You are on page 1of 17

Journal of Hydrology 590 (2020) 125267

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Experimental study and numerical verification of silted-up dam break T


a a,⁎ a b
Foad Vosoughi , Gholamreza Rakhshandehroo , Mohammad Reza Nikoo , Mojtaba Sadegh
a
Department of Civil and Environmental Engineering, Shiraz University, Shiraz, Iran
b
Department of Civil Engineering, Boise State University, Boise, ID, USA

A R T I C LE I N FO A B S T R A C T

This manuscript was handled by Corrado In this study, multiphase flood waves caused by failure of silted-up dams have been scrutinized experimentally,
Corradini, Editor-in-Chief, with the assistance and efficacy of the public-domain CFD software OpenFOAM numerical models to represent such phenomenon is
of Subashisa Dutta, Associate Editor evaluated. Experimental observations are sparse on dam break for reservoirs with high sediment depths, called
Keywords: silted-up reservoirs. Sediment layer, in this case, may behave as a viscous fluid and produce a complex multi-
Dam break layer flow of air and water overtopping a sediment layer. Massive sediment quantities can have a substantial
Silted-up reservoirs impact on the wave propagation pattern caused by dam break. This study contributes to an experiment-based
Multiphase flood wave understanding of dam break phenomenon with silted-up reservoir, and elaborates the propagation of multi-layer
Image processing
waves. Herein, different hydraulic conditions were experimentally observed including 8 different upstream
OpenFOAM
sediment depths in the range of 0–0.24 m and 4 downstream initial water levels of 0, 0.02, 0.04 and 0.05 m,
which collectively created 32 scenarios. The resulted waves were filmed by high-speed cameras and data were
extracted through image processing. An open-source CFD software, namely OpenFOAM, was employed to si-
mulate the experiments using two distinct numerical approaches: volume of fluid (VOF) and Eulerian. Results of
both numerical approaches were in good agreement with the experimental measurements; with MAE and RMSE
error values varying from 0.006 m to 0.031 m and 0.008 m to 0.043 m, respectively (2% to 10% and 2.6% to
14% with respect to the reservoir height of 0.3 m). The Eulerian approach showed marginally superior per-
formance to VOF, especially in cases with high sediment depths, which is likely due to accommodating phase
mixing. A wide collection of high-quality data is made available online in the Appendix Data and could be
utilized in future studies.

1. Introduction depth play important roles in both. But in fact, there is a large differ-
ence between the sediment depth in a silted-up reservoir and typical
Reservoir sedimentation can cause challenging problems such as sediment level in a river downstream of the dam (Duarte et al., 2011).
reducing the effective volume and triggering flood overpassing (Yang, Particularly, the authors are not aware of any numerical model ver-
1996). It may also cause failure of coffer dams built to trap sediments ifying the adequacy of the public-domain CFD software OpenFOAM on
upstream of the main dam reservoir (Vischer and Hager, 1998). Massive dam break multi-layer waves for silted-up reservoirs, which motivated
amount of sediment stored in dam reservoirs has a substantial impact this work. Accordingly, it is appropriate to investigate the dam break
on the dam break flood propagation pattern and also can bury build- issue with silted-up reservoirs, and elaborate how multi-layer waves
ings, installations and infrastructures at the downstream of the dam. In propagate therein, experimentally and verifying numerically using
such phenomenon, the sediment layer may behave like a viscous fluid, OpenFOAM CFD software (Open∇FOAM, 2015).
with clear water flowing easily over its top which produces a complex The oldest experimental works on dam break and associated flood
three-layer wave formed by air, a layer of water overtopping a satu- wave dating back to the late 20th century (Dressler, 1954; Bell et al.,
rated sediment layer (Duarte et al., 2011). 1992; Bellos et al., 1992; Fraccarollo and Toro, 1995; Lauber and
Most early studies as well as current work focus on dam break for Hager, 1998). The majority of prior experimental researches have
water filled reservoirs and this problem for silted-up reservoir is spar- evaluated this issue for a fixed channel bed (without sediment), in
sely addressed in the literature (Duarte et al., 2011). At the first glance, which different initial conditions such as dry-bed or wet-bed down-
many dam break scenarios with mobile bed may seem similar to silted- stream have been addressed (Soares Frazão and Zech, 2002; Bellos,
up reservoirs because the parameters such as water level and sediment 2004; Meile et al., 2013; Wood and Wang, 2015; Xue et al., 2011;


Corresponding author.
E-mail address: rakhshan@shirazu.ac.ir (G. Rakhshandehroo).

https://doi.org/10.1016/j.jhydrol.2020.125267
Received 25 March 2020; Received in revised form 30 June 2020; Accepted 1 July 2020
Available online 07 July 2020
0022-1694/ © 2020 Elsevier B.V. All rights reserved.
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Fig. 1. Flowchart of the methodology utilized for experimental study and numerical verification of silted-up dam break.

Kocaman and Ozmen-Cagatay, 2012; LaRocque et al., 2013; Ozmen- MacIel, 2012). In reality the multiphase flood resulted from dam break
Cagatay et al., 2014). Recently, dam break laboratory studies con- with silted-up reservoir may behave as a non-Newtonian flow, espe-
cerning mobile bed scenarios have gained more attention to assess re- cially at small shear stresses. There are different non-Newtonian models
levant physical parameters such as flow depth and wave front celerity in OpenFOAM in which various solvers such as SimpleFoam, non-
(Leal et al., 2001; Spinewine and Zech, 2007; Cochard and Ancey, 2008; NewtonianIcoFoam and PISOFoam are developed to simulate non-
Leal et al., 2009; Lal and Moustafa, 2016). To collect the required data Newtonian fluids (Open∇FOAM, 2015; Unofficial OpenFOAM, 2019).
in these studies, digital image processing has been used (Leal et al., However, since in the numerical part of this study, sudden dam break
2001; Spinewine and Zech, 2007). Although, studies of dam break are was modeled and partial gate opening was not considered, the applied
well documented, to the authors’ knowledge, just one study is available models and solvers are not capable of non-Newtonian simulation.
in the literature that experimentally examined dam break with silted-up Hence, it is one of the modeling assumptions - although imperfect – in
reservoirs (Duarte et al., 2011). In particular, that study has considered this study that the multiphase flood wave behaves Newtonian.
only dry tail water portion as the downstream condition (Duarte et al., A considerable body of literature on dam break phenomena are
2011). based on numerical modeling, without recourse to the main differences
Additionally, several experimental works have been conducted in between sudden failure of dams with highly silted-up reservoirs and
the literature to examine dam break phenomena with various kinds of dam break with usual mobile bed condition. In the late 20th century,
fluid in their reservoir (Nsom, 2002; Ancey et al., 2009; Minussi and De there has been numerous studies to investigate 1D and 2D dam break
Freitas MacIel, 2012); some focusing on Newtonian fluids with high (Gabutti, 1983; Beam and Warming, 1976; Chaudhry, 2008; Fennema
viscosities and debris flows (Nsom, 2002; Ancey et al., 2009), others on and Chaudhry, 1990; Mingham and Causon, 1998), on the basis of
similar scenarios, but for Non-Newtonian fluids (Minussi and De Freitas computational fluid dynamics (CFDs), in which specific numerical

