Sie sind auf Seite 1von 16

Smooth and Stepped Spillway Modeling

Using the SPH Method


Juliana D. Nóbrega 1; Jorge Matos 2; Harry E. Schulz 3; and Ricardo B. Canelas 4

Abstract: The smoothed particle hydrodynamics (SPH) method was used to simulate the flow over smooth and stepped spillways in the
nonaerated, skimming flow regime. Two-dimensional numerical simulations were carried out using the DualSPHysics software and
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

compared with experimental data on a 2H:1V sloping spillway for smooth and stepped inverts. Continuous inflow fluid layers with constant
height and uniform velocity were used to establish the flow rate on the broad crested weir. The ability of the SPH to reproduce the main flow
characteristics was analyzed by comparing the numerical flow depths and velocity profiles with their experimental counterparts. Firstly, a
convergence analysis was carried out for various initial particle spacing. The flow characteristics along the stepped spillway were
more sensitive to the particle spacing compared to the smooth chute one. In general, the numerical flow depths compared well with
the corresponding experimental data and with empirical formulas available in the literature. The velocity profiles, and the free-stream
velocity in particular, were also well reproduced by the SPH method; however, larger differences were obtained near the solid boundary.
DOI: 10.1061/(ASCE)HY.1943-7900.0001776. © 2020 American Society of Civil Engineers.
Author keywords: Nonaerated flow region; Skimming flow; Smooth spillway; Smoothed particle hydrodynamics; Stepped spillway.

Introduction experimental and numerical works have focused on this flow re-
gime. Various studies have addressed detailed air–water flow prop-
Stepped spillways have been increasingly used to construct new erties along the chute, namely on flat to moderate–sloped spillways
dams, as well as to rehabilitate embankment dams, especially to (i.e., chute angles from the horizontal lower than 22° and between
increase the spillway discharge capacity (Sorensen 1985; Frizell 22° and 30°, respectively), typical on embankment dams, such as
1992; McLean and Hansen 1993; Rice and Kadavy 1996; Ward (2002), Chanson and Toombes (2002a, b), Boes and Hager
McDonald and Curtis 1997; Chanson 2002; Hunt et al. 2008; (2003), André (2004), Felder and Chanson (2008, 2013), Bung
Matos and Meireles 2014; Frizell and Frizell 2015). Among their (2011), Takahashi and Ohtsu (2012), and Hunt et al. (2014), among
positive aspects, stepped spillways lead to an increased flow energy others. Experimental studies embracing the nonaerated flow region
dissipation due to the macroroughness created by the steps. De- were also conducted to date, namely by Gonzalez and Chanson
pending on the specific flow rate and geometric characteristics of (2007), Amador et al. (2009), Meireles and Matos (2009), Hunt
the chute, the flow over stepped spillways is usually classified as a and Kadavy (2010, 2011), Meireles et al. (2012), Hunt et al. (2014),
nappe, transition, or skimming flow. In the latter, the water flows as
and Zhang and Chanson (2016).
a coherent stream above the step edges, whereas the step cavities
Numerical studies of skimming flow over stepped spillways
are completely filled by the recirculating fluid (e.g., Essery and
have mainly focused on mesh-based methods, such as those by
Horner 1978; Chamani and Rajaratnam 1999; Matos 2001;
Chen et al. (2002), Bombardelli et al. (2011), Chinnarasri et al.
Chanson 2002; Ohtsu et al. 2004; Chanson et al. 2015).
(2014), Meireles et al. (2014), Lopes et al. (2017), and Toro et al.
Most stepped spillways have been designed to operate un-
(2017). The interest for meshless (or Lagrangian) methods, such as
der skimming flows, and consequently a significant number of
smoothed particle hydrodynamics (SPH) or moving particle semi-
1
Research Fellow, Dept. of Hydraulics and Sanitation, São Carlos implicit (MPS) has emerged in more recent years. Some general
School of Engineering, Univ. of São Paulo, São Carlos, SP 13566590, advantages of the SPH method over the mesh-based methods
Brazil (corresponding author). Email: junobreg@gmail.com are the effectiveness in solving complex fluid dynamic problems
2
Professor, Civil Engineering Research and Innovation for Sustainabil- with highly nonlinear deformations; the ability to deal with multi-
ity, Instituto Superior Técnico, Universidade de Lisboa, Lisbon 1049-001, phase flows and fluid-structure interactions; the natural tracking
Portugal. Email: jorge.matos@tecnico.ulisboa.pt of free surfaces and moving boundaries; and ease of numerical im-
3
Professor, Dept. of Hydraulics and Sanitation, São Carlos School of
plementation comparatively to mesh-based methods (Liu and Liu
Engineering, Univ. of São Paulo, São Carlos, SP 13566590, Brazil; Visiting
Professor, Dept. of Hydraulic and Environmental Engineering, Federal 2010; Violeau and Rogers 2016).
Univ. of Ceará, Fortaleza, CE 60451-970, Brazil. Email: harry.schulz@ A fair number of studies using the SPH method for modeling
pq.cnpq.br flows over spillways have been developed (e.g., Rebollo et al.
4
Researcher, Marine, Environment and Technology Center, Instituto 2010; Saunders et al. 2014; González-Cao et al. 2018, 2019;
Superior Técnico, Universidade de Lisboa, Lisbon 1049-001, Portugal. Moreira et al. 2019). Rebollo et al. (2010) simulated the flow over
ORCID: https://orcid.org/0000-0003-1780-6273. Email: ricardo.canelas@ a smooth gated spillway of the Alarcón Dam in Spain, using a tri-
tecnico.ulisboa.pt
dimensional SPH model and a particles feedback circuit to deter-
Note. This manuscript was submitted on March 7, 2019; approved on
February 5, 2020; published online on May 26, 2020. Discussion period
mine a range of points of the rating curve and to obtain pressure
open until October 26, 2020; separate discussions must be submitted for values on the spillway face, under different floodgate opening po-
individual papers. This paper is part of the Journal of Hydraulic Engineer- sitions. The analysis of the rating curve and pressures on the spill-
ing, © ASCE, ISSN 0733-9429. way face showed good agreement between physical and numerical

© ASCE 04020054-1 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


results, although significant differences on the pressures were i.e., increasing the ratio between the step height and the particle
obtained mainly in the vicinity of the spillway crest. spacing (h=dp) from 5.3 to 10.0. Comparisons between the nu-
Saunders et al. (2014) investigated the flow over a four-bay merical and experimental velocity profiles and pressures were also
gated spillway with a ski jump, focusing on the water depths at provided for one step position where the flow was still not strongly
the upstream reservoir. Six initial values of the spacing between affected by aeration. The greater differences for the simulated ve-
the SPH particles were tested. The accuracy of the results was locities were found for the higher discharges tested, and the veloc-
found to be dependent on the particle spacing, and the percentage ities were overestimated near the free surface.
differences between simulated and experimental data decreased Husain et al. (2014) performed numerical studies to simulate the
from 14.6% to 7.0% when reducing the initial particle spacing from flow over a stepped spillway with a broad crested weir. The numeri-
0.035 to 0.010 m. Relationships between discharge and water level cal model comprised a large upstream reservoir, the stepped spill-
at the upstream reservoir were also determined for several flow way and an open boundary at the downstream end. A sensitivity
rates, using the optimal particle spacing of 0.015 m (considering analysis was firstly performed to evaluate the effect of several par-
the computational efficiency and results accuracy). The comparison ticle spacing on the discharge prediction and a good optimization in
between simulated and experimental reservoir water depths showed terms of computation time and accuracy was achieved for the par-
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

