Professional Documents
Culture Documents
DOI 10.1007/s11071-009-9624-7
O R I G I N A L PA P E R
Received: 1 May 2009 / Accepted: 13 November 2009 / Published online: 10 December 2009
© Springer Science+Business Media B.V. 2009
dp Is subjected to a static pressure rise (m) who popularized and refined the graphical calculation
hp Head gain from a pump (m) method. Wylie and Streeter [6] combined the method
hL Combined head loss (m) of characteristics with computer modeling. The field
Eν Bulk modulus of elasticity (Pa) of fluid transients is still rapidly evolving worldwide
α Kinetic energy correction factor by Brunone et al. [7], Koelle and Luvizotto [8], Fil-
P Surge pressure (m) ion and Karney [9], Hamam and McCorquodale [10],
ρ Density (kg/m3 ) Savic and Walters [11], Walski and Lutes [12], Wu and
C Velocity of surge wave (m/s) Simpson [13]. Various methods have been developed
g Acceleration of gravity (m/s2 ) to solve transient flow in pipes. These ranges have
K Wave number been formed from approximate equations to numerical
Tp Pipe thickness (m) solutions of the non-linear Navier–Stokes equations.
Ep Pipe module of elasticity (kg/m2 ) Hydraulic transient flow is also known as unsteady
Ew Module of elasticity of water (kg/m2 ) fluid flow. During a transient analysis, the fluid and
C1 Pipe support coefficient system boundaries can be either elastic or inelastic:
Ymax Max. fluctuation (a) elastic theory describes the unsteady flow of a com-
Min. Minimum pressible liquid in an elastic system (e.g., where pipes
Max. Maximum can expand and contract); (b) rigid-column theory, de-
Lab. Laboratory scribes unsteady flow of an incompressible liquid in
a rigid system. It is only applicable to slower tran-
sient phenomena. Both branches of transient theory
1 Introduction stem from the same governing equations. The conti-
nuity equation and the momentum equation are needed
Long-distance water transmission lines must be eco- to determine V (velocity) and p (surge pressure) in a
nomical, reliable, and expandable. The present work one-dimensional flow system. Solving these two equa-
tried to show safe hydraulic input to a network. This tions produces a theoretical result that usually corre-
idea provided wide optimization and water hammer sponds quite closely to actual system measurements,
risk-reduction strategy for the Rasht city water main if the data and assumptions used to build the numeri-
pipeline in northern Iran. The important effect of wa- cal model are valid. Transient analysis results that are
ter hammer disaster is unaccounted for in the water not comparable with actual system measurements are
“UFW” phenomenon. This work had particular exper- generally caused by inappropriate system data (espe-
tise for designing safe pressurized pipeline segments. cially boundary conditions) and inappropriate assump-
This means that by reduction of unaccounted for wa- tions [2, 3]. Among the approaches proposed to solve
ter, “UFW”, energy costs can be reduced. Water ham- the single-phase (pure liquid) water hammer equations
mer disaster as a fluid dynamics phenomenon is an im- are the Method of Characteristics (MOC), Finite Dif-
portant case study for designer engineers. Water ham- ferences (FD), Wave Characteristic Method (WCM),
mer is a pressure surge or wave caused by the kinetic Finite Elements (FE), and Finite Volume (FV). One
energy of a fluid in motion when it is forced to stop or difficulty that commonly arises relates to the selec-
change direction suddenly [1]. The majority of tran- tion of an appropriate level of time step to use for the
sients in water and wastewater systems are the result analysis. The obvious trade-off is between computa-
of changes at system boundaries, typically at the up- tional speed and accuracy. In general, the smaller the
stream and downstream ends of the system or at local time step, the longer the run time but the greater the
high points. Consequently, the results of the present numerical accuracy. The challenge of selecting a time
work can reduce the risk of system damage or failure. step is made difficult in pipeline systems by two con-
The study of hydraulic transients is generally consid- flicting constraints. First, to calculate many boundary
ered to have begun with the works of Joukowsky [2] conditions, such as obtaining the head and discharge at
and Allievi [3]. The historical development of this sub- the junction of two or more pipes, it is necessary that
ject makes for good reading. A number of pioneers the time step be common to all pipes. The second con-
have made breakthrough contributions to the field, straint arises from the nature of the MOC. If the adjec-
including Angus and Parmakian [4] and Wood [5], tive terms in the governing equations are neglected (as
Some aspects of physical and numerical modeling of water hammer in pipelines 679
is almost always justified), the MOC requires that ratio components of the energy equation can be combined
of the distance x to the time step t be equal to the to express two useful quantities, the hydraulic grade
wave speed in each pipe. In other words, the Courant and the energy grade:
number should ideally be equal to one and must not
exceed one by stability reasons. For most pipeline sys- (P1 /γ ) + Z1 V12 /2g + hp
tems, having as they do a variety of different pipes
= (P2 /γ ) + Z2 + V22 /2g + hL , (1.1)
with a range of wave speeds and lengths, it is impos-
sible to satisfy exactly the Courant requirement in all
Pressure head: p/γ
pipes with a reasonable (and common) value of t.