2
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

approaches have been widely used, such as finite difference method flume consists of a smooth metallic bottom, and the walls are made of 9-
(Gabutti, 1983; Beam and Warming, 1976) and finite volume method mm-thick glass. In order to represent a reservoir, a 1.52 m long section
(Fraccarollo and Toro, 1995; Mingham and Causon, 1998). Usually of the flume was isolated by a gate (dam) to split the upstream and
simple scenarios have been the focus of researches, where straight and downstream regions of the flume, while the downstream flume reach
bending channels (Soares Frazão and Zech, 2002; Valiani et al., 2002; was 4.5 m long (Fig. 2). Initial experiments were carried out to carefully
Zoppou and Roberts, 2003; Ackerman et al., 2008) have been con- evaluate hydraulic parameters, initial conditions and required facilities.
sidered for a constant bed. Later dam break studies have emphasized on Then, by preparing, constructing and installing detailed parts to the
more complicated scenarios such as wet-bed downstream conditions flume, it was equipped to perform main dam break experiments. Dam
(Zoppou and Roberts, 2003; Cao et al., 2004; Leal et al., 2006; Wu and break was physically emulated by rapidly raising the gate downstream
Wang, 2007; Crespo et al., 2008; Zoppou and Roberts, 2000; Aliparast, of the reservoir. We followed the literature guidelines that gate opening
2009) and the role of sediment transport and vegetation (Evangelista time should be shorter than 1.25( h 0 / g ) where h 0 is the initial reservoir
et al., 2013; Postacchini et al., 2014; He et al., 2017; Kim and Sanders, water level, and g is the gravitational acceleration (Lauber and Hager,
2016; Fu and Jin, 2016) have also gained more attention. Recently, 3D 1998). Noted that, in all experiments, h 0= 0.3 m, and hence, maximum
numerical studies have been conducted to simulate such phenomenon gate opening time is 0.22 s. The gate was removed in a much shorter
(Biscarini et al., 2010; Yang et al., 2010; Hu et al., 2012; Marsooli and duration (0.08–0.16 s) as documented by high-speed imaging, and
Wu, 2015), some focusing on mobile bed condition (Marsooli and Wu, hence, this process may be considered as a sudden dam break.
2015). However, the efficacy of numerical models to represent the dam Two- and three-phase flows of water and saturated sediment layer
break multi-layer waves for silted-up reservoirs to date has not been caused by dam break phenomena were filmed by three professional and
directly evaluated. high-speed digital cameras (Canon EOS 70D). Cameras were mounted
The present study is focused on two- and three-phase flood waves on the right-hand side of the flume, without any position change in
caused by sudden dam break with silted-up reservoirs. In order to ad- more than 6 months, while main experiments were conducted.
dress the literature gaps and shortcomings outlined above, the authors Recorded videos were 50 frames per second with 1920 × 1080 pixels
have scrutinized this topic experimentally for 32 different cases and (Full HD/1080p) resolution. As shown in Fig. 2, cameras covered
modeled them numerically using the OpenFOAM software. More spe- 5.52 m along the flume, which consisted of 1.52 m upstream to 4 m
cifically, different initial hydraulic conditions were considered in- downstream of the gate location (dam). Moreover, two powerful spot-
cluding eight distinct depths for saturated sediment layer upstream of lights were mounted above the flume as light sources at both ends of the
the dam which occupied 10%–80% of the reservoir’s initial water flume. The opposite lateral wall of the flume was covered with black
depth. Four downstream boundary conditions were examined: zero plastic sheets to conceal installations in the back of the flume. Lab
(dry) and three different non-zero (wet) initial water depths of 0.02, windows and other uncontrolled light sources were covered using thick
0.04 and 0.05 m, which collectively created 32 scenarios. black curtains to eliminate light pollution and obtain adequate contrast.
Results of this study can enhance the technical understanding of Different hydraulic conditions were examined (Table 1) including
how multi-layer waves resulted from dam break phenomena with various depths of saturated sediment layer in the reservoir and different
silted-up reservoirs propagate therein, and elaborate how the saturated downstream initial water levels (dry-bed or wet-bed downstream con-
sediment cone begins to move and develops or deforms as time elapses. ditions with different water levels), which collectively created 32 dam
This study also provides a wide collection of high-quality laboratory break scenarios. As mentioned earlier, the purpose of this study was to
data that can be utilized for validation of other numerical models. A scrutinize the multiphase shock waves caused by failure of silted-up
free-licensed CFD package, namely OpenFOAM (Open∇FOAM, 2015), dams. It also addresses the failure of coffer dams, secondary dams and
was employed to simulate the experiments using two distinct numerical gravel or front basins built upstream of a main dam to protect its re-
approaches: volume of fluid (VOF) and Eulerian. Then, the experi- servoir sedimentation by trapping sediments (Vischer and Hager,
mental measurements and numerical results are comprehensively 1998).
compared and their correlation is discussed. As shown in Table 1, a variety of initial upstream sediment depths
were considered to cover 10%–80% of the reservoir’s initial water
2. Experimental procedure depth. These experimental scenarios were designed to cover different
elevation zones of the dam reservoir such as inactive zone (dead sto-
Fig. 1, represents flowchart of the methodology utilized in devel- rage), buffer and active zone (conservative) and flood control zone
oping experimental study and numerical verification of silted-up dam (Lamond and Boukhtouta, 2002). Initial downstream water levels of
break. As reflected in the figure, the methodology contains 5 main steps 0.02, 0.04 and 0.05 m, as well as dry downstream conditions were also
which will be discussed in detail in the rest of the research paper. considered. This setup was specifically designed to approximately ac-
All main experiments were carried out in the Hydraulic Lab of Civil quire logical downstream river level conditions and to evaluate and
and Environmental Engineering Department at Shiraz University, cover different states as much as possible (Vischer and Hager, 1998;
Shiraz, Iran, using a 6-m-long horizontal laboratory flume with pris- Lamond and Boukhtouta, 2002). Authors believe these initial condi-
matic rectangular section; 0.3 m wide and 0.32 m high (Fig. 2). The tions are sufficient to mimic practical applications. Percentage depth of

Fig. 2. Top view of the experimental set-up.

3
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Table 1
Scenarios formed by different upstream and downstream hydraulic conditions examined in this study.

saturated sediment stored in the reservoir was defined in Table 1 images have been enhanced in brightness, sharpness, contrast, color,
asSd/ h 0 (%) = (Sd/ h 0) × 100 , where Sd and h 0 are initial upstream sedi- and overlap.
ment and water levels, respectively. The waves caused by a dam break in silted-up reservoirs produce a
Fig. 3 (a), shows wave propagation images at different times after very complex phenomenon with a three-phase behavior, as shown in
dam break for the case with 0.075 m depth silted-up reservoir and Fig. 3 (b), including a layer of water on a saturated sediment layer and
0.02 m standing water downstream. As expected, the wave stretches the third phase being air. Additionally, once the dam breaks, the sedi-
and dampens as it moves downstream. It is worth noting that to gain ment layer may behave like a viscous fluid, with clear water flowing
higher quality and accuracy in determining free surface profiles, all easily on its top.

Fig. 3. Multiphase dam break wave propagation, (a) experimental images for different times after dam break, with 0.075 m depth silted-up reservoir and 0.02 m
standing water downstream, (b) schematic side view of a typical dam break experiment with silted-up reservoir and dry downstream.

4
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

To collect the required data, several 1-mm-resolution point-gages in available experimental and numerical studies reported in literature si-
stripe forms have been installed on the flume outside wall. These point- milar results in the (LaRocque et al., 2013; Ozmen-Cagatay and
gages located in overlap sections covered by two closed cameras were Kocaman, 2010). Fig. 5, presents a set of image comparisons at different
used as reference to match the images. For the main tests with silted-up times after dam break to visually compare the results of the present
reservoirs, a certain type of sediment was selected as mixtures of quartz study against other researchers’ results (Ozmen-Cagatay and Kocaman,
sand with a uniform grain size between 0.2 and 0.4 mm; the sediment 2010). A good agreement with literature was found in terms of water
gradation curve is presented in Fig S1, in Supplementary Material. In depths and wave front locations as illustrated in Fig. 5 and also Figs. S2
order to ensure uniformity and homogeneity of sedimentation, a stan- and S3, in Supplementary Material.
dard test preparation procedure was followed carefully (Duarte et al.,
2011). After washing and drying the sediment, it was transferred slowly
into the dry flume. Therefore, the smooth and horizontal sediment layer 3. Numerical modeling
surface was manually created by a putty knife. In order to minimize
sediment diameter separation, water was injected to the flume at a low Along with several commercial CFD software, such as Fluent and
discharge while slow sediment deposition took place simultaneously. Flow3D to solve fluid mechanics problems, free-licensed software also
gained popularity in the field; one of the most widely used being an
open-source code software named OpenFOAM (Open∇FOAM, 2015).
2.1. Data collection This software allows users to simulate a wide range of physical phe-
nomena such as two-phase and multiphase flows (Open∇FOAM, 2015;
For each experiment, three 40 s video files in MOV format were Unofficial OpenFOAM, 2019). The numerical part of this study is lim-
recorded using all cameras while each camera covered about 2 m of the ited to simulation of all experimental data collected from 32 different
flume length (Fig. 2). Relevant parameters of water level and sediment experiment scenarios using the OpenFOAM software. This software is
depth were extracted, using an automated image processing method. designed for installation on a Linux operating system and its pro-
Main advantage of digital image processing is that this method is non- gramming language is C++; an object-oriented and easily expandable
subjective and can yield accurate and high-quality results. Additionally, programming language. Simulations have been performed using both
measurement points can be easily replaced in this method and the volume of fluid (VOF) and Eulerian approaches. Although these ap-
measurements of water level and sediment depth variations can be proaches have been used previously to solve clear water dam break
estimated at any arbitrary location along the flume. For this purpose, problems (Ozmen-Cagatay et al., 2014; Biscarini et al., 2010; Issakhov
the spatially surveyed spots, including 16 distinct points have con- and Imanberdiyeva, 2019; Mokrani and Abadie, 2016; Panicker et al.,
sidered covering the entire length of the flume (Fig. 4). Gaps between 2015; Fondelli et al., 2015; Ye and Zhao, 2017; Issakhov et al., 2018), to
the points near the gate location were less than other areas along the the authors’ knowledge, they have never been used in multiphase dam
flume, to ensure enough measurements are performed in this area of break with silted-up reservoirs modeling. It should be noted that, no
high turbulence and rapid depth change. The first point was at the reduction or enlargement in width of channel downstream was con-
beginning of the reservoir (upstream of the flume) and the distances of sidered, and the longitudinal slope of channel was taken as zero.
15 other surveyed points were 0.76, 1.02, 1.27, 1.37, 1.42, 1.47, 1.52,
1.57, 1.67, 1.77, 1.87, 2.52, 3.52, 4.52 and 5.52 m downstream of the
first point, respectively. These data were extracted by the analysis of the 3.1. VOF method
first 300 frames, covering 6 s (50 frames per second) of each video file
using 15 time snaps after the dam breaks. Intervals between the snaps The VOF method is used to simulate layered flows in which the
were shorter immediately after the dam break and increased gradually. phases have explicit boundaries (clear interfaces) such as free surface
The snap times were 0.04, 0.08, 0.12, 0.2, 0.3, 0.4, 0.6, 0.8, 1, 1.5, 2, 3, flows. In order to simulate two-phase and multiphase dam break waves
4, 5 and 6 s after the dam break. with OpenFOAM, interFoam and multiPhaseInterFoam solvers have been
Given 3 cameras for shooting, 15 time snaps and 32 different dam used. In this approach, a set of conservative forms of continuity and
break scenarios, a total of 1440 photographs extracted, which were all momentum equations are simultaneously solved for all available
examined carefully. According to the 16 fixed locations and two para- phases. Computations of the water free surface by VOF also involved a
meters, these photographs contained 15,360 data (32 scenarios × 15 CFD package based on the Finite Volume Method (FVM). Continuity
times × 16 locations × 2 parameters) which are all presented in the and momentum equations for a Newtonian incompressible fluid in VOF
Appendix Data. method may be described as Eqs. (1) and (2) (Jamshidi et al., 2019;
To verify the experimental measurements and OpenFOAM results, Kassar et al., 2018; Barbosa et al., 2019; Hirt and Nichols, 1981;
particular physical models with clear water upstream and dry-bed or Rusche, 2002):
wet-bed downstream conditions were designed and modeled both ex- →
∇· U = 0 (1)
perimentally and numerically. Then their results were compared with

Fig. 4. Schematic view of the laboratory flume; L1 to L16 are depth measuring points.