that the maximum percentage difference of 11.5% occurred for the ticle spacing equal to 0.005 m. The results were validated through
highest simulated discharge at a location of 1.2 m upstream of the comparison between measured and simulated flow depths, velocity
spillway crest. A qualitative description of the flow down the spill- profiles, and inception point locations, resulting in overall relative
way ski jump and projected away was also presented by Saunders errors lower than 10%. An analysis of the normalized simulated
et al. (2014). A more complete analysis of the free trajectory jet pressure distributions on the horizontal and vertical step faces along
from a ski jump was developed by González-Cao et al. (2019). the nonaerated flow region were also presented, which were quali-
The general flow behavior obtained in their numerical simulations, tatively consistent to those reported in the literature.
as well as the maximum water elevations and the upper nappe Following the work of Husain et al. (2014), Gu et al. (2017)
profiles, were in accordance with the corresponding experimental adopted a similar model, consisting of an upstream tank and an
results. open boundary at the downstream end of the channel. To reduce
The work developed by González-Cao et al. (2018) aimed to the reservoir dimensions, the authors applied a push-paddle system
evaluate the hydrological safety of the Belesar Dam spillway in moving horizontally with a constant velocity. The simulations were
Spain under extreme overflow conditions. Two-dimensional simu- based on two physical models, and the study was mainly concerned
lations were first carried out for an ogee spillway and a broad with the velocity distribution, pressures on the step faces, and en-
crested weir, resulting in numerical free surface profiles similar to ergy dissipation. The pressure profiles on the horizontal step face
those obtained in the experimental tests. The flow behavior of the demonstrated an S-shape with positive values, while the general
real spillway was then simulated. The time series and maximum pressure pattern on the vertical step faces presented negative values
values of water depths and velocities were determined at several on the upper part and positive values on the lower part of the face.
distances from the crest on the right and left sides of the spillway. The average errors for pressures and energy dissipation were lower
The numerical velocities presented a reasonable agreement with than 4.7% and 8.2%, respectively.
those obtained from a theoretical approximation. The water depths The research developed by Wan et al. (2017a) aimed to inves-
were clearly asymmetric until the middle part of the spillway struc- tigate the flow hydrodynamics and the dissolved oxygen character-
ture and no overtopping of the lateral walls was observed, showing istics along a stepped spillway, using a multiphase SPH model.
a proper operation of the structure. Similarly, Wan et al. (2017b) used an improved multiphase SPH
Moreira et al. (2019) applied a single-phase SPH model to model to investigate the air concentration characteristics along a
simulate the flow in the energy dissipator of Crestuma Dam stepped spillway. The simulated inception point location of free
(gate-structure type followed by a stilling basin and a rock fill bed surface aeration agreed well with the experimental results in both
protection) and the complementary spillway of Caniçada Dam studies. In the study from Wan et al. (2017b), the simulated air con-
(gate-structure controlled by an ogee crest, followed by a tunnel centration profiles in different steps were consistent with the exper-
and a ski jump at the outlet), both located in Portugal. The general imental data, and the greatest discrepancies were found at the
accuracy of the SPH results in predicting the water depths, veloc- inception point location. According to the authors, the violent flow
ities, and pressures was satisfactory. The discrepancies between the deformation and break-up at this position probably influenced the
simulated and experimental water depths became significant in the experimental measurements or induced simulation inaccuracies.
final stretch of the Caniçada Dam spillway, which could be justi- The aim of the present study was to verify the applicability of
fied, to some extent, to the nonconsideration of air entrainment in the SPH method for simulating the flow over smooth and stepped
the SPH model. The pressures on the channel bottom and walls spillways using the high-performance DualSPHysics implementa-
were generally well captured, except at locations where subatmo- tion (Crespo et al. 2015), and to determine the main flow character-
spheric pressures were registered in the experiments. Other appli- istics of the nonaerated region (i.e., flow depths and velocity
cations to model dam flows using the SPH method can be found in profiles). The vorticity field in the skimming and recirculating flow
the state-of-the-art review of Moreira et al. (2020). regions was also evaluated.
Additionally, studies of SPH simulations of the flow over In previous studies, different approaches have been considered
stepped spillways were carried out by López et al. (2011), Husain for dealing with the inflow and outflow boundaries. In the present
et al. (2014), Gu et al. (2017), and Wan et al. (2017a, b), among study, a new functionality of the DualSPHysics code for a continu-
others. Different inflow and outflow conditions were considered in ous inlet condition was used to establish the flow rate at the up-
these studies. López et al. (2011) developed tridimensional numeri- stream end of the broad crested weir. A similar inlet condition
cal simulations for stepped spillways with lateral expansion. The was adopted by González-Cao et al. (2019) upstream of a ski jump
numerical model consisted of a recirculation system of the fluid to simulate the free trajectory jet of the projected flow. Using the
particles with a reservoir upstream of the spillway. The flow width continuous inlet condition, it is not necessary to consider a reservoir
occupying the lateral expansion was analyzed and the results were of large dimensions upstream of the spillway, enabling a reduction
considerably improved with reducing the initial particle spacing, in the numerical domain, and therefore the initial particle spacing.

© ASCE 04020054-2 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Schematic representations: (a) physical model, including locations along the broad crested weir and the spillway chute where the measure-
ments were taken; and (b) numerical model, with initial numerical conditions.

The lower the particle spacing, the higher the numerical resolution, (x ¼ 0.125, 0.250, 0.375, and 0.464 m), whereas the measurements
and more accurate results are expected; however, it requires greater along the spillway chute were performed at several locations de-
computational storage and longer simulation time. fined by the step edges of the stepped spillway [Fig. 1(a)]. Flow
The simulations were conducted using the same physical dimen- depths along the broad crested weir and the spillway chute were
sions and flow conditions of an experimental setup composed by a obtained using a 0.05 mm resolution point gauge, at the flume
26.6° (2H:1V) sloping spillway, with smooth or 2.5 cm high step- centerline, as well as by visual observation, using graduated rulers
chute inverts, and a relatively long broad crested weir. SPH simu- at the flume sidewalls. The velocities at the broad crested weir and
lation results are presented for a range of flow rates and compared the spillway chute were measured using a Prandtl-Pitot tube with an
with their experimental counterparts. Additionally, numerical pre- external diameter of 8 mm, connected to a manometer.
dictions of the vorticity in the skimming and recirculating flows are Additional velocity measurements were conducted at the broad
qualitatively compared with experimental results available in the crested weir with a total pressure (stagnation) probe, with a smaller
literature. external diameter (1 mm), also connected to a manometer. To es-
timate the velocity from the total pressure probe data, the pressure
distribution was assumed hydrostatic, which should not be fully
Methodology satisfied at position x = 0.125 m, and particularly at position
x = 0.464 m of the broad crested weir, due to flow curvature, as
Experimental Tests shown by static pressure measurements in identical or similar set-
ups (Cabrita 2007; Felder and Chanson 2012). Taking into account
The experimental tests were previously conducted at the Hydraulic that the effects resulting from the longitudinal separation between
and Water Resources Laboratory at the Instituto Superior Técnico the total and static tappings were not considered, the velocity es-
(IST), by André and Ramos (2003) and Cabrita (2007). The setup timation with the Prandtl-Pitot tube data in those sections, particu-
consisted of an 8.0 m long and 0.7 m wide rectangular flume with larly in cross section 4, is not accurate.
Plexiglas walls. The spillway model was installed in the flume On the stepped chute, air entrainment was observed for low flow
occupying its whole width, and the remaining system comprised rates (André and Ramos 2003; Cabrita 2007). Therein, the visually
the water supply and recirculation system, a reservoir and flow sta-
observed inception point location was defined as the vertical edge
bilizing structure at the entrance, and a flap gate at the downstream
immediately upstream of the step cavity where a continuous pres-
end of the flume. The inflow was controlled by a broad crested weir
ence of white water or air bubbles was noticed from above and also
which consisted of a 0.5 m high, 0.5 m long, and 0.7 m wide crest
through the Plexiglas sidewalls.
with a vertical upstream wall, and upstream rounded nose. The
Table 1 presents the main flow properties in cross section 1
stepped chute configuration was formed by twenty, 2.5 cm high
(x ¼ 0.125 m), namely the flow depth d, the mean flow velocity
steps, with the chute slope of 2H:1V (θ ¼ 26.6°). pffiffiffiffiffi
The flow rate was determined using an electromagnetic U, the Froude number (F ¼ U= gd) and the Reynolds number
flowmeter installed in the supply pipe. The mean and maximum (R ¼ q=ν), where g = gravity acceleration constant; q = unit
flow rate fluctuations were 0.4% and 1%, respectively (André and flow rate; and ν = kinematic viscosity of water (approximately
Ramos 2003). The accuracy of the electromagnetic flowmeter was 10−6 m2 s−1 ).
also tested with a Bazin weir temporarily installed at the down-
stream end of the flume. Maximum and mean absolute percentage
differences of the flow rate obtained from both measurement tech- Table 1. Main flow properties in cross section 1 of the broad crested weir
niques were lower than 2% and 0.9%, respectively.
Q (Ls−1 ) q (m2 s−1 ) d (m) U (ms−1 ) F R
The experimental procedure consisted of measuring the flow
depths and velocity profiles for different positions along the broad 35 0.050 0.0620 0.806 1.034 5.0 × 104
crested weir and the spillway chute for the flow rates Q ¼ 35, 42, 42 0.060 0.0729 0.823 0.973 6.0 × 104
49, and 56 Ls−1 (i.e., unit flow rates q ¼ 0.050, 0.060, 0.070, and 49 0.070 0.0841 0.832 0.916 7.0 × 104
0.080 m2 s−1 ). The measurements along the broad crested weir 56 0.080 0.0926 0.864 0.906 8.0 × 104
were taken at four positions from the upstream end of the crest Source: Data from André and Ramos (2003); Cabrita (2007).

