You are on page 1of 17

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301758546

Understanding physical clogging in drip irrigation: in situ, in-lab and


numerical approaches

Article  in  Irrigation Science · April 2016


DOI: 10.1007/s00271-016-0506-8

CITATIONS READS

12 393

7 authors, including:

Séverine Tomas Jerome Labille


National Research Institute of Science and Technology for Environment and Agric… Centre Européen de Recherche et d’Enseignement des Géosciences de l’Environn…
36 PUBLICATIONS   137 CITATIONS    80 PUBLICATIONS   2,043 CITATIONS   

SEE PROFILE SEE PROFILE

B. Molle
National Research Institute of Science and Technology for Environment and Agric…
45 PUBLICATIONS   209 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Euroclay2019 Session B2: Role of clays & modified clays in remediating environmental pollutants View project

Eco-SUN - Ecodesign of Sunscreen Using titanium dioxide Nanoparticles View project

All content following this page was uploaded by Séverine Tomas on 10 April 2017.

The user has requested enhancement of the downloaded file.


Irrig Sci (2016) 34:327–342
DOI 10.1007/s00271-016-0506-8

REVIEW

Understanding physical clogging in drip irrigation: in situ, in‑lab


and numerical approaches
Salim Bounoua1 · Séverine Tomas1 · Jérôme Labille2 · Bruno Molle3 ·
Jacques Granier3 · Pierre Haldenwang4 · Surani Nuur Izzati1 

Received: 27 November 2015 / Accepted: 19 April 2016 / Published online: 30 April 2016
© Springer-Verlag Berlin Heidelberg 2016

Abstract  Dripper clogging is a major drawback of micro- of the fluid velocity fields inside the dripper labyrinth
irrigation systems that must be addressed to improve their channel.
efficiency and durability. Particle-induced clogging is first
studied in situ. The experiments consist in observing in real
conditions the behavior of a series of drippers fitted on an Introduction
agricultural plot in the south of France. The plot is supplied
from a canal with Durance River water. The latter is loaded In microirrigation, clogging of drippers is one of the main
with sediments that gradually clog drippers and filters. problems which seriously affects water distribution uni-
Water analysis reveal that physicochemical clogging pre- formity at plot scale (Pitts et al. 1990) as well as irrigation
vails over biological clogging. This characterization helps system durability. Clogging can be of biological, chemi-
in setting in-lab experiment protocol. Indeed, besides field cal or physical origin and is often a combination of two or
observation of clogging, laboratory analyses of both the more of these factors (Gilbert et al. 1981; Adin and Sacks
irrigation water and the clogging material are performed 1991). Biological clogging is caused by the organic mat-
with reactive and inert clay: smectite and an illite–calcite ter transported in waters, sometime including the devel-
mix. A surprising tendency is observed: Salt concentration opment of algae, bacteria or fungi. Algae growth requires
in smectite seeded water decreases the clogging, whereas light, which means they generally develop at dripper out-
it increases agglomerate size. Computational fluid dynamic let. Bacteria proliferation can trap some solutes such as
simulations are carried out to investigate the impact of par- calcium bicarbonates or mineral particles in suspension
ticles on flow behavior. Results demonstrate that clay parti- in irrigation water (Gilbert et al. 1981). In certain condi-
cles interacting with the flow govern the complex structure tions, aquatic organisms may proliferate throughout the
irrigation network (Nakayama and Bucks 1991; Ravina
et al. 1992). Clogging often stems from the association of
Communicated by N. Lazarovitch.
different factors such as irrigation water quality or more
specifically its physicochemical composition. Clogging
* Séverine Tomas may also be of chemical origin if solutes precipitate due
severine.tomas@irstea.fr to changes in physicochemical conditions. Such phenom-
1 ena can occur in surface as well as underground waters,
UMR G‑EAU, IRSTEA Montpellier, 361 rue Jean‑François
Breton, BP 5095, 34196 Montpellier Cedex 5, France and most are observed with calcium carbonates, iron and
2 sometimes sulfate precipitates (Hills et al. 1989; Levy et al.
CEREGE UMR 6635, CNRS, Aix-Marseille Universit,
13545 Aix‑en‑Provence, France 2011). Finally, physical clogging is often due to suspended
3 mineral particles, such as clay and silt. These particles are
UMR G‑EAU, IRSTEA Montpellier, 361 rue Jean‑François
Breton, BP 5095, 34196 Montpellier Cedex 5, France generally too small to be retained by the filter, but they can
4 clog the emitters because of aggregation downstream from
FRE 2405 du CNRS, IMT/La Jetée/L3M, Modélisation et
simulation numérique en mécanique, 38, avenue Joliot Curie, the filtration system. This agglomeration favors the deposit
13451 Marseille Cedex 20, France process and results in clogging (Pitts et al. 1990).