5
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Fig. 5. Visual comparison of free surface profiles h(x) at different times: (a) available experimental results (Ozmen-Cagatay and Kocaman, 2010), (b) Fluent
predictions (Ozmen-Cagatay and Kocaman, 2010), (c) experimental results, and (d) OpenFOAM predictions. Lengths in cm, and β: the depth ratio of downstream
water level to reservoir initial height.

→ and 1 (Deising et al., 2016):


∂ (ρU ) →→
+ ∇ ·(ρU U ) = −∇ ·P + ∇ ·τ + ρ ·g + Fσ
∂t (2)
⎧1 filled by fluid 1
→ α (x. y. z. t) = 0 < α < 1 At the interface
where U represents the velocity field; P is pressure; t is time; τ is vis- ⎨
cous stress tensor; ρ is density; g is gravitational acceleration constant, ⎩0 filled by fluid 2 (6)
and Fσ is the momentum associated with surface tension. The viscous
To determine α as a function of time and estimate the location of
stress tensor τ is given by
interphase between the two fluids, the volume fraction equation for -
→ → α has to be solved (Hirt and Nichols, 1981):
τ = μ [(∇U ) + (∇U )T ] (3)
∂α →
where μ is the fluid dynamic viscosity (Jamshidi et al., 2019; Kassar + ∇ ·(U α ) = 0
∂t (7)
et al., 2018). The last term in momentum Eq. (2), the surface tension Fσ ,
is modeled as Continuum Surface Force (CSF) and is expressed as More theoretical details of the OpenFOAM and VOF method are
Jamshidi et al. (2019), Kassar et al. (2018), Heyns and Oxtoby (2014), presented as Eqs. (S1)–(S6) in Supplementary Material and also can be
Brackbill et al. (1992): found in Jamshidi et al. (2019), Kassar et al. (2018), Barbosa et al.
Fσ = σK ∇α (4) (2019), Hirt and Nichols (1981), Rusche (2002), Heyns and Oxtoby
(2014), Brackbill et al. (1992), Deising et al. (2016).
where σ is the surface tension constant, K is the curvature of the sur-
face, and α represents volume fraction of the fluid. The curvature K 3.2. Eulerian method
may be approximated by Heyns and Oxtoby (2014), Brackbill et al.
(1992): Eulerian method is used to solve multiphase flows while phases may
∇α1 ⎞ mix together. In this method, which is also called Two-Fluid Method
K = −∇ ⎛ ⎜ ⎟
(TFM), each phase is individually calculated and solved, making the
⎝ α1 | ⎠
|∇ (5)
computing time longer than VOF. The most widely used solvers in this
In more details, the volume fraction α (alpha in the software code) method are twoPhaseEulerFoam and multiPhaseEulerFoam, with the
is used as an index function to determine the relative proportion of the theoretical details being described in Soleimani et al. (2015), Bhusare
cell filled with fluid. α is 1 for a place filled by the first fluid with the et al. (2017), Lubchenko et al. (2018), Kia and Aminian (2017), Panda
density ρ1 and α is 0 for a place filled by the second fluid with the et al. (2017)). To verify the OpenFOAM numerical results; particular
density ρ2 . At the interphase between the two fluids α varies between 0 clear water dam break scenarios were modeled physically and

6
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

numerically under dry-bed and wet-bed downstream conditions. Re-


sults were compared (Fig. 5) with available laboratory and numerical
results in the literature (LaRocque et al., 2013; Ozmen-Cagatay and
Kocaman, 2010) and a good agreement with literature was found. Also,
as shown in Figs. S4 and S5, in the Supplementary Material, the nu-
merical results are in good agreement with measured data in the lit-
erature.

3.3. Initial conditions and solution domain

Owing to the laboratory channel restrictions, the solution domain


was 6 m long, 0.3 m wide and 0.32 m high. As the initial condition, a
volume of 1.52 m in length and 0.3 m in height was considered as the
reservoir. Channel bottom and side walls were made of steel and glass,
thus assumed smooth. As the boundary condition, the upstream,
bottom, and side boundaries along the channel were all defined as
walls, and the downstream boundary at the end of the channel, was
defined as a pressure outlet. The top boundary was defined as a pressure
inlet, due to atmospheric pressure on the free surface. The boundary
hydraulic conditions were also independently considered (meshed),
creating different downstream and upstream conditions. Downstream
conditions were completely dry-bed or wet-bed with different initial
water depths, and upstream conditions were reservoirs with various
saturated sediment layer heights.
A standard procedure was followed carefully to determine suitable
mesh size and time step (Akbari and Barati, 2012; Barati et al., 2012).
Mesh size analysis was performed using eight different sizes. Then, by
compromising the error and runtime, a uniform mesh of fixed cubes
with dx = dy = dz = 0.005 m was selected. In total, the three di-
mensional computational domain contained more than 4.6 million
cells. Time step analysis was performed using four different time steps Fig. 6. Results of simulation by VOF and Eulerian numerical approaches for
and considering the error, courant number, and the runtime, a time step 0.1, 0.2, 0.4 and 1 s after dam break.
of 0.001 s was selected. More details of the mesh size and time step
analyses are presented in Figs. S6–S8 in Supplementary Material. channel length from upstream to downstream of the dam. Six snap
times of 0.2, 0.3, 0.4, 0.6, 0.8 and 1 s after dam break were scrutinized.
4. Results and discussion In Fig. 7, water level and sediment depth profiles are shown after
the dam breaks. The simulation results by VOF and Eulerian methods
In this section, several graphs and tables are presented and dis- are located below experimental images for a span of 1.3–3.1 m from the
cussed to study the effects of various factors including dry-bed or wet- beginning of the flume; from 0.2 m upstream to 1.6 m downstream of
bed downstream and different depths of upstream saturated sediment the dam. This figure is for the case where upstream sediment depth is
layer on multiphase flood waves caused by the silted-up dam break. 0.03 m and dam downstream is wet with 0.04 m of standing water. By
Fig. 6, shows the comparison between simulation results using VOF and comparing the experimental and numerical results, it is observed that
Eulerian approaches at distinct times of 0.1, 0.2, 0.4 and 1 s after the the downstream wave propagations are similar in predicting the wave
dam breaks. Images are extracted and categorized for an approximately tip locations. Hence there is an acceptable concordance in prediction of
2-m interval covered from 0.2 m upstream to 1.8 m downstream of the the wave front celerity. At 0.4 s, the downstream positive wave causes
dam. As shown, the resulted wave moves rapidly downstream; about an increase in downstream water elevation. Also, the progressive wave
0.18 m in 0.1 sec after the dam breaks and it reaches 1.5 m downstream has a specific profile with two bulges in the body and a front wave in its
of the dam in 1 s (3-m point in the figure). forehead region. The wave tip has reached 2.4 m in downstream and it
Once this wave hits the standing downstream water, it creates is well predicted by both numerical methods in terms of size, form and
special large bulges in the water level. As the time passes after dam location, however, Eulerian results seem to be more accurate in pre-
break, water level at the dam section as well as its upstream decreases dicting the details. At 0.6 s, turbulence is increased in the forehead
quickly. Results of the two numerical methods are slightly different. For region and the wave tip has reached 2.9 m from the reservoir beginning
example, at 0.4 and 1 s, the amount of sediment layer propagation in with two large bulges in its body. Also, at 2.2 m from the reservoir
VOF method is slightly higher than that in Eulerian. Also, the boundary beginning, the water level is about 0.13 m which predicted reasonably
between the phases is sharp and clear in VOF results, while the inter- by both numerical methods.
ference of phases has occurred at some regions in Eulerian. At 0.8 s, the wave has some bulges in frontal area and at an interval
of 2.1–3 m, the water level has increased and all of these fluctuations
4.1. Image comparisons for upstream sediment depths of 0.03, 0.15, and are well predicted by the software. Investigating later times after dam
0.22 m break indicates that the sediment layer has moved downstream up to
0.08 m without any significant change in its initial form of a relatively
Several image comparisons are depicted using photographs ex- sharp step. Comparisons of water level and sediment depth profiles,
tracted from the laboratory videos and simulation results of VOF and show that laboratory experiments and results of VOF and Eulerian
Eulerian methods. All images are presented on an evaluable basis and a methods are in good agreements. Note that the water level and sedi-
complete comparison is made to better visualize different upstream and ment depth profile images after dam break, while upstream sediment
downstream hydraulic conditions. The photographs associated with depth is 0.15 m and dam downstream is wet with 0.04 m water are
Camera No. 2 (mid camera), have utilized covering about 2 m of the depicted and discussed in Fig. S9, in Supplementary Material.

7
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Fig. 7. Visual comparison of laboratory modeling results (a, d, g, j, m and p), VOF (b, e, h, k, n and q) and Eulerian predictions (c, f, i, l, o and r) in determining water
level and sediment depth profiles, at different times after dam break, when upstream sediment depth is 0.03 m and dam downstream is wet with 0.04 m water.

Fig. 8. Visual comparison of laboratory modeling results (Figs. a, d, g, j, m and p), VOF (Figs. b, e, h, k, n and q) and Eulerian predictions (Figs. c, f, i, l, o and r) in
determining water level and sediment depth profiles, at different times after dam break, when upstream sediment depth is 0.22 m and dam downstream is wet with
0.05 m water.

