus is biomechanical modeling of the brain, in particular in the context of the glymphatic system. Research proposals that address the foundations of constrained neural networks are welcome.

External partners:

  • Simula Research Laboratory
  • University Hospital of Oslo
     

Theme FM2. Sustainable and renewable offshore energy

  • Contact person: John Grue
  • Keywords: Air-sea interaction with solar islands and offshore wind

Air-sea boundary layer (wind and waves) impacts on solar islands and offshore wind geometries. Outstanding research gaps include the frictional velocity (u star) in oblique waves, wind and wave interaction with geometries of solar islands, and turbulent air flows interacting with wind turbines in the ocean space or on land. Investigations may include analysis and post processing of data retrieved from experiments in brand new 3D wave basin facility including wind resources and 3D PIV and laser methods, may involve mathematical modelling of wave and wind interaction with floating geometries, and may involve aerodynamical CFD and turbulence models for wind-geometry analysis.
 

Theme FM3. Rogue waves

  • Contact person: Karsten Trulsen
  • Keywords: Rogue waves, criteria for prediction

Rogue waves are unexpectedly large waves. They occur stochastically in the ocean and along the coast, being a threat to people and activities. We still do not have any useful criterion for their prediction and forecasting.

We propose a machine guided search, using extended data sets of observational data and computer simulated data and laboratory data, to identify precursors for rogue waves in the ocean and along the coast. Our goal is to identify useful criteria for prediction and forecasting. We propose to check the quality of our findings in our Hydrodynamic Laboratory.

Publicly available data, both observational and hindcast data, are available from met.no. Also datasets of shallow water wave measurements from the North Sea and the Baltic Sea will be available.

Theme FM4. Deep Reinforcement Learning for Active Flow Control

  • Contact person: Mikael Mortensen
  • Keywords: Active flow control, Deep reinforcement learning, nonlinear flow scenarios

Deep reinforcement learning (DRL) for active flow control leverages neural networks and reinforcement learning algorithms to manage and manipulate fluid flows. In this context, an intelligent agent interacts with the flow environment, learning optimal control strategies through trial and error to achieve goals such as drag reduction or enhanced mixing. The agent receives feedback as rewards or penalties based on its actions, refining its policy over time. DRL has the potential to outperform traditional control methods in complex, nonlinear flow scenarios, especially when several agents (multi-agent DRL) learn to operate together. We are interested in any proposals that utilise DRL for active flow control. 

Topics from methodological research: 

  • Multi-agent deep reinforcement learning
  • Optimisation and efficiency, hardware acceleration (e.g., GPU), and distributed (MPI) learning.
  • Physics-based and hybrid physics-based and data-driven approaches  
  • Model-based reinforcement learning

Topics from natural sciences or technology:  

  • Active flow control and optimisation
  • Turbulence modeling and reduction
  • Boundary layer control to control separation and reduce drag
  • Heat transfer enhancement
  • Fluid structure interaction
  • Accuracy and efficiency of climate models and weather forecasting systems