© ASCE 04020054-3 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Numerical Method dua X p p

¼ − mb 2a þ 2b ∇a W ab þ g
In the SPH method, the flow is represented by a set of computa-
dt b
ρa ρb
X  
tional moving nodes to which fluid properties such as mass, veloc- 4νrab · ∇a W ab
ity, and pressure are associated, effectively making them material þ mb uab
b
ðρa þ ρb Þðjrab j2 þ 0.01h2k Þ
X τ a τ b 
particles. The particles move according to the evolution of their
properties, which are updated for each time step of the simulation þ mb 2 þ 2 · ∇a W ab ð7Þ
following fluid conservation laws, namely mass and momentum b
ρa ρb
(Liu and Liu 2003).
The continuity and momentum equations for incompressible where uab ¼ ua − ub ; rab ¼ ra − rb ; W ab ¼ Wðra − rb ; hk Þ;
fluids are expressed by Eqs. (1) and (2), whereas the pressure is cab ¼ ðca þ cb Þ=2 = average speed of sound; ∇a = gradient of the
explicitly calculated by an equation of state for a nearly compress- kernel function in respect to the coordinates of the particle
ible flow [Eq. (3)], which relates the pressure with the fluid density a; δ = coefficient of Delta-SPH function; ν = kinematic viscosity;
and τ = SPS stress tensor.

¼ −ρ∇ · u ð1Þ
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

The discrete equations are integrated using a time integration


dt scheme, which is subjected to the Courant-Friedrich-Levy (CFL)
condition for stability. The time step of the simulation should be
du 1
¼ − ∇p þ Γ þ g ð2Þ proportional to the kernel function length (hk ), taking into account
dt ρ the viscous dissipation and the external forces (Liu and Liu 2003).
 γ 
ρ0 c2s ρ
p¼ −1 ð3Þ Numerical Model
γ ρ0
The numerical model was developed and validated based on the
where t = time; ρ = fluid density; u = velocity vector; p = pressure; previously described experimental setup, considering a scale factor
Γ = dissipative terms; g = gravity acceleration; ρ0 ¼ 1,000 kgm−3 of 10:1 (numerical model:physical model) and Froude dynamic
is the reference density; cs = numerical speed of sound (≥10 times similitude between prototype (numerical model) and physical
the maximum flow velocity); and γ ¼ 7 is the exponent of the model. Similar scale factors were previously adopted by López et al.
equation of state. (2011) and Gu et al. (2017), namely 15:1 and 10:1, respectively.
The numerical discretization of the basic physical law equations Considering that the present study focuses on the nonaerated flow
in the SPH form involves the interpolation of a function F at a region and given the experimental flow conditions (Table 1), scale
particle “a,” considering the flow properties on the neighboring effects are expected to be negligible.
particles “b” over an influence domain. The interpolation is done The continuous inlet condition adopted in the simulations is one
using a smoothing function (also known as kernel function) as rep- of the features released on the DualSPHysics 4.2 version. Using
resented by Eq. (4) (Liu and Liu 2003). Eq. (5) describes the fifth this condition, fluid particle layers of constant height and uniform
order Wendland kernel function used in this study (Wendland velocity profiles were continuously added to the upstream section
1995) of the broad crested weir. The boundary conditions are enforced
X using buffer particles, and their physical information (velocity,
m
Fðra Þ ≈ Fðrb Þ b Wðra − rb ; hk Þ ð4Þ water depth, density, and pressure) is imposed or extrapolated from
b
ρb the fluid domain. Further details about the open boundary condition
implementation in the DualSPHysics software can be found in
8 4
> Tafuni et al. (2018).
< 1 − qr ð1 þ 2q Þ 0 ≤ q ≤ 2
r r The numerical simulations were developed for the stepped
Wðr; hk Þ ¼ αD 2 ð5Þ
>
:0 (h ¼ 2.5 cm) and smooth chutes for the flow rates of 35, 42,
qr > 2 49, and 56 Ls−1 . Six fluid layers with uniform velocity were dis-
placed at the beginning of the broad crested weir, and the total
where a, b = subscripts referring to particle a and the neighboring height of these layers was equal to the measured flow depth in cross
particles b; W = smoothing kernel function; F = function at a section 1, 0.125 m from the upstream end of the broad crested weir.
particle; m, ρ = mass and density of a particle, respectively; r = The velocity of the input layers was equal to the maximum velocity
position vector; hk = kernel function length; αD ¼ 7=4π · h2k in estimated from the total pressure probe measurements, considering
2D; αD ¼ 21=16π · h3k in 3D; and qr ¼ r=hk is the normalized the prototype scale and Froude similitude. The second set of fluid
particle spacing in relation to the kernel function length. particles was used to fill the area over the spillway as the initial
The conservation equations of the flow discretized using the condition [Fig. 1(b)]. The SPH numerical results at the broad
SPH operators are given by Eqs. (6) and (7). Eq. (6) represents crested weir were also compared with those obtained by Lúcio
the continuity equation with an additional term to reduce fluid (2015) for identical geometry, using the FLOW-3D software
density fluctuations, known as Delta-SPH term (Molteni and version 11.0.3.
Colagrossi 2009). Eq. (7) is the momentum equation, based on the The relevant parameters which are detailed on the DualSPHy-
subparticle scale (SPS) turbulence model that considers both sics software (Crespo et al. 2015), such as the numerical speed
the effects of laminar and turbulent viscosity (Gotoh et al. 2004; of sound (cs ¼ 100 ms−1 ), CFL coefficient (0.15), kernel function
Crespo et al. 2015) (Wendland function), time integration algorithm (Sympletic
scheme), among others, were adopted based on the experimental
dρa X X data of the stepped spillway for the inflow rate of 35 Ls−1 and
¼ mb uab · ∇a W ab þ 2δhk mb cab
dt b b
the initial particle spacing equal to 15.0 mm. The viscosity model
  that presented better results, regarding the filling of the steps with
ρ 1
× a−1 2 · ∇a W ab ð6Þ fluid particles when simulating stepped spillways, was the SPS tur-
ρb rab þ 0.01h2k bulence model [Eq. (7)] compared to the artificial viscosity model

© ASCE 04020054-4 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Table 2. Number of fluid particles at t ¼ 20 s, and simulation time for the The values of h=dp adopted in previous studies using the SPH
smooth and stepped spillways, for Q ¼ 35 Ls−1 : convergence analysis method for modeling the flow over stepped spillways are shown in
Smooth spillway Stepped spillway Table 3. The condition of continuous inflow fluid layers used in the
a present study enabled us to adopt a higher numerical resolution,
dp Simulation Simulation
(mm) N fluid (−) time, t (h) h=dp (−) N fluid (−) time, t (h)
that is, larger values of h=dp compared to those adopted in pre-
vious studies, applying either a particle recirculation system from
5.0 236,332 10.94 50.0 298,287 11.67 the outflow to the inflow boundaries, or a push-paddle system to
7.5 105,316 3.79 33.3 134,471 4.32 establish the simulated flow rate.
10.0 58,945 2.00 25.0 77,304 2.16
12.5 37,432 1.02 20.0 49,878 1.35 Flow Depths and Velocity Profiles along the Broad Crested
15.0 26,140 0.70 16.7 34,521 0.81
Weir
a
Prototype scale (10 times lower in the model scale). The flow depths along the broad crested weir upstream of the
smooth spillway and the experimental versus numerical flow
depths are shown in Figs. 2(a and b), respectively, for Q ¼ 35 Ls−1
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