Elevation head: z
Faced with this challenge, researchers have sought for
Velocity head: V 2 /2g
ways of relaxing the numerical constraints. Two con-
trasting strategies present themselves. The method of where p—pressure (N/m2 ), γ —specific weight
wave-speed adjustment changes one of the pipeline (N/m3 ), z—elevation (m), V —velocity (m/s), g—
properties (usually the wave speed, though more rarely gravitational acceleration constant (m/s2 , ft/s2 ).
the pipe length is altered) so as to satisfy exactly the
Courant condition [14]. Unsteady or transient friction Compared to a steady
Every pipe system has a characteristic time pe-
state, fluid friction increases during hydraulic transient
riod, T = 2L/a, where L is the longest possible path
events because rapid changes in transient pressure and
through the system and a is the pressure wave speed.
flow increase turbulent shear. The MOC model can
This period is the time it takes for a pressure wave
track the effect of fluid accelerations to estimate the at-
to travel the pipe system’s greatest length two times.
tenuation of transient energy more closely than would
It is recommended that the run duration equals or ex-
be possible with quasi-steady or steady-state friction.
ceeds T . Another factor to consider when determining
Computational effort increases significantly if tran-
run duration is to allow enough time for friction to sig-
sient friction must be calculated for each time step.
nificantly dampen the transient energy. If in doubt, run
This can result in long model-calculation times for
work for a longer duration and examine the resulting
large systems with hundreds of pipes or more. Typ-
graphs and time histories. Run duration is measured
either in seconds or as a number of time steps. Time ically, transient friction has little or no effect on the
steps typically range from a few hundredths of a sec- initial low and high pressures and these are usually the
ond to a few seconds, depending on the system and largest ever reached in the system. This is illustrated
the pressure wave speeds. The run duration has a di- from the following present work simulation results
rect effect on the modeling computation time, along comparing steady, quasi-steady, and transient friction
with the time step selected for the simulation. Steady- methods. The steady-state friction method yields con-
state models, such as Water CAD or Water GEMS, servative estimates of the extreme high and low pres-
are capable of two modes of analysis: steady state and sures that usually govern the selection of pipe class
extended period simulation (EPS). EPS solves a se- and surge-protection equipment. However, if cyclic
ries of consecutive steady states using a gradient algo- loading is an important design consideration, the un-
rithm and accounting for mass in reservoirs and tanks steady friction method can yield less conservative es-
(e.g., net inflows and storage). Both methods assume timates of recurring and decaying extremes. In the
the system contains an incompressible fluid, so the to- present work, this method has provided a suitable
tal volumetric or mass inflows at any node must equal way for detecting, analyzing, and recording transient
the outflows, less the change in storage. In addition flow (down to 5 milliseconds). Transient flow has
to pressure head, elevation head, and velocity head, been solved for the pipeline in the range of approx-
there may also be head added to the system, for in- imate equations. These approximate equations have
stance, by a pump and head removed from the sys- been solved by numerical solutions of the non-linear
tem by friction. These changes in head are referred to Navier–Stokes equations in a method of characteris-
as head gains and head losses, respectively. Balancing tics “MOC”. So, experiences have been ensured for
the energy across two points in the system yields the the reliable water transmission for the Rasht city main
energy or Bernoulli equation for steady-state flow: the pipeline in northern Iran (research pilot).
680 K.H. Asli et al.
2 Materials and methods Dateline for field tests and lab. Model data collec-
tion was: at 12:00 a.m., 10/02/07, until 05/02/09. Lo-
The method of characteristics “MOC” is based on a cation of work field tests and lab. Model was at Rasht
finite difference technique where pressures are com- city in the northern part of Iran. Pilot subject was: “In-
puted along the pipe for each time step [15]. Two cases terpenetration of two fluids at parallel between plates
are considered for modeling [2, 3]: and turbulent moving in pipe”. For the data collection
(1) The combined elasticity of both the water process, the Rasht city water main pipeline has been
and the pipe walls is characterized by the pressure selected as a field test model. The pipeline was in-
wave speed (Arithmetic method—combination of the cluded in the water treatment plant pump station (in
Joukowski formula and Allievi formula): the start of water transmission line), 3.595 (km) of two
lines of 1200 (mm) diameter pre-stressed cement pipes
Joukowski formula
and one 50,000 (m3 ) water reservoir (in the end of wa-
H2 − H1 = (C/g)(v2 − v1 ) = ρC(v2 − v1 ), (2.1) ter transmission line). All of these parts have been tied
into existing water networks.
