Terascale Direct Numerical Simulations of Turbulence Generated by Fractal Grids

湍流在环境、气象和海洋学以及许多涉及流体流动的工业中，如在化学、混合、汽车、航空航天和海军工业中，都具有极其重要的意义。举个例子，通过管道抽油的成本与湍流引起的摩擦损失成正比。作为另一个例子，全球海洋运输每年消耗约21亿桶石油。了解湍流可以导致流动控制方案，以减少表面摩擦阻力。以每桶75美元的价格（现在是一个非常保守的估计），减少10%每年将节省150亿美元，更不用说随之而来的船舶排放污染物的显著减少了。对每年消耗20亿桶喷气燃料的航空业也可以做出类似的预测。在过去的60年左右，对湍流的研究主要集中在特定气流的特别建模上，进展有限。理解湍流动力学是必要的，为此需要一个特殊的实验，在这个实验中，这些湍流动力学可以被分开，以便给我们理解它们的线索。这就是最近在伦敦帝国理工学院取得的成就，那里首次在风洞中进行了由分形网格产生的湍流实验，并取得了令人印象深刻的结果。现在有必要对这些流动进行计算机模拟。这些都是非常大的模拟，特别是因为分形湍流发生器的多尺度性，需要最先进的高端并行计算。这是一个计算上的挑战。迎接这一挑战为我们提供了一个前所未有的机会来理解湍流以及如何操纵它。

Turbulence is of enormous importance in the enviroment, meteorologyand oceanography and also in many industries where fluid flow isinvolved such as in the chemical, mixing, car, aerospace and navalindustries. Just as an example, the cost of pumping oil throughpipelines is directly proportional to the frictional losses whichturbulence causes. As another example, worldwide ocean shippingconsumes about 2.1 billion barrels of oil per year. Understandingturbulence can lead to flow control schemes for reducing skin frictiondrag. At 75 US dollars per barrel (nowadays a very conservativeestimate), 10% such reduction would save 15 billion US dollars peryear, not to mention the consequent impressive reductions ofpollutants in ship emissions. Similar projections can be made forairline industries which consume about 2 billion barrels of jet fuelper year.Over the past 60 years or so the efforts in turbulence have beenmostly in ad hoc modelling of specific flows and the progress has beenlimited. An understanding of turbulence dynamics is needed, and forthis a particular experiment is needed where these turbulence dynamicscan be set out of joint so as to give us clues for understanding them.This is what has been achieved recently at Imperial College Londonwhere the first ever experiments of turbulence generated by fractalgrids in wind tunnels have been conducted with impressive results. Itis now necessary to develop computer simulations of these flows. Theseare very large simulations, in prticular because of the multiscalenature of the fractal turbulence generators, and require state-of-theart top-end parralel computing. The challenge is computational.Meeting this challenge presents us with an unprecedented opportunityto understand turbulence and how to manipulate it.