proposed by Monaghan (1994). Complementary information can and dp values of 5.0, 7.5, 10.0, 12.5, and 15.0 mm (at prototype
be found in Crespo et al. (2015). A GPU Tesla K20C board, with scale). Similar results are expected for the stepped spillway because
13 multiprocessors (2,496 cores) and global memory of 5,061 MB of the identical entrance flow conditions. From Fig. 2, it can be
was used to run the simulations. concluded that the flow depths along the broad crested weir are
close to their experimental counterparts and practically independent
of dp; the relative differences between the flow depths considering
Results the highest and lowest particle resolution (dp ¼ 5 and 15 mm,
respectively) were lower than 0.7%. Small differences may be
The simulations ran for a total of 20 s, with output intervals of observed because the gap between the fluid particles and the solid
0.01 s. The flow was initially unstable over the crest and stabilized boundary is affected by the particle spacing due to the dynamic
approximately after 12 s of simulation time. Hence, flow velocities boundary condition (DBC) implemented in the DualSPHysics
and flow depths were calculated as time-averaged values from 12 software.
to 20 s. Similar conclusions were obtained for the velocity profiles
First, a systematic convergence analysis for several initial par- along the broad crested weir, as shown in Fig. 3, where H1 is
ticle spacing was carried out. Then, for the particle spacing consid- the upstream total head above the broad crested weir. These pro-
ered adequate to reproduce the main flow properties, the study was files are not significantly influenced by the particle spacing, with
extended to other discharges. relative differences (relative to the results for dp ¼ 5 mm) lower
than 2.2% for y > 0.006 m (i.e., y=H1 > 0.061), although slightly
larger deviations can be observed close to the invert. In cross sec-
Convergence Analysis of the Initial Particle Spacing tion 4 (x ¼ 0.464 m; x=Lcrest ¼ 0.93), in particular, the experi-
mental data are not accurate, because of the nonnegligible flow
In the SPH method, the fluid and boundaries are represented by a
curvature, and the hydrostatic pressure distribution hypothesis
set of particles equally distributed as the initial condition. To verify
for estimating the velocity from the total pressure probe data.
the adequate initial particle spacing (dp, or initial particle size), a Nevertheless, the SPH velocity profile in this cross section is sim-
convergence analysis was developed considering various particle ilar to that obtained by Lúcio (2015), using the FLOW-3D software
spacings, as presented in Table 2. In an analogy with the mesh- (Fig. 3), and follows the typical profile expected near the brink
based methods, the convergence of the results is attained with the of a broad crested weir, as obtained experimentally by Felder
increase of the numerical resolution (i.e., decreasing dp), until the and Chanson (2012).
results become virtually independent of the particle spacing.
The numerical simulations for the convergence analysis were Flow Depths along the Spillway Chute
carried out for the smaller experimental flow rate (Q ¼ 35 Ls−1 ), The flow depths along the smooth and stepped spillways are pre-
for both smooth and stepped spillway configurations. The velocity sented in Figs. 4 and 5 respectively, for Q ¼ 35 Ls−1 and dp values
and flow depth of the input particle layers were respectively of 5.0, 7.5, 10.0, 12.5, and 15.0 mm, along with the experimental
2.66 ms−1 and 0.620 m. The convergence of the results was evalu- data of André and Ramos (2003), acquired in the chute centerline.
ated by comparing the flow depths and the velocity profiles along The origin of the longitudinal axis (L) corresponds to the down-
the broad crested weir and the chute spillway, for various dp. stream end of the broad crested weir [Fig. 1(a)]. The flow depth
The simulation time and number of fluid particles (N fluid ) at the data at the right and left sidewalls along the stepped chute, obtained
final instant of simulation (t ¼ 20 s) for each value of dp (or by André and Ramos (2003) through visual observations, are also
h=dp, where h is the step height) are reported in Table 2. included in Fig. 5, as well as the inception point location given by

Table 3. Number of fluid particles and particle spacing applied in some previous studies
Reference dp (mm) h=dp (−) N fluid (−)
López et al. (2011) 120 (scale 15:1) 10.0 2,960,000
Husain et al. (2014) 3; 5; 7.5; 10.0 16.67; 10.0; 6.67; 5.0 621,332; 233,691; 100,820; 57,678
Gu et al. (2017) 4 12.5 162,098
Gu et al. (2017) 40 (scale 10:1) 7.5; 5.18; 3.75 300,000
Wan et al. (2017a) 8 12.5 12,238 (water); 32,524 (air)
Wan et al. (2017a) 4 8.25 15,738 (water); 34,024 (air)
Wan et al. (2017b) 7 14.29 18,916 (water); 37,161(air)

© ASCE 04020054-5 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Convergence analysis: (a) flow depths along the broad crested weir for 5 ≤ dpðmmÞ ≤ 15 and Q ¼ 35 Ls−1 ; and (b) experimental versus
numerical flow depths.

Fig. 3. Velocity profiles along the broad crested weir for 5 ≤


dpðmmÞ ≤ 15 and Q ¼ 35 Ls−1 : convergence analysis. Fig. 5. Flow depths along the stepped spillway for 5 ≤ dpðmmÞ ≤ 15
and Q ¼ 35 Ls−1 : convergence analysis.

The flow depth profiles along the smooth spillway were not very
sensitive to the variation of dp, since the mean absolute percentage
difference between the flow depths considering the highest and
lowest particle resolution (dp ¼ 5 and 15 mm, respectively) was
2.3%. The mean absolute percentage difference between the exper-
imental and the numerical flow depths for dp ¼ 5 mm was 8.4%
for the complete data set. The higher numerical flow depths ob-
tained at the initial spillway reach (L < 0.1 m) occurred due to the
formation of a cavity of air (i.e., in the absence of fluid particles)
below the jet formed in the transition from the broad crested weir to
the spillway.
Regarding the stepped spillway simulations, the flow depths
were more influenced by the particle resolution compared to the
Fig. 4. Flow depths along the smooth spillway for 5 ≤ dpðmmÞ ≤ 15 smooth condition, with differences decreasing for smaller values
and Q ¼ 35 Ls−1 : convergence analysis. of dp, until the convergence was reached. The mean absolute per-
centage differences were 10.2% for the flow depth profiles between
dp ¼ 5 mm and dp ¼ 15 mm; and 2.6% between dp ¼ 5 mm
and dp ¼ 7.5 mm. The mean absolute percentage difference be-
Li ¼ 0.780 m. In general, slightly higher flow depths were ob- tween the numerical and experimental data in the nonaerated flow
tained on the stepped chute sidewalls through visual observation, region, at the flume centerline, for dp ¼ 5 mm, was 8.2% for
as a result of free surface unsteadiness due to turbulence and wall 0 < LðmÞ < 0.8. A better approximation of the numerical profile
effects, in accordance with previous studies (Matos 1999, 2000; was observed for the experimental values measured at the flume
André and Ramos 2003). centerline, as expected, because of the observed wall effects.

© ASCE 04020054-6 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Velocity profiles along the smooth spillway for 5 ≤ dpðmmÞ ≤ 15 and Q ¼ 35 Ls−1 : convergence analysis.

It is worth noting that the free surface exhibited a wavy pattern depending on dp, and the differences tend to increase with the lon-
along the chute, particularly near the inception point, where it gitudinal distance. This may be explained by the decreasing flow
was more difficult to take accurate measurements of the flow depth depth and consequent increase of the relative step macroroughness
(Meireles and Matos 2009). down the slope. For example, for L ¼ 0.671 m, the mean absolute
percentage difference between the velocity profiles for dp ¼ 5
Velocity Profiles along the Spillway Chute and 15 mm was around 17%. Up to cross section 9 (L ¼ 0.447 m),
The numerical velocity profiles for the smooth and stepped the velocity profiles were well reproduced by the SPH method, and
spillways are shown in Figs. 6 and 7, respectively, for Q ¼ 35 Ls−1 the differences were generally smaller for the profiles obtained with
and dp values of 5.0, 7.5, 10.0, 12.5, and 15.0 mm, along with the dp ¼ 5.0 mm. It should be noted that the experimental velocity data
experimental data of André and Ramos (2003). is not expected to be accurate in the cross sections close to the
The free-stream velocity along the spillway was calculated ac-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi transition from the broad crest to the chute, because of the flow
cording to the ideal-flow theory as V max ¼ 2gðHmax − d · cos θÞ, curvature.
where g = gravity acceleration constant; Hmax = upstream total Overall, based on the particle resolution study, the convergence
head relative to the invert; d = flow depth; and θ = chute slope. of the results was attained for an initial particle spacing equal to
The flow velocity within the boundary layer is smaller than the or lower than dp ¼ 7.5 mm. Therefore, dp ¼ 7.5 mm was used to
free-stream velocity because of the friction losses (e.g., Chanson simulate the other flow conditions, for the flow rates of 42, 49, and
2002). The free-stream velocities were calculated using the exper- 56 Ls−1 . The number of fluid particles at t ¼ 20 s and the total
imental flow depths at the flume centerline, measured by André and simulation time are presented in Table 4.
Ramos (2003), and were represented by dashed lines in Figs. 6
and 7.
Broad Crested Weir
Regarding the smooth spillway, the velocity profiles were de-
pendent on the particle resolution in the vicinity of the crest The flow depths along the broad crested weir are presented in Fig. 8,
(L ≤ 0.224 m), close to the invert, as expected. A better agreement which comprises the experimental data of Cabrita (2007), the
with experimental profiles and free-stream velocity was obtained numerical profile obtained by Lúcio (2015) using FLOW-3D, and
for dp ¼ 5 mm. Further downstream, the velocity profiles were al- the results from the SPH simulations, using dp ¼ 7.5 mm. The
most independent of dp, although slightly larger deviations with flow depth profiles shown in Fig. 8 were generated from the smooth
the experimental profiles were observed near the invert. spillway simulation results.
The velocity profiles along the stepped spillway show that For a short initial reach of the weir (x ≤ 0.125 m), the
the differences among the numerical profiles can be considerable, free surface profiles obtained by Lúcio (2015) are expected to

© ASCE 04020054-7 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Velocity profiles along the stepped spillway (h ¼ 2.5 cm) for 5 ≤ dpðmmÞ ≤ 15 and Q ¼ 35 Ls−1 : convergence analysis.

