Modelling of the Unsteady Dynamics of Turbulent Disperse Bubbly Flows

气泡流存在于大量的工业过程中，包括油井、发电以及食品、制药和化学工业。通过这种方式，全世界每年生产价值数十亿美元的化学产品。因此，一个可靠的数值模拟工具在过程安全和过程控制方面的优势是巨大的。通常，工业上的两相流模拟是基于Reynolds平均Navier-Stokes方程（RANS）。然而，即使对于单相流，经过几十年的研究，也没有令人满意的湍流闭合模型。气泡流动的特点是界面在空间和时间上的变形。气泡动力学、湍流和气泡本身的多种相互作用是如此复杂，以至于在不久的将来不太可能得到一个通用的、准确的预测气泡流动的RANS模型。大涡模拟（LES）在基于RANS的闭包和基础方程的直接数值模拟（DNS）之间提供了一个很好的折衷。结合界面平流技术，LES的优势在于可以在很大程度上解决许多物理过程，这与在RANS中对整个长度和时间尺度进行建模形成了对比。在具有移动边界的两相流的情况下，LES方程包含额外的未知项。这些术语的闭包开发才刚刚开始，文献中还没有模型能够在实际的LES中产生令人满意的结果。本课题的目标是建立一个能够准确预测具有实际意义的分散气泡流的LES模型。该项目分为两个阶段。首先，将生成一个DNS数据库。为此，将在静止流体和湍流逆流配置中模拟垂直上升的单个气泡和小气泡群。随后，数据将被过滤以获得伪LES数据（先验分析）。通过这种方式，可以将模型思想与从DNS确定的确切的未封闭项进行比较。然而，模型的真正评估需要执行真正的LES。第一阶段中最有希望的模型将在LES代码中实现，并与DNS和研究伙伴提供的非常详细的实验数据进行比较。这种所谓的后验分析将指导气泡两相流的LES模型的资格。在本研究项目的延续中，计划使用多尺度方法来模拟大型气泡群。这将使我们能够达到全局目标，即模拟具有实际意义的配置。

Bubbly flows are found in a large number of industrial processes including oil wells, power generation as well as in the food, pharmaceutical and chemical industry. In this way chemical goods worth billions of dollars are produced worldwide every year. Therefore the advantages of the ability of a reliable numerical simulation tool in regards of process safety and process control are enormous.Typically, two phase flow simulations in industry are based on the Reynolds averaged Navier-Stokes equations (RANS). However even for single phase flow, after decades of research, no satisfactory turbulence closure model is available. Bubbly flows are characterized by spatially and temporally deforming interfaces. The manifold interaction of bubble dynamics, turbulence and the bubbles itself are so complex, that the availability of a universal and accurate RANS model for the prediction of bubbly flows appears unlikely in near future. Large eddy simulation (LES) offers a good compromise between RANS based closures and the direct numerical simulation (DNS) of the underlying equations. Combined with an interface advection technique LES offers the advantage, that many physical processes can be resolved to a large extent which is in contrast to modelling the whole range of length and time scales in RANS. In the context of two phase flows with moving boundaries the LES equations contain additional unknown terms. The development of closures for these terms has just started and no model has been demonstrated in the literature to yield satisfactory results in a real LES, yet. The goal of this project is to establish an LES model which is able to predict disperse bubbly flows of practical interest with good accuracy. The project is subdivided into two phases. First a DNS data base will be generated. To this end vertically rising single bubbles as well as small bubble swarms will be simulated in a quiescent fluid as well as in a turbulent counterflow configuration. Subsequently the data will be filtered to obtain pseudo LES data (a-priori analysis). In this way model ideas can be compared with the exact unclosed terms determined from DNS. The true evaluation of a model requires however to perform a real LES. The models from phase one which appear most promising will be implemented in the LES code and compared with DNS and very detailed experimental data available from a research partner. This so called a-posteriori analysis will guide the qualification of an LES model for bubbly two phase flows. In a continuation of this research project it is planned to use a multiscale approach in order to enable the simulation of large bubble swarms. This will allow to reach the global goal, which is the simulation of configurations of practical interest.