Allievi formula The curve estimation procedure was formed by es-
timating regression statistics and producing related
1
c= (2.2) plots. The model summary and parameter estimates
2[ρ (1/k) + (dC1 /Ee))] have been provided a set of results. Also, numeri-
With the combination of the Joukowski formula and cal modeling and simulation which was defined by
Allievi formula: method of characteristics “MOC” have been provided
a set of results. The present work has compared these
λ (∂v/∂t) + (1/ρ)(∂p/∂s) two sets of results (method of characteristics “MOC”
numerical modeling and simulation results against the
+ g(dz/ds) + (f/2D)v|v|
(2.3) curve estimation procedure which was formed by es-
+ C 2 (∂v/∂s) + (1/p)(∂p/∂t) = 0, timating regression statistics). The method of charac-
teristics “MOC” approach transforms the water ham-
λ = +c and λ = − c
mer partial differential equations into the ordinary dif-
Hence, water hammer pressure or surge pressure ferential equations along the characteristic lines de-
(H ) is a function of independent variables (X) such fined as the continuity equation and the momentum
as equation are needed to determine V and P in a one-
dimensional flow system. Solving these two equations
H ≈ ρ, K, d, C1 , Ee, V , g, (2.4) produces a theoretical result that usually corresponds
quite closely to actual system measurements if the data
(2) The method of characteristics (MOC) based on and assumptions used to build the numerical model
a finite difference technique where pressures are com- are valid. Transient analysis results that are not com-
puted along the pipe for each time step, parable with actual system measurements are gener-
ally caused by inappropriate system data (especially
1
HP = C/g(VLe − Vri ) + (HLe + Hri ) boundary conditions) and inappropriate assumptions.
2
Comparisons between the models and validation data
− C/g(f t/2D) VLe |VLe | − Vri |Vri | , (2.5) can be grouped into the following three categories:
1 (a) Cases for which closed-form analytical solutions
VP = VLe + Vri ) + (g/c)(HLe − Hri )
2 exist given certain assumptions if the model can
− (f t/2D) VLe |VLe | + Vri |Vri | , (2.6) directly reproduce the solution, is considered valid
for this case.
f —friction, C—slope (deg), V —velocity (m/s), t— (b) Laboratory experiments with flow and pressure
time (s), H —head (m). data records—The model is calibrated using one
The present work used the method of characteris- set of data and, without changing parameter val-
tics “MOC” to solve virtually any hydraulic transient ues, it is used to match a different set of results. If
problems. successful, it is considered valid for these cases.
Some aspects of physical and numerical modeling of water hammer in pipelines 681
(c) Field tests on actual systems with flow and pres- ing down to one sample. Sensor input analogue sup-
sure data records—These comparisons require plied with quick fit connector, 0–35 (bar), accuracy
threshold and span calibration of all sensor groups, ±0.25% recording up to 8,000,000 readings. Logging
multiple simultaneous datum, and time base memory was programmed to read continuously (cyclic
checks and careful test planning and interpreta- mode) or for a features specific period of time (block).
tion. Sound calibrations match multiple sensor Frequency 1, 5, 10, 20, and 25 samples were per sec-
records and reproduce both peak timing and sec- ond. Logger ID had up to 8 alphanumeric characters.
ondary signals—all measured every second or Communications programmed with Serial RS232 by
fraction of a second. MIL connector for connection to laptop or desktop PC
115,200. Portable ultrasonic flow meter “UFM” with
2.1 Laboratory models pressure transient data logger pressure transient was a
specialized data logger for monitoring rapid pressure
A scale model shows transient flow in a prototype changes in water pipe systems (e.g., water hammer). It
(real) system. It was designed by Kaveh Hariri Asli. was supplied in portable mode only. Pressure transient
Its patent was recorded in the inventions organization loggers were completely waterproof, submersible, and
of Iran (record No.: 44242—date of record: 11/24/07). battery powered for at least 5 years.
It was used for PhD research laboratory tests.
Newton second law for laboratory model
2.2 Laboratory model technical specifications
Fig. 1 Laboratory experiments for Rasht city with flow and pressure data records
the pipeline. However, if a steady-state flow condition approximate equations to numerical solutions of the
is not established because of flow control operations, non-linear Navier–Stokes equations. Hydraulic tran-
then three unknowns need to be determined: H1 (t) sient flow is also known as unsteady fluid flow. For
(the left-hand head), H2 (t) (the right-hand head), V (t) laboratory models of the work, the maximum transient
(the instantaneous flow velocity in the conduit) to de- head envelope calculated (2.8)–(2.15) by rigid water
termine these unknowns, the boundary conditions at column theory (RWCT) is a straight line, as shown
both ends of the pipeline must be known. Using the in the following figure (Fig. 7a). The rigid model has
fundamental rigid-model equation, the hydraulic grade limited applications in hydraulic transient analysis be-
line can be established for each instant. The slope cause the resulting equations do not accurately model
of this line indicates the head loss between the two pressure waves caused by rapid flow control opera-
ends of the pipeline, which is also the head neces- tions. The rigid model applies to slower surge or mass
sary to overcome frictional losses and inertial forces oscillation transients, as defined in “wave propagation
in the pipeline. For the case of flow reduction caused and characteristic time”.