Table 4. Number of fluid particles at t ¼ 20 s, and simulation time for the smooth and stepped spillways
Smooth spillway Stepped spillway
Q (Ls−1 ) q (m2 s−1 ) dc =ha (−) N fluid (−) Simulation time, t (h) N fluid (−) Simulation time, t (h)
35 0.050 2.537 105,316 3.79 134,471 4.32
42 0.060 2.865 129,179 4.28 159,978 4.99
49 0.070 3.175 145,837 4.70 175,252 5.44
56 0.080 3.470 159,217 4.99 186,009 5.73
a
Stepped spillway.

be more representative of the flow pattern over a broad crested weir near the invert, mainly for y ≤ 0.01 m, probably because of the
(see, e.g., Felder and Chanson 2012), since the SPH profile is ap- DBC implemented in the DualSPHysics software.
proximately constant in this region due to the upstream boundary
condition of six layers with uniform velocity profiles. However,
for the second half of the weir (x > 0.25 m), the flow depths are Smooth Spillway
generally well estimated by the SPH profile.
The largest relative differences between numerical (SPH) and Flow Depths along the Spillway Chute
experimental flow depths were observed in cross section 2, for Q ¼ The free surface profiles for the smooth spillway simulations us-
42 and 49 Ls−1 (approximately 7%). For the other sections, ing dp ¼ 7.5 mm (based on the convergence analysis) for the
the relative differences were lower than 5%. Such differences flow rates from 35 to 56 Ls−1 are shown in Fig. 9. The experi-
were of the same order of magnitude as those obtained by Lúcio mental data from André and Ramos (2003), measured at the flume
(2015). centerline, are also included in this figure. The experimental ver-
The velocity profiles along the broad crested weir for the flow sus numerical flow depths at the chute centerline are plotted in
rates 42, 49, and 56 Ls−1 (not shown herein) were similar to those Fig. 10.
presented in Fig. 3. The most significant differences between the In general, the numerical results were found to fit well to the
SPH results and the experimental velocity profiles were observed experimental data, with relative differences lower than 10%, and

© ASCE 04020054-8 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Fig. 8. Flow depths along the broad crested weir for various simulated
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

flow rates (dp ¼ 7.5 mm).

Fig. 10. Experimental (centerline data) versus numerical flow depths


along the smooth and stepped spillways.

presented in this figure were obtained by André and Ramos


(2003) at the flume centerline and sidewalls for Q ¼ 35 and
42 Ls−1 , whereas for Q ¼ 49 and 56 Ls−1 , the flow depths were
obtained only through visual observation from the sidewalls. The
visually determined inception point locations are also indicated in
Fig. 13, namely Li ¼ 0.78 m, for Q ¼ 35 Ls−1, and Li ¼ 0.89 m,
for Q ¼ 42 Ls−1. For the largest flow rates, the inception point oc-
curred approximately at the downstream end of the chute, or did not
take place in the chute, respectively, for Q ¼ 49 and 56 Ls−1 .
Fig. 9. Flow depths along the smooth spillway for various simulated
The comparison between experimental and numerical flow
flow rates (dp ¼ 7.5 mm).
depths at the stepped chute centerline is also presented in Fig. 10.
In general, the numerical results were found to fit well to the
an overall mean value of 5.3%. Because of the formation of a void experimental data, with relative differences lower than 10%, and
under the water jet immediately downstream of the broad crested an overall mean value of 6.3% (considering data for Q ¼ 35
weir (i.e., without fluid particles), the numerical results differed and 42 Ls−1 ). Data points indicating larger differences than 10%
considerably from the experimental counterparts for L < 0.1 m refer to two cross sections located in a reach with considerable
(i.e., corresponding to smooth chute data points outside the 10% free surface oscillations, where the experimental data may be
error interval, in Fig. 10), whereas they are quite close on the inaccurate.
remaining reach. The relationship for the clear-water depth (d=di ) presented by
Meireles and Matos (2009) for L=Li < 1, as well as the relationship
Velocity Profiles along the Spillway Chute (d=dc ) developed by Hunt et al. (2014) for a broad range of
The velocity profiles for each flow condition are presented in Fig. 11, step height to critical flow depth ratio and chute slopes, were ap-
whereas the instantaneous velocity contour field for the flow rate of plied herein to compute d=di or d=dc as a function of L=Li
35 Ls−1 (at model scale) is plotted in Fig. 12. For the flow rates of (for 0.1 ≤ L=Li ≤ 1.0). The properties at the inception point (loca-
35 and 42 Ls−1 , the SPH results are compared with the experimental tion Li and flow depth di ) were then calculated from the relation-
values obtained by André and Ramos (2003). Due to the absence of ships developed in the respective studies to subsequently compute
velocity data for the flow rates of 49 and 56 Ls−1 , the experimental
the values of d and L, as plotted in Fig. 13.
values obtained by Cabrita (2007) for a spillway with two converg-
In general, the numerical profiles were close to the results ob-
ing sidewalls of 9.9° were used. It should be noted that for this
tained both from Meireles and Matos (2009) and Hunt et al. (2014)
geometry and flow rates, the flow conditions in the chute centerline
for the identical chute slope. As expected, the flow depths mea-
were not influenced by the cross waves.
sured at the flume sidewalls through visual observations were gen-
The relative velocity differences were generally small, but close
erally higher than those acquired at the flume centerline, as well as
to the pseudobottom, varying on average from 1% to 5%, for
those estimated from Meireles and Matos (2009), and Hunt et al.
L ≥ 0.224 m, whereas larger values were obtained for L ¼
0.112 m, as a result of higher flow depths in such a transition region (2014), or obtained from the SPH simulations.
from the SPH simulations. The mean absolute percentage differences between the numeri-
cal flow depths and those estimated from the equation proposed by
Meireles and Matos (2009), for 0.1 ≤ LðmÞ ≤ 0.95, were 3.6%
Stepped Spillway (Q ¼ 35 Ls−1 ), 6.4% (Q ¼ 42 Ls−1 ), 3.5% (Q ¼ 49 Ls−1 ), and
7.1% (Q ¼ 56 Ls−1 ). It should be noted that the largest flow
Flow Depths along the Spillway Chute rate was outside of the range of data used for developing such
The free surface profiles for the stepped spillway simulations using an equation. In turn, the mean absolute percentage differences be-
dp ¼ 7.5 mm are shown in Fig. 13. The experimental data tween the numerical flow depths and those calculated from the

© ASCE 04020054-9 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Velocity profiles along the smooth spillway: (a) 35; (b) 42; (c) 49; and (d) 56 Ls−1 .

Fig. 12. Instantaneous velocity contour field on the smooth spillway for Q ¼ 35 Ls−1 and t ¼ 20 s.

Hunt et al. (2014) relationship, for identical chute slope, were θ ¼ 26.6° and dc =h ¼ 2.87. The chute slope has a nonnegligible
quite small: 1.6% (Q ¼ 35 Ls−1 ), 2.6% (Q ¼ 42 Ls−1 ), 1.8% effect on the normalized flow depths, as also shown in Hunt et al.
(Q ¼ 49 Ls−1 ), and 2.5% (Q ¼ 56 Ls−1 ). (2014). However, a larger data scatter can be observed in the down-
A comparison of the normalized flow depths (d=dc ) along the stream half of both chutes (L=Li > 0.6), due to increased free
chute (L=Li ) is shown in Fig. 14, based on the experimental data surface waviness. For the identical chute slope (θ ¼ 26.6°), the
from André and Ramos (2003), measured at the flume centerline, experimental data and the SPH results are in good agreement with
on the experimental clear-water flow depth obtained by Hunt et al. the values estimated from those equations, but for the initial reach,
(2014) for θ ¼ 18.4° and 1.67 ≤ dc =h ≤ 3.23, as well as the rela- and for few cross sections in the vicinity of the inception point lo-
tionships proposed by Meireles and Matos (2009) and their em- cation, where the flow depth data may be inaccurate. In general, the
pirical relationship to compute di , and by Hunt et al. (2014), for mean absolute percentage differences between the experimental

© ASCE 04020054-10 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Flow depths along the stepped spillway: (a) 35; (b) 42; (c) 49; and (d) 56 Ls−1 .

Fig. 14. Normalized flow depths along the stepped spillway. (Data of Hunt et al. 2014 obtained through digitalization.)

data and the relationship of Meireles and Matos (2009) or Hunt (for Q ¼ 49 and 56 Ls−1 ; experimental velocity data were not
et al. (2014), for the identical chute slope (θ ¼ 26.6°), were lower available). Fig. 15 shows that the maximum numerical velocities
than 5.6% (for 35 and 42 Ls−1 , and L=Li ≥ 0.1). The mean abso- were close to the free-stream velocities for an ideal-fluid flow,
lute percentage differences between the numerical results and the as one would expect in the developing flow region, upstream of the
values predicted by the Meireles and Matos (2009) and Hunt et al. inception point. The largest differences between the experimental
(2014) relationships, for all tested flow rates, were limited to 6.8% data and the SPH results were obtained for locations closer to the
and 2.6% (L=Li ≥ 0.1), respectively. upstream and downstream ends of the nonaerated flow reach, as
Overall, the numerical flow depth results of the present study well as in the vicinity of the pseudo-bottom (y ≤ 0.01 m), similarly
compared well with those estimated from the formulas presented as obtained on the smooth chute.
by Meireles and Matos (2009) and by Hunt et al. (2014) for the Considering the simulations for Q ¼ 35 and 42 Ls−1 , the mean
identical chute slope. absolute percentage velocity differences varied from 0.7% to 6.2%
for 0.224 ≤ LðmÞ ≤ 0.671 and vertical positions y ≥ 0.010 m.
Velocity and Vorticity Fields along the Spillway Chute Larger values were obtained for L ≤ 0.112 m as a result of
Velocity profiles along the nonaerated flow region of the stepped higher numerical flow depths in this transitional region, and for
spillway for the flow rates of 35 and 42 Ls−1 are shown in Fig. 15 L ≥ 0.783 m, closer to the inception point location.