13

328 Irrig Sci (2016) 34:327–342

In addition, the different impacts that particles can where q is the flow rate of emitter, K is the constant of pro-
have on flow depend on various factors such as particle portionality that depends of used units, P is the pressure
characteristics, pipe geometry, hydrodynamics and turbu- head and x is the emitter discharge exponent that character-
lence properties to name but a few. Adding particles can izes each emitter. The value of this exponent depends on
result in different phenomena depending on their charac- the conception of the emitter. For this study, we determine
teristics. Fluid turbulence intensity is reduced by small K = 2.12 and x = 0.57.
particles while it increases for large particles (Kulick
et al. 1994). They also observe in keeping with the In‑field experiments
simulation of Vreman (2007) that turbulence intensities
decrease when the mass load ratio increases. Yuan and The experimental setup is installed on a potato plot in
Michaenidies (1992) study turbulence intensity modifica- southeast France (Vaucluse, Pertuis) irrigated with water
tions due to particles in the pipe flow. They conclude that from the river Durance, which flows from the Alps. It is
turbulence reduction is caused by energy dissipation due comprised of four drippers lines placed on the ground sur-
to particle acceleration, while turbulence augmentation is face which are operated one hour per day from five to six
the result of flow disturbance due to particle motion and pm. Two lines are supplied with raw water and the two
particle wakes. others with filtered water. Filtration is performed using
The main objective of our research is to better under- a series of three self-cleaning disk Amiad filters with a
stand drivers of the physical clogging phenomena. This 80 µm aperture size. Over the two months (June and July)
work focuses on particle agglomeration and its impact on of irrigation, water samples are regularly taken from five
drippers operation. Flow rate variations caused by modifi- drippers on each line, positioned at 10, 20, 50, 80, 90 m
cations in velocity and turbulence intensities are analyzed (line length 100 m) from the inlet. One more emitter is
through in-field and lab experiments coupled with a numer- analyzed on raw water at 95 m. Collection time is 120 s.
ical approach. Such a combination of scales constitutes an These measurements give the spatiotemporal changes in
original work to approach fluid–particles interactions in discharge, which is directly linked with the accumulation
clogging process. From in situ observations, we highlight of clogging material in the flow path. Distribution uni-
the correlation between agglomeration and clogging for formity is calculated using the DUlq (lower-quartile distri-
raw water naturally loaded with clay. In-lab experiments bution uniformity) coefficient corresponding to the ratio of
are performed to characterize clay agglomeration behav- the average discharge of the n/4 lower values (average low
ior and subsequently its impact on dripper functioning for quarter qmin ) to the average discharge of all (n) measure-
both inert and reactive clays. We attempt to model certain ments (q) ; n equals 10 (respectively 12) for filtered water
tricky observations such as flow rate modifications due to (raw water).
suspended particles.
qmin
DUlq = 100 (2)
q
Materials and methods For uniform distribution, i.e., good quality irrigation, DUlq
should not be lower than 85 %. Pilot pressure is main-
The dripper tained around 60 kPa. Pressure is recorded for each water
sampling thanks to a needle pressure gauge.
To analyze the causes of clogging and the processes The suspended particles in raw and filtered irrigation
involved, we use long-path drippers. This dripper model is waters are characterized using a laser sizer described in
the most commonly used worldwide (several billion units Sect. 2.3. Their mineralogy is determined by X-ray diffrac-
sold per year) due to its great durability and to the fact that tion on three water samples. More details of the setup can
its design patent entered the public domain many years be found in Bounoua (2010).
ago. The specific samples we use are commercialized under
John Deere Water brand. It is inserted within 16-mm-diam- In‑lab experiments
eter polyethylene pipes with a spacing of 0.3 m. The nomi-
nal operating pressure is 100 kPa resulting in a discharge For a better understanding of the clay aggregation impact
of 2 l h−1. The manufacturing variation coefficient (CVm) on dripper behavior, two kinds of clay suspensions are
is 3.17 %. The emitter flow rate increases with static pres- used:
sure in a lateral pipe according to an exponential relation
(Karmeli 1977): • a reactive smectite, which tends to aggregate depending
the water physicochemical balance (bentonite ABSO-
q = KPx (1) CLAY NAW, treated at CMMP Saint-Quentin, 02,

13
Irrig Sci (2016) 34:327–342 329

Fig. 1  Laser sizer setup

France). It is composed of natural sodium montmoril- Clogging setup at pilot scale


lonite extracted at Wyoming (USA);
• an illite mixed with calcite (Arvel Society based at This second experiment is operated to understand the
Saint-Paulien, 43, France), expected to be less active phenomena of clay-particle-induced clogging under con-
than the smectite with regard to the aggregation phe- trolled conditions (size and type of particles, ionic charge
nomenon. and temperature). We aim to identify the various processes
• Lab experiments are performed using two setups. One involved, in particular the effect of the ionic strength and
is dedicated to the characterization of clay aggregation nature of particles (illite–calcite vs. smectite) on aggrega-
mechanism. The other is built in order to test the impact tion. The pilot comprises four lines each equipped with
of agglomeration on dripper clogging. 10 drippers. These lines are fed from a 50-l reservoir con-
nected to two pumps: one for feeding the dripper units and
Clay aggregation setup the other for stirring the clay solution (200 mg l−1). The
water used to prepare the clay suspension is previously
It comprises a Taylor–Couette reactor coupled with a laser filtered at 1 µm. This filter is constituted of polypropylene
sizer (Mastersizer S, Malvern Instruments, UK). This microfibers (Spun 1 µm Big Blue) downstream of a 25-µm
technique is based on light diffraction according to the filtration (wound sediment cartridge Big Blue), to avoid
Fraunhofer theory. Our experimental facility is a closed saturation. The tank is first filled with 49 l of filtered water
loop circuit (Fig. 1) in which the suspension is pumped and then the clay, previously put in suspension into 1 l fil-
from the Taylor–Couette reactor (1 l) to the measurement tered water, is added in the tank. A cooler is immersed in
cell and then driven back to the reactor. The pump flow each of the four tanks to maintain the temperature between
is 5 × 10−4 l s−1. The container houses a Taylor–Couette 20 and 23 ◦ C. Each tank is operated using different salt
reactor which controls shear stress. The aggregate median concentrations (0, 10−3 and 10−2 M of NaCl). The pressure
size (D50) is measured as a time function at the Taylor– is maintained at 100 kPa. After 8 h, the flow rates are meas-
Couette reactor outlet. The respective effects of the ionic ured using a beaker placed under the drip outlet for five
strength and shear stress are investigated. Shear stress can minutes. This measurement is repeated daily through 40
cause agglomeration of the particles by increasing the days of experiment. The initial value of DUlq is calculated
probability of shocks. Conversely, it can also cause aggre- at the very beginning of the irrigation season and reveals to
gate fragmentation when it increases. These parameters be between 96 and 97 %, which is consistent with the value
govern the maximum size of the flocs that can be attained. of CVm measured in laboratory.
These factors are varied adding NaCl concentrations (0,
10−3, 10−2 and 10−1 M) and strain rate values of 20, 60 Numerical approach
and 210 s−1. The clay concentration is fixed at 200 mg l−1
through all the experiments. Size measurement frequency The aim of this part of the study is to deepen the analy-
is 0.33 Hz and lasts for 36 min or less if no size evolution sis of the interaction between fluid and particles in order to
is detected. characterize the impact of size, particle concentration and