by a valve closure (dQ/dt < 0), the slope is re- L dv dy
duced. If a valve is opened, the slope increases, po- × + y + hf = 0, av = A + Q,
g dt dt
tentially allowing vacuum conditions to occur. The
H2 = H1 + k, L sin θ = k,
change in slope is directly proportional to the flow
change. At fast transients, down to 1 second, surge L d A dy Q
× + + y + hf = 0,
pressure, and velocity of surge wave are found by g dt a dt a
Some aspects of physical and numerical modeling of water hammer in pipelines 683
Fluid power = δ(H − h)V A (2.13) The regression model and “MOC” model showed
L V2 water-column separation and the entrance of air (e.g.,
Head friction = f ⇒ (2.14) Fig. 3) into the pipeline. Field tests focused on the
D 2g
actual system’s model. But in the second case, water-
Fluid power = δV A H − KV h . (2.15) column separation was not what happened. This was
the effect of the air release from the leakage point lo-
2.3 Regression cation. Also, work results showed Max. transient pres-
sure line was completely over the steady flow pressure
The curve estimation procedure allows quick estimat- line Max. The pressure value was 156.181 (m). This
ing regression statistics, and producing related plots pressure was too high for old piping and it must be
for different models (Tables 2–7). Curve estimation is considered as a hazard for piping (for transmission
the most appropriate when the relationship between line with surge tank and in leakage condition).
the dependent variable(s) and the independent variable
is not necessarily linear. Linear regression is used to 2.5 Field test and regression equations definition
model the value of a dependent scale variable based
on its linear relationship to one or more predictors. The main approach of this work was the definition
Non-linear regression is appropriate when the relation- of a model by regression of the relationship be-
ship between the dependent and independent variables tween the dependent and independent data or vari-
is not intrinsically linear. Binary logistic regression is ables. (Numerical modeling and simulation of water
most useful in modeling of the event probability for hammer disaster). The variables are as follows: P —
a categorical response variable with two outcomes. surge pressure (as a dependent variable with nomen-
The auto-regression procedure is an extension of ordi- clature “Y ”), several factors (as independent variables
nary least-squares regression analysis specifically de- with nomenclature “X”) such as: ρ—density (kg/m3 ),
signed for time series. One of the assumptions under- C—velocity of surge wave (m/s), g—acceleration
lying ordinary least-squares regression is the absence of gravity (m/s2 ), V —changes in velocity of wa-
of auto-correlation in the model residuals. Time se- ter (m/s), d—pipe diameter (m), Ep —pipe module
ries, however, often exhibit first-order auto-correlation of elasticity (kg/m2 ), Ew —module of elasticity of
684 K.H. Asli et al.
Fig. 2 Laboratory experiments for Rasht city with flow and pressure data records
water (kg/m2 ), C1 —pipe support coefficient, T — monitoring. Digital uni- or bi-directional pulses were
time (s), Tp —pipe thickness (m). For the data collec- used as instrument powered or non-powered sensors,
tion process, advanced flow and pressure sensors have e.g. PD100 (Up to 128 pulses per second) for internal
been selected as field test data collection. They were pressure transducer Sensor (0–20 bar/0–200 meters
equipped with high-speed data loggers and “PLC” in head/0–300 (psig), repeatability ±0.1%). Input ana-
this work. So, by this ability, they recorded fast tran- logue external pressure transducer (volt) or transmit-
sients data, down to 5 milliseconds. Methods such ter (mA) 0–20 (bar)/0–200 meters head/0–300 (psig),
as inverse transient calibration and leak detection in accuracy ±0.1% 4–20 (mA) from isolated sensor
calculation of unaccounted for water “UFW” used 0–1 (V), 1–5 (V), or 0–100 (mV) Primary recording
such data. For district and zone monitoring, Multi- 48,720 readings (memory expandable to 245,280 read-
Log GPRSTM were used for monitoring flow, pres- ings) were used. Memory were programmed to read
sure, and/or water quality parameters were assessed continuously (cyclic mode) or for a specific period of
for demand, leakage, and conformance. MultiLog time (block). Secondary recording used 6,144 read-
GPRSTM were used to perform dynamic flow and ings, frequency 15 minute sample rate (for hourly data
pressure analysis of water pipeline modeling, particu- download). Optional alarms sent by SMS. Minimum
larly where hourly data were updated for near real time or maximum threshold alarm equipped with persis-
Some aspects of physical and numerical modeling of water hammer in pipelines 685
tence logging alarms factor per channel (7 Alarms per built based on field test data. The curve estimation pro-
logger). Each alarm had out comment field of 16 char- cedure (Figs. 3, 4) was formed by estimating regres-
acters. Features were programmed to auto dial up to 4 sion statistics (Tables 2–7) and producing related plots
telephone numbers on alarm. Logger ID had up to 8 for the field test model with two assumptions:
alphanumeric characters that were programmed with
geography Information System (GIS) number. Clock Assumption (1): p = f (V ).
on board 24-hour real time showed clock with date Assumption (2): p = f (V , T , L).
facility. It was programmed to record either fast data,
Assumption 1 p = f (V ), V —velocity (flow para-
average minimum, average maximum, or channel time
meter) is the most important variable. Dependent vari-
interval between pulses (for data smoothing). Logging
able: P —pressure (bar), for starting point of water
modes counted with Serial RS232 by MIL connector
hammer condition (Table 2). The independent variable
for connection to Rad Link hand held programming
is Velocity (m/s). Regression software “SPSS 10.0.5”
and data collection unit, laptop PC, or desktop PC.