© ASCE 04020054-11 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 15. Velocity profiles along the stepped spillway: (a) 35; and (b) 42 Ls−1 .

Fig. 16. Instantaneous velocity contour field on the stepped spillway for Q ¼ 56 Ls−1 and t ¼ 20 s.

The instantaneous velocity contour field is displayed in Fig. 16 are of the same order of magnitude as those obtained experimentally
for the flow rate of 56 Ls−1 (at model scale). The step cavities along by Amador (2005) and Amador et al. (2006) using a particle image
the chute were completely filled with fluid particles, with the ex- velocimetry (PIV) technique to characterize the nonaerated flow re-
ception of the first step. In the experimental tests, for such flow gion over a 51.3° (1V:0.8H) steeply sloping stepped spillway, for
conditions, all step cavities were completely filled by the recircu- dc =h ¼ 2.15, regardless of the dissimilar slope and normalized criti-
lating fluid. Moving voids were also observed as the flow evolved cal flow depth; they are also similar to those presented by Zhang
with the simulation time. Voids were possibly created at locations and Chanson (2018), using two local optical flow methods applied
where the free surface instantaneously rose because of the unsteadi- to an ultrahigh-speed camera data to visualize and analyze the
ness of the free surface, or as a result of local outflows from the step flow patterns in aerated skimming flows for a 45° (1V:1H) slope,
cavities to the main stream. with dc =h ¼ 0.9. The dominant vorticity reported in the study by
The instantaneous velocity field and the horizontal and vertical Amador et al. (2006) was 360 s−1 near the outer step edges, whereas
components of velocity (V x and V z ) at some steps are shown in a vorticity between 80 and 100 s−1 was observed inside the cavities.
Figs. 17(a–c), for Q ¼ 56 Ls−1, where Lcav is the step cavity The numerical results of Toro et al. (2016) for two-dimensional sim-
length, in the chute direction [Fig. 1(a)]. The flow pattern inside the ulations using OpenFOAM and two Reynolds stress models, for
step cavities suggests unstable vortices and a wake–wake interac- similar conditions to those of Amador (2005), led to a representative
tion in sequent steps, as described by Chanson (2002) and Ohtsu value of 300 s−1 for the vorticity peak, close to the outer step edges.
et al. (2004), for similar chute slopes.
Spanwise vorticity fields (ω) for Q ¼ 56 Ls−1 are presented in
Fig. 17(d). As expected, the magnitude of the recirculation was Conclusion
small in the skimming flow, whereas large vorticity values were ob-
tained in the shear layer that forms above the step tips, namely at the The SPH method was used for modeling the flow over spillways,
impact region of the flow on the outer edge of the horizontal step using the DualSPHysics software. The physical models to be
faces. For this flow rate (dc =h ¼ 3.47), the largest vorticities ex- simulated consisted of smooth and stepped spillways with a broad
ceeded 300 s−1 , particularly near the outer step edges. These values crested weir, and a slope of 2H:1V, typical of embankment dams.

© ASCE 04020054-12 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 17. Instantaneous contour fields on the stepped spillway for Q ¼ 56 Ls−1 and t ¼ 20 s: (a) velocity magnitude; (b) horizontal velocity
component V x ; (c) vertical velocity component V z ; and (d) spanwise vorticity.

The stepped geometry comprised twenty 2.5-cm-high steps. Using not observed in the experiments. Downstream of the initial reach,
continuous inflow fluid layers at the upstream end of the numeri- the relative differences between numerical and experimental results
cal domain allowed to increase the particle spacing resolution were fairly small in the studied nonaerated flow region.
compared to other studies, with reasonably limited simulation The numerical velocity profiles along the smooth and stepped
time. spillways were also found to agree well with the experimental data,
First, a convergence analysis was carried out for the lower except in the initial short reach and when approaching the point of
specific flow rate and for both smooth and stepped spillways. Five inception, as well as close to the pseudobottom. The numerical
values of initial particle spacing (dp, or particle size) were inves- free-stream velocities agreed quite well with those obtained from
tigated (from 5 to 15 mm, at 10:1 prototype scale simulations). For the ideal-flow theory for maximum velocities, using the experimen-
the stepped spillway, the ratio between the step height and particle tal flow depth data. SPH spanwise vorticity fields were comparable
spacing (h=dp) varied from 17 to 50. The numerical flow depths to those obtained experimentally by other researchers.
were found to vary with dp, and the resulting differences among
the flow depths were more significant in the stepped spillway. The
convergence analysis also revealed that the flow depth and velocity Data Availability Statement
profiles along the smooth spillway were almost independent of dp,
whereas those for the stepped spillway could vary considerably Some or all data, models, or code generated or used during the
with the adopted initial particle spacing. study are available in a repository online in accordance with funder
In general, the flow depths along the smooth and stepped spill- data retention policies (www.dual.sphysics.org).
ways were well predicted by the numerical simulations. The largest Some or all data, models, or code that support the findings
differences were observed in the initial short reach of the spillway of this study are available from the corresponding author upon
because of the flow separation under the water jet, in the transition reasonable request (experimental and numerical flow depths and
region from the broad crested weir to the spillway chute, which was velocities).

© ASCE 04020054-13 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Acknowledgments δ = coefficient of the Delta-SPH function for the continuity
equation;
The authors acknowledge the support of the Portuguese Foundation θ = spillway slope;
for Science and Technology (FCT), through the RECI/ECM-HID/
ν = kinematic viscosity;
0371/2012 Project, and Prof. Rui Ferreira in particular, for provid-
ρ = density;
ing the computational server to run the simulations. Thanks also are
extended to Luís Palma Mendes for his assistance. The first author ρ0 = reference density;
acknowledges the scholarships from the Capes, CNPq and Erasmus ω = spanwise vorticity; and
Mundus-Smart2 Project. The third author acknowledges CNPq τ = stress tensor.
Project 307105/2015-6.
Subscript

Notation a = reference particle;


b = neighboring particle; and
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