13

330 Irrig Sci (2016) 34:327–342

Fig.  2  a Dripper with close-up of the labyrinth. This dripper is then inserted within 16-mm-diameter polyethylene pipe. b Geometry of the laby-
rinth

agglomeration process on velocity and turbulence fields. Multiphase flow model


The study is performed using commercial computational
fluid dynamics software where the models studied are The two-phase flow modeling relies on an Eulerian descrip-
implemented: ANSYS/Fluent V15.0.7. tion. The continuous and dispersed phases are water and
clay, respectively. The Eulerian model solves the contin-
Labyrinth geometry uum and the momentum equations of every phase which is
defined by their volume fraction. The volume of phase q, Vq
A portion of the repeated flow path of the dripper laby- is defined as:
rinth is designed and built numerically. It is constituted of a 
n
1-mm-width inlet, six baffles and a 1-mm-width outlet. The Vq = φq dV with Σq=1 φq = 1, (3)
distance between two baffles is 3 mm, while the minimum V

section size of the flow path is 1 mm (Fig. 2). This two- where φq is the volume fraction of phase q.
dimensional geometry is meshed by triangles. The mesh Continuum equation:
size is progressively refined until the results are no more
∂
dependent to the number of meshes. The minimum mesh φq ρq + ∇ · φq ρq − → n
    
vq = Σp=1 ṁpq ṁqp + Sq , (4)
size is about 3.8 µm. ∂t
where −
→v is the velocity of phase q, ṁ characterizes mass
q pq
Boundary conditions and hypothesis transfer phase p to phase q, and ṁqp the contrary. In this
study, in a first approximation, the source term Sq is sup-
Pressure values at inlet and outlet are set to obtain a pres- posed to be zero.
sure gradient of 5 kPa between these two sections; this gra- Momentum equation (Navier–Stokes):
dient is equivalent to the pressure drop recorded at nominal
pressure (100 kPa) of the complete dripper labyrinth. The ∂
φq ρq −

vq + ∇ · φq ρq −→
vq −

vq = −φq ∇p + ∇ · τq + φq ρq −

  
g
flow is considered two-dimensional, steady and the fluid ∂t −→  −
→ −−−→ −−−→
incompressible, and its temperature is maintained at 20 ◦ C . n
+ Σp=1 Rpq + ṁpq −
v→ −→
pq − ṁqp vqp + Fq + Flift,q + Fvm,q
The fluid used is water with a density ρf = 981 kg m−3 . (5)
Clay particles (ρp = 2800 kg m−3) are then injected. The


flow rate is 1 l h−1 which corresponds to a Reynolds number where τq is the shear rate of phase q, Fq is the external
−−→ −−→
(Re) of 770. Even if this value is below the transition lami- force, Flift,q is the lift force, Fvm,q is the added mass force,
−→
nar–turbulent flow for a Poiseuille flow, we assume a turbu- Rpq is interaction force between phases and p is the pres-
lent mode as the passage sections are close to large micro- sure for all phases.
channels (Hetsroni et al. 2005; Li and Olsen 2006). For this k − ǫ transport equations:
type of flow some experiments give a laminar–turbulent  
∂  − → µt,m
transition value which is lower than the conventional Poi-

(ρm k) + ∇ · ρm vm k = ∇ · ∇k + Gk,m − ρm ǫ,
∂t σk
seuille flow value (Wu and Little 1983; Pfund et al. 2000), (6)
while others show no displacement of this transition (Qu
and Mudawar 2002; Sharp and Adrian 2004). We select the
 
∂ µt,m ǫ
(ρm ǫ) + ∇. ρm −
 →  
vm ǫ = ∇. ∇ǫ + C1ǫ Gk,m − C2ǫ ρm ǫ ,
k − ǫ model as did many CFD studies on the flow in dripper ∂t σǫ k
labyrinths (Dazhuang et al. 2007; Wei et al. 2012). (7)

13
Irrig Sci (2016) 34:327–342 331

Table 1  Simulated cases
Case Flow type dp inlet (μm) Mass load m (–) Inlet particle volume Particle–particle interaction
fraction φp (–)

L0a Without particles – – – –


L5b Monodispersed 5 1.7 × 10−4 6 × 10−5 –
L20b Monodispersed 20 1.7 × 10−4 6 × 10−5 –
L60b Monodispersed 60 1.7 × 10−4 6 × 10−5 –
L20c Monodispersed 20 1.09 × 10−1 6 × 10−2 –
Lpoly Polydispersed 1–80 1.7 × 10−4 6 × 10−5 –
Lpoly-b Polydispersed 5 1.7 × 10−4 6 × 10−5 Aggregation and breakage

where ρm and − →
vm are the density and velocity of the mix- of ±5 % with regard to the nominal discharge. This vari-
ture, respectively. The turbulent viscosity µt,m is calculated ation is partially due to the manufacturing quality of the
using the equation: product (CVm = 3.17 %) but also possibly to fluctuations
in pressure in the network. Indeed, 5 kPa are recorded
k2 which imply 3.7 % discharge variation from Eq. 1. The
µt,m = ρm Cµ , (8)
ǫ second week measurements on the filtered lines, the first
and the production of turbulent kinetic energy Gk,m is cal- two drippers show a lower discharge than others that is
culated using the equation: not observed during further measurements. This temporary
  decrease in discharge may also be related to the transfer of
Gk,m = µt,m ∇ − →
vm + (−