It was programmed up to 19,200 Baud. Communica- performs multi-dimensional scaling of proximity data
tions GPRS (email) typically 1 × email per day to to find least-squares representation of the objects in a
transmitted 1 or 2 channels of communications com- low-dimensional space
pressed data at a 15-minute sample rate. SMS (Text)
with SMS message transmitted on Alarm. GSM (Data) Linear function ∴ pressure
with SIM card was enabled for GSM service, office PC = 6.062 + 0.571 Flow, (3.1)
could establish communications real time, communi-
cations with Logger for reconfiguration, etc. Pressure Quadratic function ∴ pressure
transient logger was supplied with one input for an = 6.216 − 0.365Flow4 + 0.468Flow3 , (3.2)
external pressure transducer.
Cubic function ∴ pressure
= 6.239 − 0.057Flow2 + 0.174Flow, (3.3)
3 Results and discussion
Compound function ∴ pressure
Table 2 Rasht city model summary and parameter estimates (start of water hammer condition)
a The independent variable contains non-positive values. The minimum value is 0.00. The logarithmic and Power models cannot be
calculated.
b The independent variable contains values of zero. The Inverse and S models cannot be calculated. Regression equation defined in
stages (2–3–7–8) is meaningless. Stages (1–4–5–6–9–10–11) are accepted, because their coefficients are meaningful.
Fig. 3 Scatter diagram for Rasht city water transmission lines (start of water hammer condition)
Some aspects of physical and numerical modeling of water hammer in pipelines 687
Regression equation defined in stage (1) is ac- it had water leakage. Field test results showed water-
cepted, because its coefficients are meaningful: column separation and the entrance of air into the
pipeline. Results showed that at point P25:J28 of the
pressure = 28.762 + 0.031Flow − 0.005Distance Rasht city water pipeline, air was interred to the sys-
+ 0.731Time. (3.8) tem. Max. volume of penetrated air was 198.483 (m3 )
and current flow was 2.666 (m3 /s).
3.1 Field tests But in the second case, water-column separation
did not happen. This was the effect of air release from
Numerical modeling and simulation included three the leakage location (Figs. 10, 11). Work results have
cases: The first case was the water pipeline with water shown Max. Transient Pressure line was completely
leakage and equipped with a surge tank. The second over the steady flow pressure line. Max. Pressure was
case was the water pipeline without a surge tank, but 156.181 (m). This was pressure was too high for old
688 K.H. Asli et al.
86 2491 3390 0
86 2491 3390 1
88 2520 3291 0
90 2520 3190 1
95 2574 3110 1.4
95 2574 3110 1.4
95 2574 3110 1.5
95 2590 3110 2
95 2590 3110 2
95.7 2600 3110 2
95.7 2600 3110 3
95.7 2600 3110 4
95.7 2600 3110 5
95.7 2605 3110 0.5
100 2633 2184 1.3
100 2633 2928 1.3
101 2650 2920 1.4
106 2680 1483 1.4
107 2690 1217 1.4
109 2710 1096 1.4
109 2710 1096 1.4
110 2920 1000 1.5
(b) Model summary and parameter estimates
Model Variables Variables Method
entered removed
piping and it must be considered as a hazard for pip- creased from 3014 (l/s) down to Min. value 2520 (l/s)
ing (Rasht city water pipeline transmission line with after 0.6 (s). So, in 0.4 (s), it had grown to 3228 (l/s).
surge tank and in leakage condition). Comparison be- This was the effect of the water release from the leak-
tween three parts (Fig. 11) for cases (Flow-Time & age location. Hence, in one second, 494 (l/s) water
Head-Time & Head-Distance transient curve) proved flows have been interred and exited to the surge tank
the surge tank had an effective role. The flow was de- (for transmission line with surge tank and in leakage
Some aspects of physical and numerical modeling of water hammer in pipelines 689
Table 6 Rasht city regression model summary and parameter showed that there was minus pressure in that zone of
estimates
the transmission line. So, it must be removed from the
Model R R square Adjusted Std. error of system. Max. Transient pressure line was completely
R square the estimate over the steady flow pressure line. Max. pressure in the
system was 156.181 (m). This pressure is to high for
1 0.955a 0.912 0.897 2.283
old piping and it must be considered a hazard for the
2 0.949b 0.900 0.889 2.370
piping system (Tables 8–10).
3 0.000c 0.000 0.000 7.126 The most important points that were observed in
4 0.913d 0.834 0.826 2.973 lab. and field test results:
5 0.949b 0.900 0.889 2.370
Influence of the rate of discharge from local leak to
a Predictors: (constant), time, distance, flow total discharge in the pipeline. It has an effect on the
b Predictors: (constant), distance, flow values of the oscillation’s period and wave celerity.