In Fig. 8, water level and sediment depth profiles are shown while, At 0.8 and 1 s, the wave tip has reached 2.6 m and 2.85 m from the
upstream sediment depth is 0.22 m and downstream is wet with 0.05 m reservoir beginning point, respectively, while water level upstream is
of standing water. As in previous figure, the experimental and both decreasing and rising in the downstream region. For instance, water
numerical results are in good agreements. At 0.4 s, the positive wave level at 0.8 s, is about 0.24 m at point 1.4 m and about 0.07 m at point
hits the downstream standing water creating an increase in water ele- 2.5 m. However, at 1 s, water level at 1.4 m decreases to 0.22 m and
vation and reaches 1.95 m from the reservoir beginning and at 0.6 s, the rises to a maximum of 0.1 m at point 2.5 m. Investigating images at
wave tip reaches 2.25 m downstream point. later times after dam break indicates the saturated sediment layer

8
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

upstream has moved downstream up to 0.18 m. As previously dis- Compared to the 0.3 m depth of the reservoir water, error indices seem
cussed, the sediment depth at the dam location and dam upstream insignificant as the maximum RMSE and MAE are 4% and 2% of the
decreased as time passes. For instance, at 1 s after dam break, the se- reservoir water depth, respectively. However, as the downstream con-
diment depth decreased about 0.08 m at dam location, which is well dition gets more complicated with higher flow turbulence, the statis-
predicted by both numerical methods. tical error indices increase accordingly (Fig. 9, c and d). The experi-
As the upstream sediment depth increases, the movement of satu- mental results at 0.04, 0.08 and 0.2 s (early times), indicate that the
rated sediment layer as well as its depth in the downstream direction dam break wave propagated up to 0.7 m upstream and 0.7 m down-
increase accordingly. On the other hand, the downstream initial water stream. However, at 0.4, 1, 2, and 6 s (late times), the negative wave
presence reduces the advancement of saturated sediment layer, redu- reached the flume beginning (1.52 m upstream of the dam). The water
cing the risk of burying structures and equipment under the massive level was affected at the beginning of the reservoir within 0.4 s and
saturated sediment layer. However, downstream initial water level decreased to less than 0.04 m at 6 s there. The waves shape and pro-
leads to increase the maximum water level there after dam break, and pagation pattern differ in dry versus wet downstream conditions. In dry
hence, increases immersion risk of downstream structures and equip- downstream condition, the wave tip propagates about 0.7 m at early
ment. Comparing the predictions of VOF and Eulerian methods with times (0.2 s) and progresses smoothly. However, in wet downstream
experimental results, it is clear that Eulerian, especially for wet condition, at the same time this wave only propagates up to 0.45 m and
downstream conditions, is more accurate in predicting the details at the has several humps.
wave forehead region. Also, if the upstream sediment depth is more In the late times (0.4, 1, 2 and 6 s) after dam break, the effect of
than half of the reservoir's height, then Eulerian is more accurate and downstream conditions on the wave propagation pattern is significant.
closer to experimental results, in both terms of the sediment layer depth For instance, wet downstream in comparison with dry downstream has
and movement in the downstream direction. Despite advantages such as caused an increase in the water level at the flume beginning point,
higher accuracy, Eulerian has disadvantages; 1) longer computational created several humps in the progressive wave, and also increased the
times, 2) more complex simulation conditions, and 3) requiring pro- maximum flood wave’s depth.
fessional skills to design and run a multiphase model. Fig. 10 (a–d) shows water surface profiles after dam break while the
upstream sediment depth is 0.03 m and the downstream is dry (Fig. 10
a and c) and wet with 0.02 m of sitting water (Fig. 10 b and d). All
4.2. Water surface and sediment depth profile
figures depict experimental measurements as well as VOF (Fig. 10 a and
b) and Eulerian (Fig. 10 c and d) results. The measured data and both
Fig. 9 shows water surface profiles at different times after dam break
VOF and Eulerian results are in good agreement, and the maximum
when the upstream reservoir is filled up by clear water (without sedi-
RMSE and MAE values (which occurred in the VOF method) are 0.013
ment) and the downstream is either dry (Fig. 9 a and b) or wet with
and 0.0098 m, respectively. These errors are negligible compared to the
0.05 m of sitting water (Fig. 9 c and d). Figures depict the profiles for
0.3 m depth of the reservoir water. Profile results at 0.4, 1, 2, and 6 s
early (Fig. 9 a and c) and late (Fig. 9 b and d) times after dam break. All
(late times) indicate that the negative wave has reached the flume be-
figures show both laboratory measurements (points) and VOF results
ginning in 0.4 s, and the water level has changed there from 0.3 m to
(lines).
around 0.05 m in 6 s. Although many of the analyses are similar to
The experimental and VOF results are in good agreements and the
Fig. 9, comparisons of water level in Fig. 10 (a–d) show that Eulerian
maximum RMSE and MAE are 0.012 m and 0.006 m, respectively.

Exp. Data @ 0.04 Sec Num. Results @ 0.04 Sec Exp. Data @ 0.4 Sec Num. Results @ 0.4 Sec
Exp. Data @ 0.08 Sec Num. Results @ 0.08 Sec Exp. Data @ 1 Sec Num. Results @ 1 Sec
Exp. Data @ 0.2 Sec Num. Results @ 0.2 Sec Exp. Data @ 2 Sec Num. Results @ 2 Sec
Exp. Data @ 6 Sec Num. Results @ 6 Sec
(a) (b)
0.35 0.35
0.3 RMSE = 0.0061 (m) 0.3 RMSE = 0.0036 (m)
Water depth (m)

Water depth (m)

0.25 MAE = 0.0031 (m) 0.25 MAE = 0.0028 (m)


0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Distance along flume (m) Distance along flume (m)

(c) (d)
0.35 0.35
0.3 RMSE = 0.0122 (m) 0.3 RMSE = 0.0063 (m)
Water depth (m)

Water depth (m)

0.25 MAE = 0.0063 (m) 0.25 MAE = 0.0049 (m)


0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Distance along flume (m) Distance along flume (m)
Fig. 9. Comparison of laboratory results and VOF predictions, in determining water level profiles at early (0.04, 0.08 and 0.2 s) and late (0.4, 1, 2 and 6 s) times after
the dam break, when the upstream reservoir is filled up with clear water (without any sediment) and the downstream is, a) dry (early times), b) dry (late times), c)
wet (early times) and d) wet (late times).

9
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Exp. Data @ 0.4 Sec Num. Results @ 0.4 Sec


Exp. Data @ 1 Sec Num. Results @ 1 Sec
Exp. Data @ 2 Sec Num. Results @ 2 Sec
Exp. Data @ 6 Sec Num. Results @ 6 Sec
(a) (b)
0.35 0.35

Water depth (m)


RMSE = 0.0131 (m) RMSE = 0.0115 (m)
Water depth (m)

0.3 0.3
0.25 MAE = 0.0098 (m) 0.25 MAE = 0.0094 (m)
0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Distance along flume (m) Distance along flume (m)

(c) (d)
0.35 0.35
RMSE = 0.0085 (m) RMSE = 0.0091 (m)
Water depth (m)

0.3 0.3

Water depth (m)


0.25 MAE = 0.0073 (m) 0.25 MAE = 0.0076 (m)
0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Distance along flume (m) Distance along flume (m)

(e) (f)
0.08 0.08
Sediment depth (m)

Sediment depth (m)

RMSE = 0.0055 (m) RMSE = 0.0083 (m)


0.06 MAE = 0.0034 (m) 0.06 MAE = 0.0047 (m)

0.04 0.04

0.02 0.02

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Distance along flume (m) Distance along flume (m)

(g) (h)
0.08 0.08
Sediment depth (m)

Sediment depth (m)

RMSE = 0.0053 (m) RMSE = 0.0046 (m)


0.06 MAE = 0.0029 (m) 0.06 MAE = 0.0023 (m)
0.04 0.04

0.02 0.02

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Distance along flume (m) Distance along flume (m)
Fig. 10. Comparison of laboratory measurements and simulation predictions by VOF (a, b, e & f) and Eulerian (c, d, g & h) methods in determining water level
profiles (a, b, c & d), when upstream sediment depth is 0.03 m, and in determining sediment depth profiles (e, f, g & h), when upstream sediment depth is 0.075 m, at
0.4, 1, 2 and 6 s after the dam break under different downstream conditions; dry downstream condition (a, c, e & g), wet downstream with 0.02 m of standing water
depth (b & d), or with 0.05 m of standing water depth (f & h).

results are more accurate than VOF in predicting the measured data. All figures depict measured data as well as VOF (Fig. 10 a and b) and
RMSE and MAE values for Eulerian are 0.0091 m and 0.0076 m, re- Eulerian (Fig. 10 c and d) results. The experimental measurements are
spectively (RMSE is 3% and MAE is 2.53% with respect to the 0.3 m in good agreement with VOF and Eulerian results and the maximum
depth of the reservoir water). While wet downstream condition has not RMSE and MAE values are 0.0083 m and 0.0047 m, respectively that
had much of an effect on the upstream wave pattern, it has had a occurred in VOF when the downstream was initially wet. Profile results
profound effect on the downstream wave pattern and created a bumpy at 0.4, 1, 2, and 6 s indicate that the sediment negative wave has ad-
profile. vanced back to upstream by 0.15 m in 0.4 s to a maximum of 0.5 m in
Fig. 10 (e–h) shows sediment depth profiles after dam break while 6 s and its positive wave has moved up to 0.35 m to the dam down-
the upstream sediment depth is 0.075 m and the downstream is dry stream. Also, the sediment layer propagation differs in dry versus wet
(Fig. 10 a and c) or wet with 0.05 m of sitting water (Fig. 10 b and d). downstream conditions. However, in dry bed condition, the sediment

10
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Table 2
Statistical error values in predicting water surface profile by the VOF and Eulerian numerical methods, at times 0.4, 1, 2 and 6 s after dam break.