vm )T : ∇ −

vm , (9) particles trapped in the labyrinth which may take time to
reach the dripper’s outlet. After four weeks of daily appli-
By default, the constants are: C1ǫ = 1.44; C2ǫ = 1.92; cation of raw water (Fig. 3), the discharge of all drippers
Cµ = 0.09; σk = 1 and σǫ = 1.3. decreases, whereas with filtered water, only a few drippers
have lower discharge. For raw waters, a 5 % decrease in
Numerical tests average discharge occurs from the second week (Fig. 3).
During the fourth week, the decrease is more marked, due
To better understand the impact of particles in the flow, to heavy storms during this period in the mountains where
different simulations are carried out by varying different water is coming from. After seven weeks, we observe a
parameters. In particular, we use both monodispersed and 19 % decrease in the average discharge of operating drip-
polydispersed particles. Table 1 summarizes all the cases pers (clogged dripper is not taken into consideration) with
studied. For the polydispersed cases, 20 diameter classes raw water, while a 6 % decrease only is observed with fil-
from 1 to 80 µm are injected at the inlet, with 3 × 10−6 tered water. The 80-µm filtration is thus shown to signifi-
volume fractions for each one. In the Lpoly-b case, aggre- cantly improve discharge stability, while the unfiltered
gation and breakage are simulated using the Luo model water induces progressive clogging. Only two drippers
(Luo 1993). supplied with raw water among the 12 analyzed are totally
clogged from the seventh week. They are situated at 95 m ,
i.e., close to the end of the line. Thus, when the system is
Results and discussion turned on, the high initial velocity of water may flush some
particles downstream. This results in a higher exposure of
Field characterization of physical clogging terminal drippers to particle deposit. Indeed, these por-
tions of irrigation blocks are characterized by lower water
These in situ experiments try to link suspended load with velocity which favor particle settling (Capra and Scicolone
dripper flow rate variations. 1998). Moreover, in the case of filtered water, the distri-
bution uniformity remains high (DUlq > 94 %, Table 2)
Analysis of discharge throughout the whole duration of the experiment despite
the decrease in discharge (Fig. 3). We can thus suspect
Figure  3 shows variations in the discharge of the drippers that the initial ±5 % DUlq is mainly due to the manufac-
supplied with raw and filtered waters. At the very begin- turing quality (CVm = 3.17 %), pressure variation (3.7 %)
ning on the four lines of the experiment, there is a variation and pressure drop along the lateral (headloss), but not to

13

332 Irrig Sci (2016) 34:327–342

Fig. 3  Time evolution of
drippers discharge according
to their position (distance from
inlet) as a function of time with
2 water qualities (filtered water
f, raw water r). Error bars show
the variability of the measure-
ments obtained

Table 2  Time evolution of DUlq with filtered and raw water


1 June (%) 11 June (%) 19 June (%) 28 June (%) 5 July (%) 11 July (%) 19 July (%)

Filtered water 96 94 96 96 96 95 98
Raw water 97 95 97 93 69 67 69

clogging. It suggests that the uniformity coefficient alone • Drippers supplied with raw water clog from the begin-
is not sufficient to reveal clogging phenomena if these are ning of the observations (4 % after two weeks). The
homogeneous. According to Chossat (1995), in this case it same trend is observed on lines irrigated with filtered
would not be necessary to undertake maintenance of the waters, but a 4 % decrease only is observed six weeks
system. Consequently, the uniformity coefficient needs to after the drippers have been operating. It means that
be associated with the calculation of deviations of the dis- despite filtration effectiveness, particles are passing that
charge from initial value (Ec) as expressed in Eq. 10 and could settle inside the drippers path.
presented in Fig. 4: • With raw water, we observe a sharp drop in discharge
q − qn and uniformity [DUlq = (69 − 67) %] after a stormy
Ec = 100 (10) period that occurs during the fourth week. It results in
qn
the complete clogging of two drippers that are disre-
where q is measured discharge (l/h) and qn is the initial garded in the following periods. This sudden discharge
discharge. drop is not observed in filtered water, although a slight
Figure 3 and Table 2 highlight the following: and uniform decrease occurs.

13
Irrig Sci (2016) 34:327–342 333

diameters reach 500 µm or even more in raw water. Particles


may have benefited from favorable hydrodynamic and chemi-
cal conditions for aggregation resulting in the emergence of
a new population of particle size greater than the filter mesh
size (80 µm). Our hypothesis is that the new population with
larger size is formed during the no-irrigation phase, when the
fine suspended materials and/or the solutes precipitating with
the increasing temperature can deposit inside the pipe. During
the irrigation phase, the re-suspension induced by the flow is
transported through the lateral in downstream samples. This
phenomenon is logically more pronounced with raw water,
where the amount suspended particulate matter is higher.
These size distribution measurements are only qualitative as
they do not reflect the concentrations of particles involved.
Fig. 4  Variations in average discharge at the outlet of operating drip- Raw water concentration is higher than filtered water ones.
pers
Characterization of the mineralogy of suspended particles
• These observations make it necessary to investigate the
characteristics of the clogging material. Some raw water is sampled and analyzed with X-ray dif-
fraction so as to characterize the mineralogy of its sus-
pended particulate matter. The three measurements are in
close agreement. The diffractogram presented corresponds
Particle size analysis to one sample (Fig. 6). It reveals a relative heterogeneity
with the dominating presence of quartz and calcite accom-
Particle size is measured in samples of raw and filtered panied by phyllosilicate such as micas, chlorite and illite.
water taken from the upstream and the downstream of dis- Infrared spectroscopy is also used to characterize the chem-
tribution laterals using two draining pipes inserted in the line ical composition of the suspended particulate matter. It con-
(Fig. 5). Those draining pipes are open only when sampling firms the major presence of silicates and alumino-silicates
is required. At the first outlet (upstream), we observe a peak and does enable to detect any additional material such as
in the population of particles suspended in filtered water cen- natural organic matter. Such dominance of inorganic mate-
tered at 4 µm. In raw water, the distribution curve is bimodal rial is expected in river water. Thus, we assume that in our
with peaks at 5 and 30 µm. For both types of water, we context, physicochemical clogging prevails over biological
observe an increase in particle diameter between the upstream clogging. This hypothesis is consistent with particle size
and downstream samples. The peaks are both located at measurements (Fig. 5) given that clay is considered respon-
80 µm in filtered and raw water, respectively, but the outlier sible for most coagulation phenomena (Labille et al. 2003).