The following symbols are used in this paper: ab = values between particle a and b.
c = speed of sound of a particle;
cs = numerical speed of sound;
d = flow depth; References
dc = critical flow depth;
Amador, A., M. Sánchez-Juny, and J. Dolz. 2006. “Characterization of the
di = flow depth at the inception point; nonaerated flow region in a stepped spillway by PIV.” J. Fluids Eng.-
dp = initial particle spacing or initial particle size; Trans. ASME 128 (6): 1266–1273. https://doi.org/10.1115/1.2354529.
F = continuous function; Amador, A., M. Sánchez-Juny, and J. Dolz. 2009. “Developing flow region
F = Froude number; and pressure fluctuations on steeply sloping stepped spillways.” J. Hy-
g = gravity acceleration; draul. Eng. 135 (12): 1092–1100. https://doi.org/10.1061/(ASCE)HY
.1943-7900.0000118.
Hmax = upstream total head relative to the chute invert; Amador, A. T. 2005. “Comportamiento hidráulico de los aliviaderos esca-
H1 = upstream total head above the broad crested weir; lonados en presas de hormigón compactado.” [In Spanish.] Ph.D. thesis,
h = step height; Dept. de Ingeniería Hidráulica, Marítima y Ambiental, Universitat
hk = kernel function length; Politècnica de Catalunya.
L = longitudinal position in the chute direction, originating at André, M., and P. Ramos. 2003. Hidráulica de descarregadores de cheia
the downstream end of the broad crested weir;pffiffiffiffiffiffiffiffiffiffiffiffiffiffi em degraus: Aplicação a descarregadores com paredes convergentes.
[In Portuguese.] Graduate Research Rep. Lisbon, Portugal: Instituto
Lcav = step cavity length in the chute direction: Lcav ¼ l2 þ h2 ; Superior Técnico, Universidade de Lisboa.
Lcrest = length of the broad crested weir; André, S. 2004. “High velocity aerated flows on stepped chutes with
Li = longitudinal position of the inception point, originating at macro-roughness elements.” Ph.D. thesis, École Polytechnique
the downstream end of the broad crested weir; Fédérale de Lausanne.
l = horizontal step length; Boes, R. M., and W. H. Hager. 2003. “Two-phase flow characteristics of
stepped spillways.” J. Hydraul. Eng. 129 (9): 661–670. https://doi.org
m = particle mass;
/10.1061/(ASCE)0733-9429(2003)129:9(661).
N fluid = number of fluid particles; Bombardelli, F. A., I. Meireles, and J. Matos. 2011. “Laboratory measure-
p = pressure; ments and multi-block numerical simulations of the mean flow and tur-
Q = flow rate; bulence in the non-aerated skimming flow region of steep stepped
q = unit flow rate; spillways.” Environ. Fluid Mech. 11 (3): 263–288. https://doi.org/10
.1007/s10652-010-9188-6.
qr = normalized particle spacing in relation to the kernel
Bung, D. B. 2011. “Developing flow in skimming flow regime on embank-
function length: qr ¼ r=hk ; ment stepped spillways.” J. Hydraul. Res. 49 (5): 639–648. https://doi
R = Reynolds number; .org/10.1080/00221686.2011.584372.
r = position vector; Cabrita, J. 2007. “Caracterização do escoamento deslizante sobre turbil-
t = time; hões em descarregadores de cheias em degraus com paredes conver-
U = mean flow velocity; gentes.” [In Portuguese.] M.Sc. thesis, Instituto Superior Técnico,
Universidade de Lisboa.
u = velocity vector; Chamani, M. R., and N. Rajaratnam. 1999. “Characteristics of skimming
V = flow velocity; pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi flow over stepped spillways.” J. Hydraul. Eng. 125 (4): 361–368.
V max = free-stream velocity: V max ¼ 2gðHmax − d · cos θÞ; https://doi.org/10.1061/(ASCE)0733-9429(1999)125:4(361).
V x = horizontal velocity component; Chanson, H. 2002. The hydraulics of stepped chutes and spillways. Lisse,
V z = vertical velocity component; Netherlands: A.A. Balkema.
Chanson, H., D. B. Bung, and J. Matos. 2015. “Stepped spillways and
W = smoothing kernel function;
cascades.” In Energy dissipation in hydraulic structures, IAHR Mono-
x = horizontal coordinate originating at the upstream end of graph, edited by H. Chanson, 45–64. Leiden, Netherlands: CRC Press,
the broad crested weir; Taylor & Francis Group.
y = transverse coordinate originating at the bottom or the Chanson, H., and L. Toombes. 2002a. “Air-water flows down stepped
pseudo-bottom; chutes: Turbulence and flow structure observations.” Int. J. Multiphase
z = vertical coordinate; Flow 28 (11): 1737–1761. https://doi.org/10.1016/S0301-9322(02)
00089-7.
αD = coefficient of the Wendland kernel function:
Chanson, H., and L. Toombes. 2002b. “Energy dissipation and air entrain-
αD ¼ 7=4π · h2k in 2D, αD ¼ 21=16π · h3k in 3D; ment in stepped storm waterway: Experimental study.” J. Irrig. Drain.
Γ = dissipative terms of the momentum equation; Eng. 128 (5): 305–315. https://doi.org/10.1061/(ASCE)0733-9437
γ = exponent of the equation of state; (2002)128:5(305).

© ASCE 04020054-14 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Chen, Q., G. Dai, and H. Liu. 2002. “Volume of fluid model for turbulence Liu, G. R., and M. B. Liu. 2003. Smoothed particle hydrodynamics: A
numerical simulation of stepped spillway overflow.” J. Hydraul. Eng. meshfree particle method. Singapore: World Scientific Publishing.
128 (7): 683–688. https://doi.org/10.1061/(ASCE)0733-9429(2002) Liu, M. B., and G. R. Liu. 2010. “Smoothed particle hydrodynamics (SPH):
128:7(683). An overview and recent developments.” Arch. Comput. Methods Eng.
Chinnarasri, C., D. Kositgittiwong, and P. Y. Julien. 2014. “Model of flow 17 (1): 25–76. https://doi.org/10.1007/s11831-010-9040-7.
over spillways by computational fluid dynamics.” Proc. Inst. Civ. Eng.- Lopes, P., J. Leandro, R. F. Carvalho, and D. B. Bung. 2017. “Alternating
Water Manage. 167 (3): 164–175. https://doi.org/10.1680/wama.12 skimming flow over a stepped spillway.” Environ. Fluid Mech. 17 (2):
.00034. 303–322. https://doi.org/10.1007/s10652-016-9484-x.
Crespo, A. J. C., J. M. Domínguez, B. D. Rogers, M. Gómez-Gesteira, S. López, D., M. de Blas, R. Marivela, J. J. Rebollo, R. Díaz, M. Sánchez-
Longshaw, R. Canelas, R. Vacondio, A. Barreiro, and O. García-Feal. Juny, and S. Strella. 2011. “Estudio hidrodinámico de vertederos y
2015. “DualSPHysics: Open-source parallel CFD solver based on rápidas escalonadas con modelo numérico tridimensional SPH:
Smoothed Particle Hydrodynamics (SPH).” Comput. Phys. Commun. Proyecto ALIVESCA.” [In Spanish.] In Proc., II Jornadas de Ingen-
187 (Feb): 204–216. https://doi.org/10.1016/j.cpc.2014.10.004. iería del Agua: Modelos Numéricos en Dinámica Fluvial, 1–10.
Essery, I. T. S., and M. W. Horner. 1978. The hydraulic design of stepped Barcelona: FLUMEN.
spillways. Rep. No. 33. London: Construction Industry Research and Lúcio, I. 2015. “Modelação numérica do escoamento deslizante sobre tur-
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Information Association. bilhões em descarregadores de cheias em degraus: Aplicação a peque-