layer tip propagates around 0.35 m to downstream, in wet downstream presented in Table 2, covering upstream sediment depths of
condition, it only moves up to 0.25 m there. 0.03–0.24 m and dry-bed or wet-bed downstream initial conditions. As
Comparisons of sediment depth profiles in Fig. 10 (e–h) show that reflected in Table 2, the experimental measurements are in good
Eulerian is more accurate than VOF in predicting measured data. RMSE agreement with VOF and Eulerian results, and the maximum RMSE and
and MAE values for the Eulerian method are 0.0046 m and 0.0023 m, MAE were 0.043 m and 0.03 m respectively. The maximum error values
respectively; RMSE is 6.1% and MAE is 3.1% with respect to the occurred in VOF for wet downstream condition with 0.04 and 0.05 m
0.075 m depth of the sediment layer. Deeper investigation indicates, in water. In these cases, the upstream sediment depths were 0.22 m and
early times after dam break (up to 0.2 s), the reservoir sediment layer 0.2 m. However, according to simulation assumptions such as con-
was not influenced by downstream dry-bed or wet-bed conditions and sidering the saturated sediment layer as a viscous fluid, the results are
no significant changes has occurred in the sediment layer initial form of satisfactory and the maximum amounts of error seems acceptable as
a relatively sharp step. In the latter case, low sediment depth and tur- compared to 0.3 m depth of reservoir water. As previously mentioned,
bulence of the wave, may prevent significant deformation in sediment it seems, as the downstream condition gets more complicated, the error
layer at early times. Note that water surface profiles when the upstream indices increase accordingly. It should be noted that, the RMSE value
sediment depth is 0.15 m with different downstream conditions are presented for each dam break scenario in Table 2 is an average of the
shown in Fig. S10, in Supplementary Material. The sediment depth errors calculated in 16 distinct surveyed locations along the channel
profiles when the upstream sediment depth is 0.2 m are also depicted and also at 4 time snaps (0.4, 1, 2 and 6 s) after the dam break.
and discussed in Fig. S11, in Supplementary Material. Deeper investigation showed that in the lower sediment depth
The statistical error values of VOF and Eulerian in predicting water conditions (0.03 m and 0.075 m), the ability of numerical models is
surface profile at late times (0.4, 1, 2, and 6 s) after dam break are higher than other cases. Moreover, in dry downstream condition, the

11
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Water Level @ 0.76 (m), Exp. Water Level @ 1.52 (m), Exp. Water Level @ 2.52 (m), Exp.

Water Level @ 0.76 (m), Num. Water Level @ 1.52 (m), Num. Water Level @ 2.52 (m), Num.

(a) (b)
0.35 0.35
0.3
RMSE = 0.0041 (m) RMSE = 0.0073 (m)
0.3
MAE = 0.0019 (m) MAE = 0.0046 (m)
0.25 0.25
0.2 0.2
Y (m)

Y (m)
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t (sec) t (sec)

(c) (d)
0.35 0.35
RMSE = 0.0079 (m) RMSE = 0.0065 (m)
0.3 0.3
MAE = 0.0057 (m) MAE = 0.004 (m)
0.25 0.25
0.2 0.2
Y (m)

0.15 Y (m) 0.15


0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t (sec) t (sec)
Fig. 11. Comparison of laboratory modeling results and VOF predictions, in determining variation of water level with passing time after dam break at 3 fixed
locations, including upstream reservoir midpoint (0.76 m), dam location (1.52 m) and dam downstream (2.52 m) when, reservoir is filled up with clear water
(without upstream sediment), under various downstream conditions: a) dry downstream, b) wet downstream with 0.02 m water, c) wet downstream with 0.04 m
water and d) wet downstream with 0.05 m water.

best fit between experimental and numerical results is achieved. By (Section 2.1). All graphs are presented for three fixed locations and the
comparing the statistical error indices, it is observed that the Eulerian distances between locations from reservoir beginning are 0.76 m (up-
had better fit to the measured data than the VOF. For instance, when stream reservoir midpoint), 1.52 m (dam location) and 2.52 m (dam
the upstream sediment depth was 0.2 m and downstream was wet with downstream). The channel side view and three fixed locations are il-
0.05 m water, the Eulerian yielded error indices that are 17% lower lustrated in Fig. S12 in Supplementary Material. For data extraction
than the VOF, pointing to a superior performance. Nevertheless, the from these locations, both first and second camera (mid camera) are
insignificant errors reported from different cases indicate the high used covering about 3.7 m of the flume length.
ability of OpenFOAM in simulating the multiphase shock waves. The In Fig. 11, the water level variation as time elapses after dam break
statistical error values of VOF and Eulerian in predicting sediment at 3 fixed locations are shown. Here, the reservoir is filled up by clear
depth profiles at 0.4, 1, 2 and 6 s after dam break are presented in Table water with different downstream conditions. Points and lines represent
S1 in Supplementary Material. experimental measurements and VOF results, respectively. The ex-
In general, the calculated MAEs vary between 0.006 m and 0.031 m perimental and VOF results are in good agreement and as the down-
(2%–10% with respect to the reservoir height of 0.3 m) and the RMSEs stream condition gets more complicated, the error indices increase ac-
vary between 0.008 m and 0.043 m (2.6%–4% with respect to the cordingly. The maximum RMSE and MAE were 0.0079 m and
maximum reservoir height of 0.3 m). Results of all numerical simula- 0.0057 m, respectively; 2.6% and 1.9% with respect to the 0.3 m depth
tions indicate that, especially in some cases with high upstream sedi- of reservoir water, respectively.
ment depths, the Eulerian method showed slightly superior perfor- As expected, after dam break for up to 3 s, the upstream water level
mances than the VOF and the error values in the Eulerian were lower (at 0.76 m) decreases rapidly. While, at the same time, the downstream
than the VOF by 5%–20%. Accommodating mixing of the phases in water level (2.52 m) rises rapidly and then slowly decreases along the
Eulerian may have prompted its superior results. flume. Once the dam breaks, the water depth at dam location (1.52 m)
sharply reduced and after about 0.5 s, it reached less than half of its
initial depth, and then gradually decreases. As reflected in the figure,
4.3. Investigation of water level and sediment depth variations over time presence of downstream initial water has led to an increase in water
level in the dam downstream. Particularly, during late times (4–6 s), the
In this part, the results of VOF and Eulerian methods are compared effect of downstream conditions on the water level in all three surveyed
with experimental measurements in predicting the variation of water points (upstream, dam location and downstream) was evident, and the
level and sediment depth as time elapses. The MAE and RMSE indices water level at all points for wet downstream condition was higher than
are also calculated and presented for all cases to assess the agreement dry downstream condition.
between experimental and numerical results. For this purpose, 15 time In Fig. 12, both water level and sediment depth variations have been
snaps are considered in a wide range, of 0.04–6 s after dam break

12
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Water Level @ 0.76 (m), Exp. Sed. Depth @ 0.76 (m), Exp. Water Level @ 1.52 (m), Exp.
Sed. Depth @ 1.52 (m), Exp. Water Level @ 2.52 (m), Exp. Sed. Depth @ 2.52 (m), Exp.
Water Level @ 0.76 (m), Num. Sed. Depth @ 0.76 (m), Num. Water Level @ 1.52 (m), Num.
Sed. Depth @ 1.52 (m), Num. Water Level @ 2.52 (m), Num. Sed. Depth @ 2.52 (m), Num.
(a) (b)
0.4 0.4
0.35 RMSE = 0.0338 (m) 0.35 RMSE = 0.0365 (m)
MAE = 0.0239 (m) MAE = 0.0255 (m)
0.3 0.3
0.25 0.25
Y (m)

Y (m)
0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t (sec) t (sec)

(c) (d)
0.4 0.4
0.35 RMSE = 0.0311 (m) 0.35 RMSE = 0.0315 (m)
MAE = 0.0206 (m) MAE = 0.0204 (m)
0.3 0.3
0.25 0.25
Y (m)

0.2 Y (m) 0.2


0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t (sec) t (sec)
Fig. 12. Comparison of laboratory experimental results and both VOF (a and b) and Eulerian (c and d) predictions, in determining water level and sediment depth
variation with passing time after dam break at 3 fixed locations, including upstream reservoir midpoint (0.76 m), dam location (1.52 m) and dam downstream
(2.52 m) when, upstream sediment depth was 0.24 m, under various downstream conditions: (a and c) dry downstream, (b and d) wet downstream with 0.02 m
water.

examined for three fixed locations along the flume, and results of VOF cases with high sediment depths upstream, the error values in Eulerian
and Eulerian numerical methods have been compared with experi- were lower, showing slightly superior performance than the VOF. Noted
mental measurements when upstream sediment depth was 0.24 m with that, in Fig. S13, in Supplementary Material, the time variation results
different downstream conditions. The experimental and both VOF and of VOF and Eulerian methods have been compared with experimental
Eulerian results are in acceptable agreement with maximum RMSE and measurements when upstream sediment depth is 0.03 m with different
MAE of 0.0365 and 0.0255 m, respectively; 12.1% and 8.5% of initial downstream conditions.
reservoir water level (0.3 m). Based on the simulation assumptions such
as considering the saturated sediment layer as a viscous fluid, the re-
4.4. Investigation of correlation factor in experimental and numerical
sults are acceptable.
results
The water level variation in dam upstream and downstream has
behaved in the same manner as previously discussed. As expected, once
In this section, by presenting several graphs for different upstream
the dam breaks, the water level at dam location (1.52 m) sharply re-
and downstream conditions, and comparing the coefficient of
duced and after 0.5 s reached less than 0.2 m and then gradually de-
determination(R2) , the correlation between experimental results and
creased. Experimental results indicate that the sediment depth at the
numerical simulations is evaluated. In Fig. 13 (a), the correlation be-
dam upstream (0.76 m) and downstream (2.52 m) do not change much
tween experimental measurements and VOF results in determining
in the first 2 s after the dam break. However, in the same time the
water level after dam break is shown for different times. In this case, the
sediment depth rapidly decreased (about 0.1 m) at the dam section
reservoir is filled up by clear water and downstream is wet with 0.04 m
(1.52 m) and then slowly decreased. Moreover, at 2 s after dam break
water. As it can be seen, the experimental measurements are highly
the downstream sediment depth, slowly started to increase, and
correlated with VOF results and the R2 is 0.989.
reached 0.025 m at 6 s.
Fig. 13 (b) and (c) indicates the correlation of laboratory mea-
It is also observed in Fig. 12 that different downstream conditions
surements versus VOF or Eulerian numerical methods in determining
(dry-bed or wet-bed) at early times after dam break (up to 0.4 s) did not
water level variation at different times after dam break for the case that
have any effect on water level variation, at upstream and dam section
the upstream sediment depth is 0.075 m and downstream is wet with
(1.52 m). However, presence of downstream initial water has led to an
0.05 m water. Experimental measurements are correlated with both
increase in the water level at dam downstream, and the effect of wet
VOF and Eulerian results with R2 values of 0.983 and 0.987, respec-
downstream condition on the water level at dam section is obvious in
tively, which indicates an acceptable accuracy of results. However, a
late times (6 s).
small difference in the determination coefficients shows the Eulerian
In general, the experimental measurements are in good agreement
results have a higher precision as compared to VOF results. The sta-
with VOF and Eulerian results. Also, as previously discussed, in some
tistical error value presented in this study for each scenario, is an