Fig. 5  Particle size distributions in filtered (a) and raw (b) water between the beginning and the end of the lateral (sample collected at dripper’s
outlet)

13

334 Irrig Sci (2016) 34:327–342

Fig. 6  X-ray diffractogram of
the suspended particles occur-
ring in the raw water

Agglomeration process and physical clogging collision frequency. At 10−1 M NaCl, the fastest aggrega-
tion kinetics are measured, moreover increasing with strain
To better understand how the agglomeration process rate value. This corresponds to the typical diffusion-limited
impacts the clogging mechanism, experiments in con- aggregation (DLA) process, i.e., when each interparticle
trolled laboratory conditions are performed with solutions collision leads to aggregation. The largest aggregates result
(200 mg l−1 or 0.710−5 mass load) of reactive (smectite) from this condition, with average size decreasing as strain
and inert (illite–calcite) clays. The main objective is to rate increases (40, 30 and 20 µm measured for 20, 60 and
determine how salt concentration and strain rate influence 210 s−1, respectively). In this case, strain rate does not only
the above. favor aggregation following particle collision, it also causes
the aggregate fragmentation or erosion when the internal
Characterization of the aggregation process aggregate tensile strength becomes insufficient with regard
to the size. The steady state obtained thus results from the
In this section, only smectite size variations are detailed. quasi-equilibrium between the aggregation forces and the
The illite–calcite suspension does not agglomerate (data fragmentation forces. The balance is moved to the aggre-
not shown) but remains totally dispersed at the 5 µm ele- gation side when the salt concentration, i.e., interparticle
mentary size whatever the physicochemical conditions attraction, increases. It is moved to the opposite fragmenta-
tested, even at the highest salt concentration (Bounoua tion side when the strain rate increases. Stabilized aggre-
2010). For smectite suspensions, the aggregation kinet- gate sizes are plotted in Fig. 8 for all salt concentrations. As
ics are plotted in Fig. 7 for the three strain rates and three already observed, when the strain rate increases, smectite
salt concentrations tested. In the absence of salt, no aggre- tends to slightly aggregate at 10−3 M NaCl or to break up at
gation is measured at any strain rate, and the D50 remains 10−1 M NaCl, whereas the particles remain dispersed when
5 µm. It appears that 10−3 M NaCl concentration is just salt is absent. At 10−2 M NaCl, size decrease is observed
sufficient to induce slow aggregation. Indeed, the average subsequent to slight increase. Despite opposite tendencies,
size at 30 min increases to 8 µm for the three strain rate i.e., aggregation or fragmentation, it appears that an asymp-
values tested. At 10−2 M NaCl, the aggregation kinetics totical size could be reached beyond 1000 s−1 strain rate
increases with the strain rate. This behavior is a character- values. Indeed, log-linear regressions fit the median size
istic of the reaction-limited aggregation (RLA) process for variation for 10−3 and 10−1 M NaCl:
which only a certain fraction of the total collisions leads
10−3 M NaCl

to aggregation. In this case, the steady-state aggregate 1.63 ln(G) + 0.28
D50 = (11)
size, between 10 and 20 µm, is not correlated with the −6.11 ln(G) + 59.23 10−1 M NaCl
strain rate intensity. However, the greater the strain rate, The correlation coefficients are 0.89 and 0.82 for 10−3 and
the sooner steady state is reached, certainly due to higher 10−1 M NaCl, respectively. At an estimated dripper strain

13
Irrig Sci (2016) 34:327–342 335

Fig. 8  Stabilized smectite sizes as a function of strain rate for several


salt concentrations

the maximum strain rate is 250 s−1), we cannot assert that


huge strain rates would not break the aggregate and thus
decrease the asymptotical D50 values.

Impact of clay and salt concentration on dripper clogging

For eight hours operating time per day, Fig. 9 represents the
evolution over time of the ratio between the nominal drip-
per flow rate, without clay particles, and with the illite–cal-
cite solution. The mean flow rate increases by some 5 %
when illite–calcite particles are added. We suppose that this
increase is related to the interaction between the particles
and small-scale turbulence which decreases the turbulent
kinetic energy. The resulting flow laminarization tends to
increase the mean velocity in the labyrinth. During the test,
there are small flow fluctuations, but the flow rate remains
higher than the nominal. The decreases are all reversible
and may therefore be related to the evacuation of particles
at system turn on. For eight hours operating time per day,
Fig.  10 represents the time evolution of the ratio between
the dripper nominal flow rate, without clay particles, and
the three smectite suspensions at varying salt concentra-
tion (0, 10−3 and 10−2 M NaCl). Globally, the mean flow
rates progressively decrease over the entire test. Nonethe-
less, intermittent increases are observable due to unclog-
ging phenomena occurring at system turn on. For each of
the above case, we note the following:
Fig. 7  Impact of salt concentration (light gray 10−3 M NaCl, gray
10−2 M NaCl, black 10−1 M NaCl) on time evolution of D50 for
strain rate values of 20 s−1 (a), 60 s−1 (b) and 210 s−1 (c), respectively • without salt, flow rate decrease reaches 60 % after 40
days. Clogged dripper observation (Fig. 11) reveals that
the clogging material is a sort of gel containing clay
rate of approximately 3000 s−1 (3.1), this equation predicts particles. Gel formation can be related to functioning
16 and 12 µm median smectite aggregate size for 10−3 and intermittence: During the stop phase, water leaks out
10−1 M NaCl, respectively. However, without experiments and smectite concentration increases beyond the solid–
at such high strain rates (for this Taylor–Couette reactor, gel transition phase (Abend and Lagaly 2000).

