Session Spotlight

Title: “Scientific Machine Learning for Wave-Structure Interaction Modelling: Uncertainty Quantification, Predictable Zone, and Structural Responses”

Speaker: Jincheng Zhang

Dr. Jincheng Zhang is an Assistant Professor at the School of Engineering, University of Warwick. He obtained his B.S. and M.S. degrees in Mechanical Engineering from Tsinghua University in 2015 and 2018, respectively, and his PhD from the University of Warwick in 2021. His research focuses on data-driven and physics-informed deep learning, wave and wind energy, CFD simulations, and uncertainty quantification.

Title: “Foundational AI Directions for CFD and Multi-Physics”

Speaker: Christopher Pain

Prof Christopher Pain (Earth Science and Engineering Department, Imperial College London) leads the Applied Modelling and Computation Group (AMCG) at ICL. He is the director of the data assimilation lab in the Data Science Institute (DSI) at ICL and is co-director of the Centre for AI-Physics Modelling at Imperial-X. His interests are in foundational AI modelling for the environment, health and wellbeing, energy and industry.

Abstract: Recent developments in AI are transforming a large number of fields and are now starting to make a major impact in computational physics. Here we describe some of these innovative AI techniques that have been recently developed and how they can work together. We will indicate how AI may be deployed for modelling of environmental flows. Recent advances have enabled AI software to solve, to within numerical tolerances, the discrete differential equations that govern the physics of fluids (AI4PDEs). Important also is the use of AI to solve particle systems (e.g. AI for Discrete Element Modelling (AI4DEM)) which are expressed by interparticle forces and Lagrangian particle motion. The presentation will provide a summary of these forward models and a view on how these new approaches can be used with trained foundational AI models to form even more powerful methods.

Title: “AI enhanced offshore vessel coordination. A transformation from deep learning research to commercialisable AI” 

Speaker: Jana Stella

Jana is NeuWave’s CEO. She is a PhD Researcher in Environmental Engineering at UoM and her research has pushed the boundaries of Oceanography, where she has developed multiple numerical and AI wave models to understand environmental conditions with limited computer resources. Her innovation experience also stretches across medicine and nuclear fusion.

Abstract: Offshore vessels are the back-bone to our offshore renewables transition. With floating offshore wind posing challenges in construction and operations and maintenance coupled with a lack of recorded data, where can AI solutions be useful? 

Title: “Bayesian and Surrogate-Assisted Optimisation for CFD”

Speaker: Gavin Tabor

Professor Gavin Tabor is a member of the Centre for Water Systems and the Computational Engineering group. He graduated from Christs College Cambridge in 1990 with a 1st in Theoretical Physics, then did a Ph.D. in Theoretical Astrophysics at the Department of Physics at Oxford. Changing both location and research area, he then worked as a RA in Prof. David Gosman’s research group at Imperial College, London, for 5 years. During this time he worked on Computational Fluid Dynamics (CFD) of multiphase flows and the modelling of premixed turbulent combustion, and contributed towards the CFD code now known as OpenFOAM. He is  a Fellow of the Institute of Physics and member of the Computational Physics Group committee of the IOP, Chair of the Joint Technical Committees for the OpenFOAM Governance effort, and a member of the international OpenFOAM Workshop Committee.

Title: “Leveraging Uncertainty Quantification in Complex Environments“

Speaker: Daniel Greenhouse 

Daniel Greenhouse has a background in plasma physics conducting his PhD across the University of York and the UK Atomic Energy Authority. Specifically he looked at using machine learning for diagnosing the crucial interaction between fusion plasmas and surrounding wall material. Daniel applies this insight at digiLab: digiLab uses explainable AI to transform complex challenges into innovation in safety critical industries where uncertainty matters.

Abstract: digiLab have applied numerous advanced artificial intelligence techniques across a range of safety critical industries. These demand crucial components that are often overlooked in the world of machine learning: explainability and uncertainty quantification (UQ). This talk will outline how UQ brings additional benefits, transforming the approach to problems such as sensor placement, emulation of complex components, and inverse methods for finding key, underlying quantities. Specifically, we will look at how they can be applied to the world of wave structure interaction.  

Title: “Hybrid PE-ML Method for nonlinear WSI“

Speaker: Qingwei Ma

Qingwei Ma, Professor of Hydrodynamics, is working with City St George’s, University. He has leaded development of numerical methods for wave structure interaction including QALE-FEM, qaleFOAM and ESBI. Recently he leads to develop the Hybrid PE-ML method.  His team has been awarded by international bodies a few times. 

Abstract: Machine learning (ML) applications have become more and more popular, and helped address many issues at a surprising speed we are faced in many areas across our society, such as finding the fastest route to a destination.  Wave-structure interaction is a common physical problem associated with offshore structures for utilising ocean resources, such as oil and gas and offshore wind/wave/tidal energy.  It has been studied by several generations of hydrodynamic researchers, who have proposed different approaches, such as linear, weak nonlinear and fully nonlinear approaches.  However, there is still a long way to go to fully meet the demands of operators, developers, and designers.  They demand accurate modelling of the nonlinear WSI in more efficient way, considering complex physics such as breaking waves and turbulence.  The state of art is that the CFD modelling tools can handle the complex physics at least in theory, but they are too computationally expensive and thus are limited to modelling some simplified and deterministic cases at the speed developers expect.  The question is if the machine learning could accelerate the computing with sufficient accuracy, and thus help us deal with the complex WSI problems in a desired efficient way.  This presentation will overview the efforts made by the team of the presenters, who have proposed Hybrid PE-ML (hybrid physics-equation-based method with ML algorithm) method and applied it to dealing with various nonlinear WSI problems.  The presentation will summarise several interesting findings, such as ML algorithm trained on simpler cases can be employed on more complex WSI cases.

Title: “Isambard-AI: a new national AI supercomputer resource for the UK”

Speaker: Simon McIntosh-Smith

Professor Simon McIntosh-Smith is the founder and Director of the Bristol Centre for Supercomputing, which runs the UK’s Isambard-AI service. He began his career in industry as a microprocessor architect, first at Inmos and STMicro in the 1990s, before co-designing the world’s first fully programmable GPU at Pixelfusion in 1999. In 2002 he co-founded ClearSpeed Technology where, as Director of Architecture and Applications, he co-developed the first modern many-core HPC accelerators. He previously founded the HPC Research Group in Bristol, where his research interests include advanced computer architectures and performance portability. 

Workshop Materials Available

Selected presentation slides and workshop notes from the workshop are available for free access on the CCP-WSI Repository.

Register

Registration for this event has closed.