13
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

(a)
0.3
0.04 Sec
0.08 Sec
0.25
0.2 Sec
0.4 Sec
1 Sec

VOF results (m)


0.2
2 Sec
6 Sec
0.15

0.1 0.9889

0.05

0
0 0.05 0.1 0.15 0.2 0.25 0.3
Experimental data (m)

(b) (c)
0.3 0.3
0.04 Sec 0.04 Sec
0.08 Sec 0.08 Sec
0.25 0.25
0.2 Sec 0.2 Sec
0.4 Sec 0.4 Sec
Eulerian results (m)

1 Sec 1 Sec
VOF results (m)

0.2 0.2
2 Sec 2 Sec
6 Sec 6 Sec
0.15 0.15

0.1 0.9834 0.1 0.9873

0.05 0.05

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
Experimental data (m) Experimental data (m)
Fig. 13. Correlation analysis of experimental measurements and the results of VOF and Eulerian numerical methods in determining water level variation along the
flume, for different times after dam break; a) when the reservoir is filled up with clear water and downstream is wet with 0.04 m standing water depth, b) and c)
when the upstream sediment depth is 0.075 m and downstream is wet with 0.05 m standing water depth.

average of the errors calculated for different sections of the flume at front celerity values for these intervals are provided in 4 tables (due to 4
various time snaps after the dam break. However, according to Fig. 13, downstream conditions) presented in Supplementary Material (Section
in some locations along the flume, high error values are observed. For S11). However, the mean measured wave front celerity along dam
example, in Fig. 13(b), at 6 s after the dam break, the error value has downstream for all 32 dam break scenarios are classified in Table 3.
reached 40%–50% in some locations along the flume. The reason seems Investigating Table 3, it is concluded that the upstream sediment
to be the difference between threshold shear stress in a fluid with non- depth is an important factor which determine the celerity values. The
Newtonian behavior (reality) and the assumption of numerical mod- deeper the upstream sediment layer, the slower the dam break flood
eling in this study in which Newtonian fluid is assumed. This has led to wave propagates. It seems that shallower water flow on the top of the
significant differences between observed data and numerical results at sediment layer is one of the reasons resulting in lower wave front cel-
some locations at the beginning of the dam reservoir. It should be noted erity values. Downstream initial standing water also leads to decrease in
that the correlation analysis for the case that the upstream sediment the flood celerity; deeper downstream standing water reduced the wave
depth is 0.2 m while downstream is dry has been presented in Fig. S14, front celerity further. Reduced wave velocity due to downstream initial
in Supplementary Material. water level is also reported in previous studies (Leal et al., 2006; Deng
et al., 2018). As reflected, the mean wave front celerity values along the
4.5. Investigation of wave front celerity flume, are in a range of 1.08–2.32 m/s, based on different dam up-
stream and downstream conditions. The variance of each mean celerity
The wave front celerity values were rigorously measured by means value was also calculated and classified in Table 3 (VAR line). As it can
of extracted images at four 1-m-long intervals through dam down- be seen, the calculated variance values are insignificant and varied from
stream. These four intervals totally cover 4 m length of dam down- 0.004 to a maximum of 0.042 m2/s2 and the maximum variance values
stream (1.52 m–5.52 m from reservoir beginning). The measured wave have occurred in the cases with deeper initial upstream sediment.

14
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

Table 3 from 0.006 m to 0.031 m (2% to 10% with respect to the reservoir
Mean measured wave front celerity along the flume (m/s). height of 0.3 m), and in terms of Root Mean Square Error (RMSE) varied
Mean wave front celerity (m/s) from 0.008 m to 0.043 m (2.6% to 14% with respect to the maximum
reservoir height of 0.3 m). Although, comparing coefficient of de-
Upstream Different downstream conditions termination (R2 ) showed acceptable correlations between the measured
sediment data and the results of VOF and Eulerian approaches, the Eulerian
depth (m) Dry downstream Downstream Downstream Downstream
with 0.02 m with 0.04 m with 0.05 m
method indicated a better correlation with experimental results.
standing standing standing However, required computational time for Eulerian approach was far
water water water more than that of VOF; in some cases, twice as long. It requires more
complex simulation conditions, especially when utilizing as a multi-
0 2.32 1.95 1.83 1.78
VAR* 0.012 0.021 0.020 0.017
phase modeling tool.
0.03 2.26 1.85 1.76 1.73 Future research could examine the potential effects of changing
VAR 0.014 0.007 0.019 0.022 downstream condition along the flume on such a multiphase shock
0.075 2.17 1.80 1.71 1.65 wave propagation. Simulating the saturated sediment layer as a fluid
VAR 0.014 0.008 0.017 0.017
with non-Newtonian behavior or as solid particles might prove im-
0.15 2.04 1.62 1.54 1.48
VAR 0.007 0.013 0.004 0.006 portant, too. Moreover, evaluating the application of expert systems,
0.175 1.92 1.55 1.44 1.42 particularly artificial neural networks (ANNs) on predicting such
VAR 0.012 0.016 0.011 0.009 complex issues may constitute the object of future studies. First of all,
0.2 1.75 1.47 1.34 1.31 this is desirable to apply artificial intelligence or genetic algorithm
VAR 0.029 0.018 0.008 0.008
0.22 1.65 1.35 1.23 1.20
models to make an accurate estimation for such multiphase flood waves
VAR 0.042 0.025 0.011 0.011 (Nikoo et al., 2015, 2016). Since, artificial intelligence models may
0.24 1.28 1.18 1.13 1.08 have different accuracies, fusion-based methods are common ap-
VAR 0.040 0.014 0.006 0.005 proaches could be conducted on the wide database provided in this
research (Alizadeh and Nikoo, 2018; Ghazali et al., 2018; Taravatrooy
* VAR: Variance value (m2 /s2 ).
et al., 2018). In the last, this is very much the key component in future
attempts to discuss the uncertainties in the peak estimation as well as
5. Summary and conclusion
the hydrograph of dam break flood, using interval optimization
methods, factorial analysis and also the method of conditional value at
In the present study, two- and three-phase flood waves caused by
risk (/CVaR) (Naserizade et al., 2018; Soltani et al., 2016; Tavakoli
sudden dam break with silted-up reservoirs have been studied both
et al., 2014). Above all, this issue would need to be explored further,
experimentally and numerically. Different initial conditions including
using different types of sediments with various grain sizes and grada-
various upstream sediment depths and dry-bed or wet-bed downstream
tions and this is desirable for future work to take into account the
have been considered which created a total of 32 scenarios. Multiphase
consolidation of sediment.
dam break shock waves were filmed by cameras mounted along the
flume and image processing method was employed to extract water
CRediT authorship contribution statement
level and sediment depth data from 16 locations along the flume and 15
snap times after the dam breaks. Numerical simulations were conducted
Foad Vosoughi: Conceptualization, Software, Validation,
using OpenFOAM via VOF and Eulerian numerical approaches.
Investigation, Resources, Data curation, Writing - original draft,
Authors concluded that upstream sediment depth is an important
Visualization, Writing - review & editing. Gholamreza
factor in determining flood wave propagation velocity. Particularly,
Rakhshandehroo: Conceptualization, Methodology, Supervision,
higher upstream sediment depths (and less water depth on it), caused a
Writing - review & editing. Mohammad Reza Nikoo:
slower flood wave propagation downstream. It was postulated that the
Conceptualization, Methodology, Supervision, Writing - review &
high sediment viscosity is a factor for the wave front celerity decrease
editing. Mojtaba Sadegh: Supervision, Writing - review & editing.
under these conditions. On the other hand, by increasing upstream
sediment depth, both length and depth of the saturated sediment pro-
Declaration of Competing Interest
pagation to downstream area increase, raising the risk of burying
downstream facilities. The results demonstrate a strong effect of
downstream initial water on the flood wave, which reduces the ad- The authors declare that they have no known competing financial
vancement of saturated sediment layer, leading to a decrease in flood interests or personal relationships that could have appeared to influ-
wave velocity. The deeper the downstream initial water is, the slower ence the work reported in this paper.
the flood wave propagates (Leal et al., 2006; Deng et al., 2018).
Nevertheless, presence of downstream initial water has led to increase Appendix A. Supplementary data
in maximum flood wave height in downstream, which could raise the
immersion risk of structures and equipment located in prone areas. Supplementary data to this article can be found online at https://
Despite limitations such as assuming the upstream saturated sedi- doi.org/10.1016/j.jhydrol.2020.125267.
ment layer as a viscous fluid in numerical modeling, comparing the
numerical results of VOF and Eulerian approaches with laboratory References
measurements showed that they are in good agreements. The Eulerian
approach slightly outperformed VOF, especially in cases with high Yang, C.T., 1996. Sediment Transport: Theory and Practice. McGraw-Hill.
Vischer, D.L., Hager, W.H., 1998. Dam Hydraulics. Wiley, Chichester, New York.
upstream sediment depths. Accommodating phase mixing in Eulerian Duarte, R., Ribeiro, J., Boillat, J.L., Schleiss, A., 2011. Experimental study on dam-break
method is believed to have caused its superior results. Further, good waves for silted-up reservoirs. J. Hydraul. Eng. https://doi.org/10.1061/(ASCE)HY.
controls over the conditions in the lab such as no leakage from gate 1943-7900.0000444.
Open∇FOAM, The Open Source CFD Toolbox – User Guide, OpenFOAM Found. 15th
edges, short gate opening time (0.08–0.16 s), high imaging quality and December. (2015).
precise computer processing of the images appeared to be among rea- Dressler, R.F., 1954. Comparison of theories and experiments for the hydraulic dam-break
sons leading to good matching of experimental and numerical results. wave. Int. Assoc. Sci, Hydrol.
Bell, S.W., Elliot, R.C., Hanif Chaudhry, M., 1992. Experimental results of two-dimen-
Maximum error values in terms of Mean Absolute Error (MAE) varied sional dam-break flows. J. Hydraul. Res. https://doi.org/10.1080/