13

336 Irrig Sci (2016) 34:327–342

Fig. 9  Time evolution,
8 h day−1 functioning, of the
ratio between the nominal flow
rate, without clay particles, and
the illite–calcite solution one

Fig. 10  Time evolution,
8 h day−1 functioning, of the
ratio between the nominal flow
rate, without clay particles, and
the three smectite solution salt
concentrations

• with 10−3 M NaCl, it appears that the presence of salt • with 10−2 M NaCl, as for 10−3 M NaCl, flow rate
generates a lower flow rate decrease in comparison with decreases by up to 30 % at the beginning of the experi-
the clear water case over the first 30 days of experiment. ment. This decrease can be linked whether to direct
Then, a fast flow rate decrease is observed reaching particle interaction with flow, which is numerically
almost a nil value after 40 days of operation. As for the studied in 3.3, or to slight clogging due to agglomer-
previous concentration, clogged dripper observations ates which are not yet sedimented in the tank. There is
(Fig. 11) reveal that the clogging material is a sort of gel a slight increase (17 %) on the fortieth days due to drip-
containing clay particles. per unclogging. Clogged dripper observations (Fig. 11)

13
Irrig Sci (2016) 34:327–342 337

Fig. 11  Clogged dripper
labyrinth photograph after 40
days of 8 h day−1 operation for
the three smectite solution salt
concentrations

Fig. 12  Streamlines colored
by water velocity for a flow
without particles (case L0a of
Table 1)

reveal biofilm presence. In this particular case, no mineral cake hydraulic conductivity, certainly due to the salt-
gel is observed. With this high salt concentration, smec- induced aggregation of the clay that leads to larger inter-
tite has already aggregated in the feeding tank. Thus, its connected porosity in the cake (Santiwong et al. 2008).
transfer, through the supply line and then the dripper, is However, our results have to be balanced by the fact that
different from the previous salt concentration where clay we work with only one kind of salt. Indeed, the aggre-
particles are dispersed. At 10−2 M NaCl, it is highly prob- gation process could have been different when using
able that aggregate sedimentation prevents the smectite another electrolyte solution such as CaCl2 instead of
from circulating through the irrigation system. NaCl (Quirk and Schofield 1955; Stawinski et al. 1990).
• The clogging mechanism seems to be governed by clay
gel presence. Moreover, one unexpected phenomena is Fluid–particle interaction modeling
observed: When salt concentration increases, the clog-
ging mechanism appears to be reduced. This effect is In this section, we analyze and discuss the results obtained
already observed with clay gel cake filtration due to par- by numerical simulations. The discussion focuses on fluid–
ticles deposition on upstream membrane surface. In this particle interaction as it is one of the hypothesis to explain
case, step increase in ionic strength leads to increase in the variations in clogging observations. We also investigate

13

338 Irrig Sci (2016) 34:327–342

Fig. 13  Streamlines colored by
water velocity for a monodis-
persed flow with 5 µm particle
diameter, 6 × 10−5 particle
volume fraction at inlet and a
mass load of 1.7 × 10−4 (case
L5b of Table 1)

Fig. 14  Water velocity (a) and turbulent kinetic energy (b) profiles L5b; 20 µm, case L20b and 60 µm, case L60b), 6 10−5 particle vol-
for a flow without particles (case L0a of Table 1) or for a monodis- ume fraction at inlet and a mass load of 1.7 10−4 (Table 1)
persed flow with different seeding particle diameters (5 µm, case

particle dispersion within labyrinth depending on their size ǫav correspond to the averaged value of ǫ on the whole laby-
and their concentration. We quantify in detail the veloc- rinth fluid domain (surface). For an average value of turbu-
ity and turbulence changes due to particle agglomeration. lent dissipation ǫav of 129 m2 s−3 (numerically estimated), η
This approach is innovative as very few studies tackle this is approximately 10 mm. Subsequently, particles larger and
process. smaller than η are numerically injected in order to charac-
terize their effect on flow structure as presented further.
Without particles
Monodispersed laden flow
Before injecting particles into the water, the pressure flow
relationship observed is validated. Figure 12 illustrates Cases L5b, L20b, L60b and L20c (Table 1) are analyzed.
that the flow is composed of a high-velocity regions with Only particle size and mass load impacts are studied with-
a maximum at 2.2 m s−1 and low swirl effect, which is out aggregation and breakage mechanisms. Mass load m is
in agreement with previous work (Wei et al. 2006, 2012; defined according to the following formula:
Dazhuang et al. 2007 ). The size of the smallest eddy struc- ρp
ture, known as the Kolmogorov microscale η, is calculated m = φp (13)
ρf
using the following formula:
1/4 Figure 13 shows that the velocity fields are maintained even
ν3

η= though particles are added. Water velocity appears slightly
ǫav (12) decreased, especially at the edges of the baffles. To better

13
Irrig Sci (2016) 34:327–342 339

turbulence reduction is consistent with Kulick et al.


(1994), who observe that turbulent intensity is reduced
when small particles are added to a flow. Yuan and
Michaenidies (1992) assume that this reduction is related
to additional dissipation, which is caused by the particle
drag force.
Figure  15 reveals that the particle volume fractions for
case L20b and L60b are close to zero in the recirculation
zone, whereas for L5b it is well distributed over the line a.
Moreover, the particle volume fraction variation for these
two cases is very significant in the main flow. These results
are to be expected because L5b particles are smaller than
η and can enter a vortex with a size of η, unlike particles
L20b and L60b which are bigger than η and thus follow the
Fig. 15  Particle volume fraction distribution for different seeding main flow without entering the recirculation zone. Looking
particle diameters (5 µm, case L5b; 20 µm, case L20b and 60 µm, further in the recirculation zone (y < 0.5 mm), it appears
case L60b), 6 10−5 particle volume fraction at inlet and a mass load
of 1.7 10−4 (Table 1)
that 20-µm-diameter particles can be trapped in the swirl;
contrary to 60-µm-diameter ones. This observation agrees
with Niu et al. (2013) that underline that there exists a sen-
understand particles impact on the flow, profiles are plot- sitive particle range (around 30 µm diameter) which favor
ted on a transversal line (line a in Fig. 13). This line goes deposit in swirl region. Cases L20b and L20c are compared
through the main flow and a recirculation zone. to discuss particle volume fraction impact on the flow prop-
Regarding impact of particle diameters, Fig. 14 clearly erties. Figure 16 shows the profiles of water velocity and
shows that when particles are injected water velocity and turbulent kinetic energy. Even though the particle volume
turbulent kinetic energy decrease. However, particle size fraction at the inlet is increased by 1000 times, the overall
does not have any specific influence: Intensities are the profile remains almost the same: Water velocity has slightly
same for all diameters. Unexpectedly, 5-µm-diameter increased (1 %) in the recirculation zone. The impact on
particles do not increase the flow as observed in the turbulent kinetic energy is higher; the maximum value in
clogging test with illite–calcite particles. When cal- the recirculation zone increases by about 20 %. This result
culating the Reynolds number for each case, it appears contradicts previous works (Kulick et al. 1994; Vreman
that it decreases from 770 without particles (case L0a) 2007) which assume that turbulence decreases as mass load
to 730 with particles (cases L5b, L20b and L60b). This increases. This discrepancy may be explained by the fact