c Predictor: (constant)
d Predictors: (constant), distance Influence of the rate of discharge from the local leak
on the maximal value of pressure
condition). The surge pressure was 110 (m) nearby
the pump station (i.e. at the start of transmission line). The reason for the high decease in water transmission
The leakage happened near the location of the water pressure was related to the leakage condition in the
treatment plant. So, water flow was decreased from transmission of the Rasht city waterline (local leak-
3000 (l/s) to 2500 (l/s). This was unaccounted for the age effect for a high decrease in water pressure at
water “UFW” alarm. the pipeline) [17]. This was done to explain repeated
This work found the location and rate of unac- pipe breaks. Water hammer has been analyzed in three
counted for water “UFW” in the pipeline. Results have manners: (1) water leakage assumption for transmis-
shown the location and the rate of unaccounted for sion line, (2) no leakage assumption for transmission
water “UFW” in the pipeline (Fig. 11). Min. pressure line, (3) water leakage assumption for transmission
line curve was under the transmission line profile, in line which was equipped by a pressure vessel or a one-
the near of 50,000 (m3 ) water reservoir. Hence, this way surge tank (actual condition). Head-time tran-
690 K.H. Asli et al.
Fig. 5 Analysis & comparison results for calculations of modeling results and field test results Rasht city water pipeline pilot research
sient curve for transmission line without surge tank 3.2 Comparison of present work results with other
and in leakage condition showed 1.2 (s), head value expert’s research
rises from 1.2 (m) to 146 (m), and down to 131 (m) at
1.5 (s). In the end, the transient curve showed 135 (m) Comparison of the present work results (water ham-
at 5 (s). mer software modeling and SPSS modeling), with
Some aspects of physical and numerical modeling of water hammer in pipelines 691
Fig. 6 Flow (l/s) and time (s) records: (a) for total length of pipeline, (b) for leakage effects observation, at water hammer field tests
of Rasht city water pipeline
other expert’s research results show similarity and ad- pipeline with the local leak was considered, the water
vantages: Apoloniusz Kodura and Katarzyna Weine- hammer phenomenon was influenced by some addi-
rowska, 2005: In the present work, water hammer has tional factors [18]. Detailed conclusions were drawn
been run in pressurized pipeline with the local leak. on the basis of experiments and calculations for the
The experiments in Fig. 7a and numerical analysis re- pipeline with a local leak. Hence, the most impor-
sults in Tables 8, 9, 10, 11, 12 were presented. If the tant effects that have been observed were as follows:
692 K.H. Asli et al.
Fig. 7 Rasht city water pipeline (modeling results and field test results, comparison)
The influence of the ratio of discharge from the lo- with the valve at the end of the main pipe, which was
cal leak has been restricted. Total discharge effect in joined with the closure time register. The water ham-
the pipeline on the values of period wave oscillations mer pressure characteristics were measured by exten-
has been investigated. The outflow to the overpres- someters, and recorded in the computer’s memory. The
sure reservoir from the leak affected the value of wave supply of the water to the system was realized with the
celerity (Fig. 8b). use of the reservoir which enabled inlet pressure sta-
bilization.
3.3 Experimental equipment comparison of present The experiments were carried out for four cases:
work results with other expert’s research
• Simple positive water hammer for the straight
Experiments were carried out in the laboratory of the pipeline of constant diameter; the measured char-
Warsaw University of Technology, Environmental En- acteristics were the basis for estimation of the influ-
gineering Faculty, Institute of Water Supply and Water ence of the diameter change and local leak on water
Engineering (Fig. 8b). hammer run.
The physical model is schematically shown • Positive water hammer in pipeline with single
in Fig. 8b. The main element was the pipeline single change of diameter: contraction and extension.
straight pipe of the length L, extrinsic diameter D, and • Positive water hammer in pipeline with local leak in
the wall thickness e, or the pipeline consisted of sec- two scenarios: with the outflow from the leak to the
tions of varied parameters. The pipeline was equipped overpressure reservoir and with free outflow from
Some aspects of physical and numerical modeling of water hammer in pipelines 693
Table 8 Field test model System (water pipeline of Rasht city Table 9 Field test model nodes (water pipeline of Rasht city in
in northern Iran—Software hammer version 07.00.049.00) northern Iran—Software hammer version 07.00.049.00)
Type Keyword Value Category and type Node Elevation (m) Branch pipes
Table 10 Field test model pipes (water pipeline of Rasht city the wave speed distribution along with a pipeline (with
in northern Iran—Software hammer version 07.00.049.00)
node points i = 0, 1, . . . , N ) was given by Lee (1991)
Pipe Length Diameter Velocity Hazen–Williams [19]. In this work, at first was a simulated transient
(m) (mm) (m/s) friction coef. pressure in the system due to an emergency power fail-
ure without any protective equipment in service. After
P3 311 1200 2.21 90 a careful examination of results, it was selected pro-
P4 1 1200 2.65 90 tective equipment and the system was simulated again
P5 0.5 1200 2.65 91 using modeling to assess the effectiveness of the de-
P6 108.7 1200 2.65 67 vices which selected to control transient pressures. At
P7 21.5 1200 2.21 90 present, work analysis and comparison were included
P8 15 1200 2.21 86 in first and second model results. It was shown that
P9 340.7 1200 2.21 90 at point P24:J28 of the water pipeline, air was in-
P10 207 1200 2.21 90 terred to the system. Max. vol. of penetrated air was
P11 339 1200 2.21 90 equal to 198.483 (m3 ) and current flow was equal
P12 328.6 1200 2.21 90 to 2.666 (m3 /s). Treated or modeled air entrainment
P13 47 1200 2.21 90 problems in real prototype systems and results was
P14 590 1200 2.21 90 shown in Tables 10, 11, 12, 13, 14.