15
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

00221689209498936. methodology for land subsidence estimation. Nat. Hazards 94, 905–926. https://doi.
Bellos, C.V., Soulis, V., Sakkas, J.G., 1992. Experimental investigation of two-dimensional org/10.1007/s11069-018-3431-8.
dam-break induced flows. J. Hydraul. Res. https://doi.org/10.1080/ Tavakoli, A., Kerachian, R., Nikoo, M.R., Soltani, M., Malakpour Estalaki, S., 2014. Water
00221689209498946. and waste load allocation in rivers with emphasis on agricultural return flows: ap-
Fraccarollo, L., Toro, E.F., 1995. Experimental and numerical assessment of the shallow plication of fractional factorial analysis. Environ. Monit. Assess. 186, 5935–5949.
water model for two-dimensional dam-break type problems. J. Hydraul. Res. https:// https://doi.org/10.1007/s10661-014-3830-6.
doi.org/10.1080/00221689509498555. Valiani, A., Caleffi, V., Zanni, A., 2002. Case study: Malpasset dam-break simulation using
Lauber, G., Hager, W.H., 1998. Experiments to dambreak wave: horizontal channel. J. a two-dimensional finite volume method. J. Hydraul. Eng. https://doi.org/10.1061/
Hydraul. Res. https://doi.org/10.1080/00221689809498620. (ASCE)0733-9429(2002)128:5(460).
Soares Frazão, S., Zech, Y., 2002. Dam break in channels with 90° bend. J. Hydraul. Eng. Zoppou, C., Roberts, S., 2003. Explicit schemes for dam-break simulations. J. Hydraul.
https://doi.org/10.1061/(ASCE)0733-9429(2002)128: 11(956). Eng. https://doi.org/10.1061/(ASCE)0733-9429(2003)129:1(11).
Bellos, C.V. 2004. Experimental Measurements of Flood Wave Created by a Dam Break. Ackerman, C.T., Fleming, M.J., Brunner, G.W., 2008. Hydrologic and hydraulic models
Eur. Water. for performing dam break studies. In: World Environ. Water Resour. Congr. 2008
Meile, T., Boillat, J.L., Schleiss, A.J., 2013. Propagation of surge waves in channels with Ahupua’a – Proc. World Environ. Water Resour. Congr. https://doi.org/10.1061/
large-scale bank roughness. J. Hydraul. Res. https://doi.org/10.1080/00221686. 40976(316)285.
2012.738876. Cao, Z., Pender, G., Wallis, S., Carling, P., 2004. Computational dam-break hydraulics
Wood, A., Wang, K.H., 2015. Modeling dam-break flows in channels with 90 degree bend over erodible sediment bed. J. Hydraul. Eng. https://doi.org/10.1061/(ASCE)0733-
using an alternating-direction implicit based curvilinear hydrodynamic solver. 9429(2004)130:7(689).
Comput. Fluids. https://doi.org/10.1016/j.compfluid.2015.03.011. Leal, J.G.A.B., Ferreira, R.M.L., Cardoso, A.H., 2006. Dam-break wave-front celerity. J.
Xue, Y., Xu, W.L., Luo, S.J., Chen, H.Y., Li, N.W., Xu, L.J., 2011. Experimental study of Hydraul. Eng. https://doi.org/10.1061/(ASCE)0733-9429(2006)132:1(69).
dam-break flow in cascade reservoirs with steep bottom slope. J. Hydrodyn. https:// Wu, W., Wang, S.Y., 2007. One-dimensional modeling of dam-break flow over movable
doi.org/10.1016/S1001-6058(10)60140-0. beds. J. Hydraul. Eng. https://doi.org/10.1061/(ASCE)0733-9429(2007)133:1(48).
Kocaman, S., Ozmen-Cagatay, H., 2012. The effect of lateral channel contraction on dam Crespo, A.J.C., Gómez-Gesteira, M., Dalrymple, R.A., 2008. Modeling dam break behavior
break flows: Laboratory experiment. J. Hydrol. https://doi.org/10.1016/j.jhydrol. over a wet bed by a SPH technique. J. Waterw. Port CoastOcean Eng. https://doi.org/
2012.02.035. 10.1061/(ASCE)0733-950X(2008)134: 6(313).
LaRocque, L.A., Imran, J., Chaudhry, M.H., 2013. Experimental and numerical in- Zoppou, C., Roberts, S., 2000. Numerical solution of the two-dimensional unsteady dam
vestigations of two-dimensional dam-break flows. J. Hydraul. Eng. https://doi.org/ break. Appl. Math. Model. https://doi.org/10.1016/S0307-904X(99)00056-6.
10.1061/(ASCE)HY.1943-7900.0000705. Aliparast, M., 2009. Two-dimensional finite volume method for dam-break flow simula-
Ozmen-Cagatay, H., Kocaman, S., Guzel, H., 2014. Investigation of dam-break flood tion. Int. J. Sediment Res. https://doi.org/10.1016/S1001-6279(09)60019-6.
waves in a dry channel with a hump. J. Hydro-Environ. Res. https://doi.org/10. Evangelista, S., Altinakar, M.S., Di Cristo, C., Leopardi, A., 2013. Simulation of dam-break
1016/j.jher.2014.01.005. waves on movable beds using a multi-stage centered scheme. Int. J. Sediment Res.
Leal, J., Ferreira, R.M.L., Franco, A.B., Cardoso, A.H., 2001. Dam-break waves over https://doi.org/10.1016/S1001-6279(13)60039-6.
movable bed channels experimental study. In: Proc. Congr. Assoc. Hydraul. Res. Postacchini, M., Othman, I.K., Brocchini, M., Baldock, T.E., 2014. Sediment transport and
Soltani, M., Kerachian, R., Nikoo, M.R., Noory, H., 2016. A conditional value at risk-based morphodynamics generated by a dam-break swash uprush: Coupled vs uncoupled
model for planning agricultural water and return flow allocation in river systems. modeling. Coast. Eng. https://doi.org/10.1016/j.coastaleng.2014.04.003.
Water Resour. Manag. 30, 427–443. https://doi.org/10.1007/s11269-015-1170-0. Ghazali, M., Honar, T., Nikoo, M.R., 2018. A fusion-based neural network methodology
Spinewine, B., Zech, Y., 2007. Small-scale laboratory dam-break waves on movable beds. for monthly reservoir inflow prediction using MODIS products. Hydrol. Sci. J. 63,
J. Hydraul. Res. https://doi.org/10.1080/00221686.2007.9521834. 2076–2096. https://doi.org/10.1080/02626667.2018.1558365.
Cochard, S., Ancey, C., 2008. Tracking the free surface of time-dependent flows: Image He, Z., Wu, T., Weng, H., Hu, P., Wu, G., 2017. Numerical simulation of dam-break flow
processing for the dam-break problem. Exp. Fluids. https://doi.org/10.1007/s00348- and bed change considering the vegetation effects. Int. J. Sediment Res. https://doi.
007-0374-3. org/10.1016/j.ijsrc.2015.04.004.
Leal, J.G.A.B., Ferreira, R.M.L., Cardoso, A.H., 2009. Maximum level and time to peak of Kim, B., Sanders, B.F., 2016. Dam-break flood model uncertainty assessment: Case study
dam-break waves on mobile horizontal bed. J. Hydraul. Eng. https://doi.org/10. of extreme flooding with multiple dam failures in Gangneung, South Korea. J.
1061/(ASCE)HY.1943-7900.0000099. Hydraul. Eng. https://doi.org/10.1061/(ASCE)HY.1943-7900.0001097.
Lal, A.M.W., Moustafa, M.Z. 2016. Dam-break wave fronts in vegetated wetlands. In: Fu, L., Jin, Y.C., 2016. mproved multiphase Lagrangian method for simulating sediment
World Environ. Water Resour. Congr. 2016 Hydraul. Waterw. Hydro-Climate/ transport in dam-break flows. J. Hydraul. Eng. https://doi.org/10.1061/(ASCE)HY.
Climate Chang. – Pap. from Sess. Proc. 2016 World Environ. Water Resour. Congr. 1943-7900.0001132.
https://doi.org/10.1061/9780784479872.012. Biscarini, C., Di Francesco, S., Manciola, P., 2010. CFD modelling approach for dam break
Naserizade, S.S., Nikoo, M.R., Montaseri, H., 2018. A risk-based multi-objective model for flow studies. Hydrol. Earth Syst. Sci. https://doi.org/10.5194/hess-14-705-2010.
optimal placement of sensors in water distribution system. J. Hydrol. 557, 147–159. Yang, C., Lin, B., Jiang, C., Liu, Y., 2010. Predicting near-field dam-break flow and impact
https://doi.org/10.1016/j.jhydrol.2017.12.028. force using a 3D model. J. Hydraul. Res. https://doi.org/10.1080/00221686.2010.
Nikoo, M.R., Hatami Bahman Beiglou, P., Mahjouri, N., 2016. Optimizing multiple-pol- 531099.
lutant waste load allocation in rivers: an interval parameter game theoretic model. Hu, K.C., Hsiao, S.C., Hwung, H.H., Wu, T.R., 2012. Three-dimensional numerical mod-
Water Resour. Manag. 30, 4201–4220. https://doi.org/10.1007/s11269-016-1415-6. eling of the interaction of dam-break waves and porous media. Adv. Water Resour.
Nikoo, M.R., Khorramshokouh, N., Monghasemi, S., 2015. Optimal design of detention https://doi.org/10.1016/j.advwatres.2012.06.007.
rockfill dams using a simulation-based optimization approach with mxed sediment in Marsooli, R., Wu, W., 2015. Three-dimensional numerical modeling of dam-break flows
the llow. Water Resour. Manag. 29, 5469–5488. https://doi.org/10.1007/s11269- with sediment transport over movable beds. J. Hydraul. Eng. https://doi.org/10.
015-1129-1. 1061/(ASCE)HY.1943-7900.0000947.
Nsom, B., 2002. Horizontal viscous dam-break flow: experiments and theory. J. Hydraul. Lamond, B.F., Boukhtouta, A., 2002. Water Reservoir Applications of Markov Decision
Eng. https://doi.org/10.1061/(ASCE)0733-9429(2002)128:5(543). Processes, pp. 537–558. https://doi.org/10.1007/978-1-4615-0805-2_17.
Alizadeh, M.R., Nikoo, M.R., 2018. A fusion-based methodology for meteorological Ozmen-Cagatay, H., Kocaman, S., 2010. Dam-break flows during initial stage using SWE
drought estimation using remote sensing data. Remote Sens. Environ. 211, 229–247. and RANS approaches. J. Hydraul. Res. https://doi.org/10.1080/00221686.2010.
https://doi.org/10.1016/j.rse.2018.04.001. 507342.
Ancey, C., Cochard, S., Andreini, N., 2009. The dam-break problem for viscous fluids in Issakhov, A., Imanberdiyeva, M., 2019. Numerical simulation of the movement of water
the high-capillary-number limit. J. Fluid Mech. https://doi.org/10.1017/ surface of dam break flow by VOF methods for various obstacles. Int. J. Heat Mass
S0022112008005041. Transfer. https://doi.org/10.1016/j.ijheatmasstransfer.2019.03.034.
Minussi, R.B., De Freitas MacIel, G., 2012. Numerical experimental comparison of dam Mokrani, C., Abadie, S., 2016. Conditions for peak pressure stability in VOF simulations of
break flows with non-newtonian fluids. J. Braz. Soc. Mech. Sci. Eng. https://doi.org/ dam break flow impact. J. Fluids Struct. https://doi.org/10.1016/j.jfluidstructs.
10.1590/S1678-58782012000200008. 2015.12.007.
Unofficial OpenFOAM wiki. This page was last modified on 27 July 2019 (accessed 23 Panicker, P.G., Goel, A., Iyer, H.R., 2015. Numerical modeling of advancing wave front in
May 2020). https://openfoamwiki.net/index.php/Main_Page. dam break problem by incompressible Navier-stokes solver. Aquat. Proc. https://doi.
Gabutti, B., 1983. On two upwind finite-difference schemes for hyperbolic equations in org/10.1016/j.aqpro.2015.02.108.
non-conservative form. Comput. Fluids. https://doi.org/10.1016/0045-7930(83) Fondelli, T., Andreini, A., Facchini, B., 2015. Numerical simulation of dam-break problem
90031-2. using an adaptive meshing approach. Energy Proc. https://doi.org/10.1016/j.egypro.
Beam, R.M., Warming, R.F., 1976. An implicit finite-difference algorithm for hyperbolic 2015.12.038.
systems in conservation-law form. J. Comput. Phys. https://doi.org/10.1016/0021- Ye, Z., Zhao, X., 2017. Investigation of water-water interface in dam break flow with a
9991(76)90110-8. wet bed. J. Hydrol. https://doi.org/10.1016/j.jhydrol.2017.02.055.
Chaudhry, M.H., 2008. Open-channel flow: second ed. https://doi.org/10.1007/978-0- Issakhov, A., Zhandaulet, Y., Nogaeva, A., 2018. Numerical simulation of dam break flow
387-68648-6. for various forms of the obstacle by VOF method. Int. J. Multiph. Flow. https://doi.
Fennema, R.J., Chaudhry, M.H., 1990. Explicit methods for 2-D transient free-surface org/10.1016/j.ijmultiphaseflow.2018.08.003.
flows. J. Hydraul. Eng. https://doi.org/10.1061/(ASCE)0733-9429(1990) Jamshidi, F., Heimel, H., Hasert, M., Cai, X., Deutschmann, O., Marschall, H., Wörner, M.,
116:8(1013). 2019. On suitability of phase-field and algebraic volume-of-fluid OpenFOAM® solvers
Mingham, C.G., Causon, D.M., 1998. High-resolution finite-volume method for shallow for gas–liquid microfluidic applications. Comput. Phys. Commun. 236, 72–85.
water flows. J. Hydraul. Eng. https://doi.org/10.1061/(ASCE)0733-9429(1998) https://doi.org/10.1016/j.cpc.2018.10.015.
124:6(605). Kassar, B.B.M., Carneiro, J.N.E., Nieckele, A.O., 2018. Curvature computation in volume-
Taravatrooy, N., Nikoo, M.R., Sadegh, M., Parvinnia, M., 2018. A hybrid clustering-fusion of-fluid method based on point-cloud sampling. Comput. Phys. Commun. https://doi.