Fig. 16  Water velocity (a) and turbulent kinetic energy (b) profiles pared with 6 10−2 particle volume fraction at inlet and a mass load of
for 20-µm-diameter particles and different mass load: 6 10−5 particle 1.09 10−1 (case L20c) (Table 1)
volume fraction at inlet and a mass load of 1.7 10−4 (case L20b) com-

13

340 Irrig Sci (2016) 34:327–342

that the flow structure inside a dripper is not representative


of cylindrical pipe flow and that our mass loads are much
lower.

Polydispersed laden flow

Aggregation and breakage mechanism impacts are ana-


lyzed (cases Lpoly and Lpoly-b) in this section. This
section is an innovative work as very few work are deal-
ing with multi-phase flow modeling in drippers (Lin and
Yang 2007; Qingsong et al. 2008) and none of them are
approaching aggregation and breakage mechanisms.
When comparing the two figures (Fig. 17), we note
that the recirculation center changes position when aggre-
gation and breakage mechanism are activated: Lpoly-b
main flow area is narrower than Lpoly, and consequently,
swirl zones are larger. This may be due to the increase of
smaller particles produced by aggregation and breakage
in swirl zones (not shown here) which thus modify this
area. Particle–particle interaction seems to change velocity
contours. When compared to Lpoly (Fig. 18), the velocity
profile of Lpoly-b decreases whereas the turbulent kinetic
energy increases significantly (fourfold) in the main flow
area. The turbulent kinetic energy reflects velocity fluctua-
tions. Therefore, according to the results obtained, it would
appear that velocity fluctuations are more significant when
there are particle–particle interactions. Collision and coag-
ulation between particles consistently influence mixture
turbulence.

Conclusions and perspectives

The concept on which drip irrigation relies makes it sensi-


tive to clogging phenomena. Filtration is usually installed
upstream to mitigate such occurrences. In situ observations
Fig. 17  Streamlines colored by water velocity for 6 10−5 particle
demonstrate that the filtration technique is not totally pre- volume fraction at inlet, a total mass load of 1.7 10−4 and different
venting any clogging when small particles go through the polydispersed particles: a Lpoly-b [(an initial diameter classes of
filter and agglomerate further in bigger aggregates. The pre- 5 µm with a mass load of 1.7 10−4 and aggregation and breakage pro-
vailing hypothesis is that such aggregation is governed by cess simulated by Luo model (Luo 1993)] and b Lpoly (20 diameter
classes from 1 to 80 µm injected at the inlet with 3 10−6 volume frac-
the presence of clay particles whose electric charges allow tions for each one) (Table 1)
agglomeration in the presence of salts. The role of salt con-
centration and flow path strain rate appear to be determinant.
Such phenomena are investigated through experiments in of clogging in the field that may rely on the mineralogy of
controlled laboratory conditions with reactive (smectite) and clay transported by the water. Clogging tests are performed
inert (illite–calcite mix) clays. Taylor–Couette experiments with clay in suspension in water. The flow rate is measured
show that smectite particle size varies depending on strain on four lines of ten drippers. The first observation is that
rate and salt concentrations but tends to have an asymptotical particles impact the flow rate whatever their size. A flow
diameter of 20 µm at high strain rate (higher than 100 s−1) rate decrease or increase for smectite and illite–calcite mix,
whatever the salt concentrations. As expected, the illite–cal- respectively, is observed. This supports the idea that there is
cite mix particle size does not evolve even for high strain an interaction between the flow turbulent structure and the
rate and high salt concentrations. Such sharp difference may particles. Second, the results obtained support the view that
explain the high variability of drippers behavior in terms salt concentrations, and thus particle size, are not directly

13
Irrig Sci (2016) 34:327–342 341

Fig. 18  Water velocity (a) and turbulent kinetic energy (b) pro- gation and breakage process simulated by Luo model (Luo, 1993)]
files for 6 10−5 particle volume fraction at inlet, a total mass load of and without (Lpoly) (20 diameter classes from 1 to 80 µm injected at
1.7 10−4 and different polydispersed particles: with [(Lpoly-b: an ini- the inlet with 3 10−6 volume fractions for each one) aggregation and
tial diameter classes of 5 µm with a mass load of 1.7 10−4 and aggre- breakage mechanisms (Table 1)