P15 49 1200 2.21 90 Consistency between the observed values of max-
P16 224 1200 2.21 90 imal pressure in the first amplitude and correspond-
P17 18.4 1200 2.21 90 ing values were calculated according to Joukowski’s
P18 14.6 1200 2.21 90 formula, irrespective of the rate of discharge from the
P19 12 1200 2.21 90 leak.
P20 499 1200 2.21 90 Significant influence of the rate of the discharge
P21 243.4 1200 2.21 90 from the leak on the vivid decrease of duration of the
P22 156 1200 2.21 90
water hammer phenomenon, which suggests the pos-
P23 22 1200 2.21 90
sibility of utilization of this fact to the pipeline leak
tightness assessment, especially that the duration time
P24 82 1200 2.21 90
decreased with the increase of the outflow from the
P25 35.6 1200 2.21 90
leak. This was the strong reason for the high deceas-
P0 0.5 1200 2.65 90
ing in duration time for Rasht city water pipeline.
P1 0.5 1200 2.65 90
Column separations due to the turned off pump for
Note: Results showed at point P25:J28 of Rasht city water the Rasht city water pipeline were carried out for two
pipeline air was interred to pipeline. Max. Vol. of air was cases:
198.483 (m3 ) and current flow was 2.666 (m3 /s). Data table cre-
ated by hammer—Version 07.00.049.00 compared to the equa- (a) With surge tank and local leakage condition as-
tion of regression software SPSS sumption: In this case, air was sucked into the
pipeline at a negative phase (Fig. 10a).
proposed in the literature in predicting the transient (b) Without surge tank and local leakage condition as-
pressure behavior: the concentrated vaporous cavity sumption: In this case, air was not sucked into the
model (Brown, 1968; Provoost, 1976) [19] and the dis- pipeline at a negative phase (Fig. 10b).
crete air release model (Lee, 1991–1995; Wylie, 1993)
[19]. The concentrated vaporous cavity model pro-
duces satisfactory results in slow transients, but pro- 4 Conclusion
duces unstable solutions for rapid transients, such as
the pump’s stoppage with reflux valve closure. The It is always a good idea to run research to check ex-
discrete air release model produces satisfactory re- treme transient pressures. It is necessary for the sys-
sults in pump shut down cases, but is susceptible to tem with large changes in elevation and long pipelines
long term numerical damping (Ewing, 1980; Jonsson, with large diameters (i.e., mass of water). Also, it is
1985) [19]. Typically, in the discrete air release model, necessary for the initial (e.g., steady-state) velocities
Some aspects of physical and numerical modeling of water hammer in pipelines 695
Table 11 (Continued)
End Upsurge Max. pressure Min. pressure Max. head Min. head
point ratio (m) (m) (m) (m)
Table 12 Field test model pipes (water pipeline of Rasht city in northern Iran—Software hammer version 07.00.049.00)
Node Label Category Type Elev. X-cord. Y -cord. Branch Vapour Maximum Type of Code
ID pipes pressure volume volume
Fig. 8 Scheme of the water hammer experimental equipment (a) Rasht city water hammer laboratory model, (b) water hammer
laboratory model with local leak (Kodura and Weinerowska)
in excess of 1 (m/s). In some cases, hydraulic tran- showed in the leakage condition at a negative phase
sient forces can result in cracks or breaks, even with that air was sucked into the Rasht city water pipeline
low steady-state velocities. In this work, positive wa- (Fig. 10a). There was minus pressure in the zone of the
ter hammer in the pipeline was introduced with local near 50,000 (m3 ) water reservoir. This volume of air
leaks in two scenarios; first, with the outflow from must be removed from the system; it was shown that
the leak to the overpressure reservoir, and secondly, at 110 M surge pressure in the near to pump station
with free outflow from the leak (to atmospheric pres- (start of transmission line). The leakage has happened
sure, with the possibility of sucking air in the nega-
near the location of the water treatment plant. So, wa-
tive phase). Consistency between observed values of
ter flow was decreased from 3,000 (l/s) to 2,500 (l/s).
maximal pressure in first amplitude and correspond-
This was unaccounted for water “UFW” alarm. This
ing values were calculated according to the explicit
method of characteristics. This was related to the rate work found the location and rate of unaccounted for
of discharge from the leak. The discharge rate of the water “UFW” in the pipeline. Results have shown lo-
leakage point, reduced the time duration significantly. cation and the rate of unaccounted for water “UFW” in
So, the duration time was decreased due to the in- the pipeline (Fig. 11). This was the main practical aim
crease of water outflow from the leak. The compar- of the present work. In the final procedure, a condition
ison showed similarity in results between this work base maintenance (CM) method has been found for all
discussion and other expert works. Field test results water transmission systems.