Felder, S., and H. Chanson. 2008. Turbulence and turbulent length and nas barragens de aterro.” [In Portuguese.] M.Sc. thesis, Instituto
time scales in skimming flows on a stepped spillway. Dynamics simi- Superior Técnico, Universidade de Lisboa.
larity, physical modelling and scale effects: Hydraulic model report Matos, J. 1999. “Emulsionamento de ar e dissipação de energia do
CH series. Rep. No. CH64/07. Brisbane, Australia: Div. of Civil escoamento em descarregadores em degraus.” [In Portuguese.] Ph.D.
Engineering, Univ. of Queensland. thesis, Instituto Superior Técnico, Universidade de Lisboa.
Felder, S., and H. Chanson. 2012. “Free-surface profiles, velocity and pres- Matos, J. 2000. “Hydraulic design of stepped spillways over RCC dams.”
sure distributions on a broad-crested weir: A physical study.” J. Irrig. In Proc., 1st Int. Workshop on Hydraulics of Stepped Spillways, edited
Drain. Eng. 138 (12): 1068–1074. https://doi.org/10.1061/(ASCE)IR by H.-E. Minor and W. H. Hager, 187–194. Rotterdam, Netherlands:
.1943-4774.0000515. A.A. Balkema.
Felder, S., and H. Chanson. 2013. “Aeration, flow instabilities, and residual Matos, J. 2001. “Discussion of ‘Onset of skimming flow on stepped spill-
energy on pooled stepped spillways of embankment dams.” J. Irrig. ways’ by M. R. Chamani and N. Rajaratnam.” J. Hydraul. Eng. 127 (6):
Drain. Eng. 139 (10): 880–887. https://doi.org/10.1061/(ASCE)IR 519–525. https://doi.org/10.1061/(ASCE)0733-9429(2001)127:6(519).
.1943-4774.0000627. Matos, J., and I. Meireles. 2014. “Hydraulics of stepped weirs and dam
Frizell, K. H. 1992. “Hydraulics of stepped spillways for RCC dams and spillways: Engineering challenges, labyrinths of research.” In Proc.,
dam rehabilitations.” In Proc., 3rd Specialty Conf. on Roller Compacted 5th IAHR Int. Symp. on Hydraulic Structures, edited by H. Chanson and
Concrete, 423–439. New York: ASCE. L. Toombes, 1–30. Brisbane: Univ. of Queensland.
Frizell, K. W., and K. H. Frizell. 2015. Guidelines for hydraulic design McDonald, J. E., and N. F. Curtis. 1997. Applications of roller-compacted
of stepped spillways. Hydraulic Laboratory Rep. No. HL-2015-06. concrete in rehabilitation and replacement of hydraulic structures.
Denver: Bureau of Reclamation. Rep. No. REMR-CS-53. Washington, DC: USACE.
Gonzalez, C., and H. Chanson. 2007. “Hydraulic design of stepped McLean, F. G., and K. D. Hansen. 1993. “Roller compacted concrete for
spillways and downstream energy dissipators for embankment dams.” embankment overtopping protection.” In Proc., Specialty Conf., Geo-
Dam Eng. 17 (4): 223–244. technical Practice in Dam Rehabilitation. New York: ASCE.
González-Cao, J., O. García-Feal, J. M. Domínguez, A. J. C. Crespo, and Meireles, I., F. A. Bombardelli, and J. Matos. 2014. “Air entrainment onset
M. Gómez-Gesteira. 2018. “Analysis of the hydrological safety of dams in skimming flows on steep stepped spillways: An analysis.” J. Hy-
combining two numerical tools: Iber and DualSPHysics.” J. Hydrodyn. draul. Res. 52 (3): 375–385. https://doi.org/10.1080/00221686.2013
30 (1): 87–94. https://doi.org/10.1007/s42241-018-0009-6. .878401.
González-Cao, J., O. García-Feal, J. M. Domínguez, A. J. C. Crespo, and Meireles, I., and J. Matos. 2009. “Skimming flow in the nonaerated region
M. Gómez-Gesteira. 2019. “Numerical analysis of ski jumps using of stepped spillways over embankment dams.” J. Hydraul. Eng. 135 (8):
DualSPHysics.” In Proc., 14th Int. SPHERIC Workshop, 308–312. 685–689. https://doi.org/10.1061/(ASCE)HY.1943-7900.0000047.
Exeter, England: Univ. of Exeter. Meireles, I., F. Renna, J. Matos, and F. Bombardelli. 2012. “Skimming,
Gotoh, H., S. Shao, and T. Memita. 2004. “SPH-LES model for nonaerated flow on stepped spillways over roller compacted concrete
numerical investigation of wave interaction with partially immersed dams.” J. Hydraul. Eng. 138 (10): 870–877. https://doi.org/10.1061
breakwater.” Coastal Eng. J. 46 (1): 39–63. https://doi.org/10.1142 /(ASCE)HY.1943-7900.0000591.
/S0578563404000872. Molteni, D., and A. Colagrossi. 2009. “A simple procedure to improve the
Gu, S., L. Ren, X. Wang, H. Xie, Y. Huang, J. Wei, and S. Shao. 2017. pressure evaluation in hydrodynamic context using the SPH.” Comput.
“SPHysics simulation of experimental spillway hydraulics.” Water Phys. Commun. 180 (6): 861–872. https://doi.org/10.1016/j.cpc.2008
9 (12): 973. https://doi.org/10.3390/w9120973. .12.004.
Hunt, S. L., and K. C. Kadavy. 2010. “Energy dissipation on flat-sloped Monaghan, J. J. 1994. “Simulating free surface flows with SPH.” J. Com-
stepped spillways: Part 1. Upstream of the inception point.” Trans. put. Phys. 110 (2): 399–406. https://doi.org/10.1006/jcph.1994.1034.
ASABE 53 (1): 103–109. https://doi.org/10.13031/2013.29506. Moreira, A., A. Leroy, D. Violeau, and F. Taveira-Pinto. 2019. “Dam spill-
Hunt, S. L., and K. C. Kadavy. 2011. “Inception point relationship for flat- ways and the SPH method: Two case studies in Portugal.” J. Appl.
sloped stepped spillways.” J. Hydraul. Eng. 137 (2): 262–266. https:// Water Eng. Res. 7 (3): 228–245. https://doi.org/10.1080/23249676
doi.org/10.1061/(ASCE)HY.1943-7900.0000297. .2019.1611496.
Hunt, S. L., K. C. Kadavy, and G. J. Hanson. 2014. “Simplistic design Moreira, A. B., A. Leroy, D. Violeau, and F. A. Taveira-Pinto. 2020. “Over-
methods for moderate-sloped stepped chutes.” J. Hydraul. Eng. view of large-scale smoothed particle hydrodynamics modeling of dam
140 (12): 04014062. https://doi.org/10.1061/(ASCE)HY.1943-7900 hydraulics.” J. Hydraul. Eng. 146 (2): 03119001. https://doi.org/10
.0000938. .1061/(ASCE)HY.1943-7900.0001658.
Hunt, S. L., D. Reep, and K. C. Kadavy. 2008. “RCC stepped spillways for Ohtsu, I., Y. Yasuda, and M. Takahashi. 2004. “Flow characteristics
Renwick Dam—A partnership in research and design.” Dam Saf. J. of skimming flows in stepped channels.” J. Hydraul. Eng. 130 (9):
6 (2): 32–40. 860–869. https://doi.org/10.1061/(ASCE)0733-9429(2004)130:9(860).
Husain, S. M., J. R. Muhammed, H. U. Karunarathna, and D. E. Reeve. Rebollo, J. J., M. de Blas, D. López, R. Díaz, and R. Marivela. 2010.
2014. “Investigation of pressure variations over stepped spillways using “Hydrodynamic verification with SPH of under gate flow in the Alarcón
smooth particle hydrodynamics.” Adv. Water Resour. 66 (Apr): 52–69. spillway (Spain).” In Proc., 1st European IAHR Congress, 1–9.
https://doi.org/10.1016/j.advwatres.2013.11.013. Edinburgh, UK: School of the Built Environment, Heriot-Watt Univ.

© ASCE 04020054-15 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054


Rice, C. E., and K. C. Kadavy. 1996. “Model study of a roller compacted Fluid Mech. 16 (6): 1195–1221. https://doi.org/10.1007/s10652-016
concrete stepped spillway.” J. Hydraul. Eng. 122 (6): 292–297. https:// -9472-1.
doi.org/10.1061/(ASCE)0733-9429(1996)122:6(292). Violeau, D., and B. D. Rogers. 2016. “Smoothed particle hydrodynamics
Saunders, K., M. Prakash, P. W. Cleary, and M. Cordell. 2014. “Application (SPH) for free-surface flows: Past, present and future.” J. Hydraul. Res.
of smoothed particle hydrodynamics for modelling gated spillway 54 (1): 1–26. https://doi.org/10.1080/00221686.2015.1119209.
flows.” Appl. Math. Modell. 38 (17-18): 4308–4322. https://doi.org/10 Wan, H., R. Li, C. Gualtieri, H. Yang, and J. Feng. 2017a. “Numerical
.1016/j.apm.2014.05.008. simulation of hydrodynamics and reaeration over a stepped spillway
Sorensen, R. M. 1985. “Stepped spillway hydraulic model investigation.” by the SPH method.” Water 9 (8): 565. https://doi.org/10.3390
J. Hydraul. Eng. 111 (12): 1461–1472. https://doi.org/10.1061/(ASCE) /w9080565.
0733-9429(1985)111:12(1461). Wan, H., R. Li, X. Pu, H. Zhang, and J. Feng. 2017b. “Numerical simu-
lation for the air entrainment of aerated flow with an improved multi-
Tafuni, A., J. M. Domínguez, R. Vacondio, and A. J. C. Crespo. 2018. “A
phase SPH model.” Int. J. Comput. Fluid Dyn. 31 (10): 435–449.
versatile algorithm for the treatment of open boundary conditions in
https://doi.org/10.1080/10618562.2017.1420175.
smoothed particle hydrodynamics GPU models.” Comput. Methods
Ward, J. 2002. “Hydraulic design of stepped spillways.” Ph.D. thesis, Dept.
Appl. Mech. Eng. 342 (Dec): 604–624. https://doi.org/10.1016/j.cma
of Civil Engineering, Colorado State Univ.
.2018.08.004.
Downloaded from ascelibrary.org by NUS-Central Library on 05/26/20. Copyright ASCE. For personal use only; all rights reserved.

Wendland, H. 1995. “Piecewise polynomial, positive definite and com-


Takahashi, M., and I. Ohtsu. 2012. “Aerated flow characteristics of skim- pactly supported radial functions of minimal degree.” Adv. Comput.
ming flow over stepped chutes.” J. Hydraul. Res. 50 (4): 427–434. Math. 4 (1): 389–396. https://doi.org/10.1007/BF02123482.
https://doi.org/10.1080/00221686.2012.702859. Zhang, G., and H. Chanson. 2016. “Hydraulics of the developing flow re-
Toro, J. P., F. A. Bombardelli, and J. Paik. 2017. “Detached eddy simulation gion of stepped spillways. II: Pressure and velocity fields.” J. Hydraul.
of the nonaerated skimming flow over a stepped spillway.” J. Hydraul. Eng. 142 (7): 04016016. https://doi.org/10.1061/(ASCE)HY.1943-7900
Eng. 143 (9): 04017032. https://doi.org/10.1061/(ASCE)HY.1943-7900 .0001136.
.0001322. Zhang, G., and H. Chanson. 2018. “Application of local optical flow
Toro, J. P., F. A. Bombardelli, J. Paik, I. Meireles, and A. Amador. methods to high-velocity free-surface flows: Validation and application
2016. “Characterization of turbulence statistics on the non-aerated to stepped chutes.” Exp. Therm Fluid Sci. 90 (Jan): 186–199. https://doi
skimming flow over stepped spillways: A numerical study.” Environ. .org/10.1016/j.expthermflusci.2017.09.010.

© ASCE 04020054-16 J. Hydraul. Eng.

J. Hydraul. Eng., 2020, 146(8): 04020054

Das könnte Ihnen auch gefallen