correlated with the clogging mechanism. A surprising ten- Adin A, Sacks M (1991) Dripper-clogging factors in wastewater irri-
dency is observed: Increasing salt concentrations in smectite gation. J Irrig Drain Eng 117:813–826
Bounoua S (2010) Etude du colmatage des systèmes d’irrigation
suspension decreases clogging, whereas it increases agglom- localisée. PhD thesis, Aix-Marseille University
erate size. Further experiments based on optical technologies Capra A, Scicolone B (1998) Water quality and distribution uni-
could greatly improve our understanding. Such works are formity in drip/trickle irrigation systems. J Agric Eng Res
ongoing with also others kinds of salt and clay (kalonite). 70(4):355–365
Chossat J (1995) Entretient en micro-irrigation, vol 19, 2e edn., Col-
Fluid–particles interaction has been studied thanks to an lection étude du Cemagref serie Equipement pour l eau et l envi-
Eulerian model of the particle–fluid mix. Particles in suspen- ronnement, Antony
sion impact turbulent kinetic energy and velocity: (1) inject- Dazhuang Y, Peiling Y, Shumei R, Yunkai L, Tingwu X (2007)
ing inert particles reduces overall water velocity and turbu- Numerical study on flow property in dentate path of drip emit-
ters. N Zeal J Agric Res 50:705–712
lent kinetic energy of the mixture in the flow. It also reduces Gilbert R, Nakayama F, Bucks D, French O, Adamson K (1981)
the turbulence of the flow (decrease of Reynolds number); Trickle irrigation: emitter clogging and other flow problems.
(2) inert particle size does not affect either water velocity or Agric Water Manag 3:159–178
turbulent kinetic energy and hence does not modify the flow Hetsroni G, Mosyak A, Pogrebnyak E, Yarin L (2005) Fluid flow in
micro-channels. Int J Heat Mass Transf 48:1982–1998
rate. However, particles are not distributed in the same way Hills D, Nawaar F, Waller P (1989) Effects of chemical clogging on
depending on their size relative to the Kolmogorov micro- drip-tape uniformity. Trans ASAE 32(4):1202–1206
scale; (3) mass load (in the range studied) does not influ- Karmeli D (1977) Classification and flow regime analysis of drippers.
ence water velocity, but it increases turbulent kinetic energy J Agric Eng Res 22:165–173
Kulick J, Fessler J, Eaton J (1994) Particle response and turbulence
in the recirculation zone; (4) agglomeration and break-up modification in fully developed channel flow. J Fluid Mech
mechanisms influence turbulence, which result in a veloc- 277:109–134
ity decrease. These numerical results need to be further ana- Labille J, Thomas F, Bihannic I, Santaella C (2003) Destabilization of
lyzed and extended to a tridimensional modeling of the flow. montmorillonite suspensions by Ca2 + and succinoglycan. Clay
Miner 38:173–185
In future works, we plan to use low Reynolds number mod- Levy G, Fine P, Bart-Tal A (2011) Treated wastewater in agriculture:
els which should greatly improve our results. use and impacts on the soil environments and crops. Wiley-
Blackwell, Hoboken
Acknowledgments This work has been funded by the Region Li H, Olsen M (2006) Micropiv measurements of turbulent flow in
Provence-Alpes-Cotes d’Azur and IRSTEA, the National Research square microchannels with hydraulic diameters from 200 µm to
Institute of Science and Technology for Environment and Agriculture. 640 µm. Int J Heat Fluid Flow 27:123–134
Lin K, Yang J (2007) Chaotic mixing of fluids in a planar serpentine
channel. Int J Heat Mass Transf 50:1269–1277
Luo H (1993) Coalescence, breakup and liquid circulation in bubble
References column reactors. Ph.D. thesis, Norwegian Institute of Technol-
ogy, Trondheim, Norway
Abend S, Lagaly G (2000) Dsol-gel transitions of sodium montmoril- Nakayama F, Bucks D (1991) Water quality in drip/trickle irrigation.
lonite dispersions. Appl Clay Sci 16(34):201–227 Irrig Sci 12:187–192

13

342 Irrig Sci (2016) 34:327–342

Niu W, Liu L, Chen X (2013) Influence of fine particle size and assemblages during microfiltration of montmorillonite suspen-
concentration on the clogging of labyrinth emitters. Irrig Sci sions. J Colloid Interface Sci 317(1):214–227
31:545–555 Sharp K, Adrian R (2004) Transition from laminar to turbulent flow in
Pfund D, Rector D, Shekarriz A (2000) Pressure drop measurements liquid filled microtubes. Exp Fluids 36:741–747
in a microchannel. Fluid Mech Transp Phenom 46(8):1496–1507 Stawinski J, Wierzchos J, Garcia-Gonzalez M (1990) Influence of cal-
Pitts D, Haman D, Smajstra A (1990) Causes and prevention of cium and sodium concentration on the microstructure of benton-
emitter plugging in microirrigation systems, Florida Coopera- ite and kaolin. Clays Clay Miner 38(6):617–622
tive Extension Service Bulletin, vol 258. University of Florida, Vreman A (2007) Turbulence characteristics of particle-laden pipe
Florida flow. J Fluid Mech 584:235–279
Qingsong W, Gang L, Jie L, Yusheng S, Wenchu D, Shuhuai H (2008) Wei Q, Yusheng S, Wenchu D, Gang L, Shuhuai H (2006) Study on
Evaluations of emitter clogging in drip irrigation by two-phase hydraulic performance of drip emitters by computational fluid
flow simulations and laboratory experiments. Comput Electron dynamics. Agric Water Manag 84:130–136
Agric 63:294–303 Wei Z, Cao M, Liu X, Tang Y, Lu B (2012) Flow behaviour analysis
Qu W, Mudawar I (2002) Experimental and numerical study of pres- and experimental investigation for emitter micro-channels. Chin
sure drop and heat transfer in a single-phase micro-channel heat J Mech Eng 25:729–737
sink. Int J Heat Mass Transf 45:2549–2565 Wu P, Little W (1983) Measurement of friction factor for flow of
Quirk J, Schofield R (1955) The effect of electrolyte concentration on gases in very fine channels used for micro-miniature Joule
soil permeability. J Soil Sci 6(2):163–178 Thompson refrigerators. Cryogenics 23:273–277
Ravina I, Paz E, Sofer Z, Marcu A, Schisha A, Sagi G (1992) Control Yuan Z, Michaenidies E (1992) Turbulence modulation in par-
of emitter clogging in drip irrigation with reclaimed wastewater. ticle flows—a theoretical approach. Int J Multiph Flow
Irrig Sci 13:129–139 18(15):779–785
Santiwong S, Guan J, Waite T (2008) Effect of ionic strength and
ph on hydraulic properties and structure of accumulating solid

13

View publication stats

You might also like