Efficient capture of the dominant periodic orbits underlying turbulent fluid flow.

点击翻译按钮获取中文摘要

We are familiar with turbulence, through its affect on the stability of aircraft during flight. Fluids, in this case air, are generally regarded as exhibiting two states of flow - a 'laminar' state and a 'turbulent' state. Turbulence is characterised by chaotic variations in the direction of the flow, through the appearance of whirls or 'eddies'. In industrial applications, turbulence typically leads to a loss of performance, as significant energy can be lost to the generation of eddies. A typical example is in pipelines, important for domestic water supply, irrigation, cooling systems, oil and gas supply. Rather than energy being expended in moving fluid directly from A to B, almost all the energy is lost to the creation and sustenance of turbulence! The question of how to model turbulence, therefore, is consistently listed among the most important outstanding problems of applied mathematics and theoretical physics (e.g. http://en.wikipedia.org/wiki/List_of_unsolved_problems_in_physics). This work builds on recent progress in understanding turbulence, made possible by the recent discovery of solutions to the equations governing flow in pipes and channels. These solutions are in the form of waves. Although they travel with the flow, their structure is otherwise static in time. Turbulence is chaotic in time, however. A radical step-change in this approach will be to model turbulence in terms of solutions that vary in time and that repeat after a period of time. Substantially new computational methods will be required to isolate such solutions in the future. There is strong motivation for isolating repeating cycles, otherwise called periodic orbits - from dynamical systems theory they are known to efficiently capture complex dynamics, filtering out activity that is otherwise a distraction. Often only a handful of periodic orbits are required to reproduce the statistical properties of a seemingly complex system.By extracting periodic orbits directly from simulations of turbulence itself, this project aims to capture those periodic orbits that are dynamically most important. So far it has only been possible to find orbits via numerical continuation methods, where there is no clear link between the orbits and the actual dynamics of the system. Capturing periodic cycles in a 'large' system such as turbulence, however, has been a challenging task. In this work, a new symmetry projection method will be developed to enable meaningful visualisations of the underlying dynamics. It has been shown that this particular method dramatically improves our ability to spot recurring cycles, i.e. periodic orbits. Collaboration with a leading European experimental facility will enable further application of these methods, plus theoretically guided searches to be performed more rapidly than is possible in simulation.This work will have great impact on our understanding of dynamical processes underlying turbulence, where periodic orbits will provide a basis for describing and predicting fluid flow patterns. This will open new avenues of future research in methods of prediction and control.