698 K.H. Asli et al.
Fig. 10 Column separations due to the turned off pump—Rasht city water pipeline: (a) with surge tank—pipeline in leakage condi-
tion—pipeline sucked air in the negative phase, (b) without surge tank—pipeline in no leakage condition
Table 14 Modeling of air influence on hydraulic similarity (water pipeline of Rasht city in northern Iran—Software hammer version
07.00.049.00)
Node Label Category Type Elevation X-cord. Y -cord. Branch Vapour Max. Type of Code
ID pipes pressure volume volume
24 J28 Prot equip Air valve 95.2 211.71 35.02 2 −10 198.483 Air 76
700 K.H. Asli et al.
Fig. 11 (a) location and rate of unaccounted for water “UFW”, (b) flow variation, (c) head variation related to “UFW” in Rasht city
water pipeline
The personal contribution of the corresponding au- sponding author) at contract No: 2, dated: 05/02/2007, with the
thor: National Academy of Science of Azerbaijan “AMEA”, Institute
of Mathematics and Mechanics.
All substantive provisions of the work are received
by the authors personally. All parts of the experiments
were made by the authors of the manuscript.
References
Acknowledgements The authors thank Prof. Soltan Aliev,
Prof. Mir Ahmad Lashteneshaee, Prof. Amir Hossien Mahvi,
Dr. Alireza Pendashteh, Dr. Babak Noroozi, and all specialists 1. Asli, K.H., Nagiyev, F.B., Hagi, A.K.: Computational
for their valuable observations and advice, and to the referees for methods in applied science and engineering. In: Interpen-
recommendations that improved the quality of this paper. This etration of Two Fluids at Parallel Between Plates and Tur-
work was a part of the Ph.D. thesis of Kaveh Hariri Asli (corre- bulent Moving in Pipe. pp. 115–128. Nova Science, New
Some aspects of physical and numerical modeling of water hammer in pipelines 701
York (2009). https://www.novapublishers.com/catalog/ 12. Walski, T.M., Lutes, T.L.: Hydraulic transients cause low-
product_ info.php?products_id=10681. Chapter 7 pressure problems. J. Am. Water Works Assoc. 75(2), 58–
2. Joukowski, N.: Paper to Polytechnic Soc. Moscow (1898) 62 (1994)
3. Allievi, L.: General Theory of Pressure Variation in Pipes 13. Wu, Z.Y., Simpson, A.R.: A self-adaptive boundary search
(1902) genetic algorithm and its application to water distribution
systems. J. Hydraul. Res. 40, 191–203 (2002)
4. Parmakian, J.: Water Hammer Analysis. Dover, New York
14. Arturo Leon, S.: Improved modeling of unsteady free sur-
(1963)
face, pressurized and mixed flows in storm-sewer systems.
5. Wood, D.J., Jones, S.E.: Water hammer charts for vari- Submitted in partial fulfillment of the requirements for the
ous types of valves. J. Hydraul. Div., Proc. ASCE 167–178 degree of Doctor of Philosophy in Civil Engineering in the
(January, 1973) Graduate College of the University of Illinois at Urbana-
6. Wylie, E.B., Streeter, V.L.: Fluid Transients. FEB Press, Champaign, pp. 57–58 (2007)
Ann Arbor (1993) 15. Joukowski, N.: Paper to Polytechnic Soc. Moscow, Spring
7. Brunone, B., Karney, B.W., Mecarelli, M., Ferrante, M.: of 1898. English translation by Miss O. Simin. Proc.
Velocity profiles and unsteady pipe friction in transient AWWA (1904)
flow. ASCE J. Water Resour. Plan. Manag. 126(4), 236– 16. Asli, K.H., Nagiyev, F.B., Hagi, A.K.: Physical mod-
244 (2000) eling of fluid movement in pipelines. In: Nanomateri-
8. Koelle, E., Luvizotto, E. Jr., Andrade, J.P.G.: Complete op- als Yearbook—2009. From Nanostructures, Nanomaterials
erational simulation of pumped storage schemes. In: Wa- and Nanotechnologies to Nanoindustry, pp. 210–225. Nova
ter Power95—International Conference on Hydropower, Science, New York (2009). https://www.novapublishers.
vol. 3, pp. 2403–2412 (1995) com/catalog/product_info.php?products_id=11587. Chap-
ter 17
9. Filion, Y.R., Karney, B.W.: Extended period analysis with
17. Wylie, E.B., Streeter, V.L.: Fluid Transients. FEB Press,
a transient model. ASCE J. Hydraul. Eng. 128(6), 616–624
Ann Arbor (1983). Corrected copy: 166–171 (1982)
(2002)
18. Kodura, A., Weinerowska, K.: Some Aspects of Physical
10. Hamam, M.A., McCorquodale, A.: Transient conditions in and Numerical Modeling of Water Hammer in Pipelines,
the transition from gravity to surcharged sewer flow. Can. pp. 125–133 (2005)
J. Civil Eng. 189–196 (1982) 19. Lee, T.S., Pejovic, S.: Air influence on similarity of hy-
11. Savic, D.A., Walters, G.A.: Genetic algorithms techniques draulic transients and vibrations. ASME J. Fluid Eng.
for, calibrating network models. Report No. 95/12, Centre 118(4), 706–709 (1996)
for Systems and Control Engineering (1995)