16
F. Vosoughi, et al. Journal of Hydrology 590 (2020) 125267

org/10.1016/j.cpc.2017.10.003. Bhusare, V.H., Dhiman, M.K., Kalaga, D.V., Roy, D.S., Joshi, J.B., 2017. CFD simulations
Barbosa, D.V.E., Santos, A.L.G., dos Santos, E.D., Souza, J.A., 2019. Overtopping device of a bubble column with and without internals by using OpenFOAM. https://doi.org/
numerical study: Openfoam solution verification and evaluation of curved ramps 10.1016/j.cej.2017.01.128.
performances. Int. J. Heat Mass Transfer. https://doi.org/10.1016/j. Lubchenko, N., Magolan, B., Sugrue, R., Baglietto, E., 2018. A more fundamental wall
ijheatmasstransfer.2018.11.071. lubrication force from turbulent dispersion regularization for multiphase CFD ap-
Hirt, C.W., Nichols, B.D., 1981. Volume of fluid (VOF) method for the dynamics of free plications. Int. J. Multiph. Flow. https://doi.org/10.1016/j.ijmultiphaseflow.2017.
boundaries. J. Comput. Phys. https://doi.org/10.1016/0021-9991(81)90145-5. 09.003.
Rusche, H., 2002. Computational Fluid Dynamics of Dispersed Two-Phase Flows at High Kia, S.A., Aminian, J., 2017. Hydrodynamic modeling strategy for dense to dilute gas-
Phase Fractions PhD thesis, Imperial College of Science, Technology and Medicine. solid fluidized beds. Particuology. https://doi.org/10.1016/j.partic.2016.06.004.
Heyns, J.A., Oxtoby, O.F., 2014. Modelling surface tension dominated multiphase flows Panda, S.K., Singh, K.K., Shenoy, K.T., Buwa, V.V., 2017. Numerical simulations of liquid-
using the VOF approach. In: 11th World Congr. Comput. Mech. WCCM 2014, 5th Eur. liquid flow in a continuous gravity settler using OpenFOAM and experimental ver-
Conf. Comput. Mech. ECCM 2014 6th Eur. Conf. Comput. Fluid Dyn. ECFD 2014. ification. Chem. Eng. J. https://doi.org/10.1016/j.cej.2016.10.102.
Brackbill, J.U., Kothe, D.B., Zemach, C., 1992. A continuum method for modeling surface Akbari, G.H., Barati, R., 2012. Comprehensive analysis of flooding in unmanaged
tension. J. Comput. Phys. https://doi.org/10.1016/0021-9991(92)90240-Y. catchments. Proc. Inst. Civ. Eng. Water Manag. https://doi.org/10.1680/wama.10.
Deising, D., Marschall, H., Bothe, D., 2016. A unified single-field model framework for 00036.
Volume-Of-Fluid simulations of interfacial species transfer applied to bubbly flows. Barati, R., Rahimi, S., Akbari, G.H., 2012. Analysis of dynamic wave model for flood
Chem. Eng. Sci. https://doi.org/10.1016/j.ces.2015.06.021. routing in natural rivers. Water Sci. Eng. https://doi.org/10.3882/j.issn.1674-2370.
Soleimani, A., Schneiderbauer, S., Pirker, S., 2015. CFD study of the gas-particle flow in a 2012.03.001.
horizontal duct: the impact of the solids wall boundary conditions. Proc. Eng. https:// Deng, X., Liu, H., Lu, S., 2018. Analytical study of dam-break wave tip region. J. Hydraul.
doi.org/10.1016/j.proeng.2015.01.225. Eng. https://doi.org/10.1061/(ASCE)HY.1943-7900.0001453.

17

You might also like