Hydrodynamic river modelling with D-Flow Flexible Mesh : case study of the side channel at Afferden and Deest
Hydrodynamic river modelling with D-Flow Flexible Mesh : case study of the side channel at Afferden and Deest
复制标题
使用 D-Flow 柔性网格进行水动力河流建模：Afferden 和 Deest 侧河道案例研究
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
E. Hagen
中科院分区：
文献类型：
--
作者：
E. Hagen
准确的水位预报在防洪安全管理中起着重要的作用。如今
为此目的使用多维数值流体动力学模型已经成为普遍的做法。
目前，WAQUA和Delft 3D是荷兰的标准工具，它们基于结构化的
曲线网格曲线网格可以跟踪大范围的地形变化，并使用类似的网格
在整个计算域中的分辨率。结构化曲线网格的缺点
方法是封闭边界的阶梯表示有时是不可避免的，因为网格
像元不与流向对齐，在蜿蜒河流的内弯处，网格线可能
变得集中于不必要的小网格单元。为了改善这些问题，Delaware正在开发
基于非结构网格的流体动力学模型Flexible Mesh（也称为“D-Flow-FM”）。的
非结构化网格方法使用户能够使用空间可变网格分辨率。通过组合
曲线网格单元与三角形网格单元，建模器可以增加网格分辨率的位置
由于局部地形的变化，这是最需要的。在本研究中，对柔性补片进行了测试，
与基于WAQUA的结构化网格和Flexible的非结构化网格相比，
网格应用于Deest的Afferden的侧通道，其中WAQUA网格被认为是
不准确。本研究的主要目的是：
通过以下方式评估柔性网的性能（水位、流速和排放量）：
与WAQUA进行比较，并评估模拟结果对网格分辨率的敏感性，
柔性网格。
在研究的第一步中，将Flexible Mesh模型与校准的WAQUA模型进行比较，
重点关注水位、流量和流速。柔性网格模型中的水位为
与WAQUA模型中的水位结果相当。对于低放电，几乎没有
水位差和高排放的水位约12厘米高，
柔性网格模型。河漫滩和侧槽中的流量在
柔性网格模型。WAQUA和Flexible之间的差异有两个重要来源
网格首先，柔性网格默认使用不同的、校正的Colebrook-White粗糙度公式
这导致柔性网中的更大摩擦和更高的水位。第二，能源损失，
Flexible Mesh中对堰上的水流进行了不同的建模，这导致Flexible Mesh中的水位较高
在洪泛区和在阿费尔登和迪斯特的侧槽中降低流量。
在第二步中，在阿费尔登和迪斯特对瓦尔河的主河道进行了局部网格细化
和侧通道。瓦尔河主河道的网格细化没有显示出明显的效果
在连续的网格细化之间。局部网格细化也应用于侧通道，
其中假定原始网格不准确地示意了侧信道。所述差与所述
参考网格对于具有最大细化侧信道的示意图是最大的。不过效果
在较高的网格分辨率下，网格细化的减少，这表明模型结果的收敛。
经过四次网格的淬炼，效果已经几乎不受网格淬炼的影响。
因此，收敛似乎是围绕四次细化的边通道达成的。的
由于网格细化，计算时间增加。对于高网格分辨率，时间步长必须为
降低，以满足模型条件稳定性（默认Courant数<0，7）。网格细化是
当模型结果尚未收敛时有效，因此进一步细化对模型仍有影响
结果，计算时间仍然是可以接受的。
本研究结果显示了应用非结构化网格进行局部网格加密的潜力
适用于复杂几何形状的D-Flow Flexible Mesh。侧流计算的精度
通道似乎通过局部网格细化得到改善。然而，需要进一步的研究来评估
柔性网格的准确性。
Accurate predictions of water levels play an important role in the management of flood safety. Nowadays, it
has become common practice to use multi-dimensional numerical hydrodynamic models for such purposes.
Currently, WAQUA and Delft3D are standard tools in the Netherlands, which are based on a structured
curvilinear grid. The curvilinear grid can follow large-scale topographical changes and uses similar grid
resolution throughout the entire computational domain. Drawbacks of the structured curvilinear grid
approach are that staircase representation of closed boundaries is sometimes unavoidable, because grid
cells are not aligned with the flow direction and in the inner bends of meandering rivers, gridlines may
become focussed to unnecessarily small grid cells. To improve on these issues, Deltares is developing the
unstructured-grid-based hydrodynamic model Flexible Mesh (also referred to as “D-Flow-FM”). The
unstructured grid approach enables the user to use a spatially variable grid resolution. By combining
curvilinear grid cells with triangular grid cells, the modeller can increase grid resolution on the locations
where, because of local topographical variations, it is most desired. In this study Flexible Mesh is tested and
compared with the structured grid based WAQUA and the possibilities of the unstructured mesh of Flexible
Mesh are applied on a side channel at Afferden at Deest, where the WAQUA grid is considered to be
inaccurate. The main objective of this research is:
Evaluate the performance (water levels, flow velocities and discharges) of Flexible Mesh by
comparing with WAQUA and assess the sensitivity of the modelling results for the grid resolution in
Flexible Mesh.
In the first step of the study the Flexible Mesh model is compared to the calibrated WAQUA model with
focus on the water levels, discharges and flow velocities. The water levels in the Flexible Mesh model are
comparable to the results of the water levels in the WAQUA model. For low discharges there is almost no
difference in the water level and for high discharges the water levels are about 12 centimeters higher in the
Flexible Mesh model. The discharges over the floodplains and in the side channel are much smaller in the
Flexible Mesh model. There are two important sources for the differences between WAQUA and Flexible
Mesh. First, Flexible Mesh default uses a different, corrected formula for the Colebrook-White roughness
which results in a larger friction in Flexible Mesh and higher water levels. Second, the energy losses due to
flow over weirs is modelled different in Flexible Mesh, which results in higher water levels in Flexible Mesh
and lower discharges over the floodplain and in the side channel at Afferden and Deest.
In the second step local grid refinement was applied at Afferden and Deest to the main channel of the Waal
and to the side channel. The grid refinement of the main channel of the Waal showed no clear effects
between consecutive grid refinements. The local grid refinement was also applied for the side channel,
where the original grid is assumed to schematize the side channel inaccurate. The difference with the
reference grid is maximal for the schematization with the largest refined side channel. However, the effect
of grid refinement decreased at higher grid resolutions which indicates convergence of the model results.
After the grid was refined four times, the results were hardly affected by a grid refinement anymore.
Therefore, convergence seems to be reached around the four times refined side channel. The
computational time increases because of grid refinement. For high grid resolutions, the time step has to be
decreased in order to meet the model condition stability (default Courant number < 0,7). Grid refinement is
efficient when model results are not yet converged, so further refinement has still effect on the model
results, and computational time is still acceptable.
The results of this study show potential for application of local grid refinements with the unstructured grid
of D-Flow Flexible Mesh for complex geometries. The accuracy of the computation of the flow in the side
channel seems to be improved by the local grid refinement. However, further research is required to assess
the accuracy of Flexible Mesh.