Major Water Distribution Project and Leak Detection Modeling
1. Major and Final project presentation
Submitted at Budapest University of Technology and
Economic, Department of Hydrodynamic Systems
Mahbod Shafiei
HHSCIQ
2. Introduction
The network of distribution mains is nearly the most
expensive item of equipment in a water undertaking
The analysis of a pipe network can be one of the most
important part in designing ,maintenance and optimization
of the underground network system
few basic principles of fluid mechanics' have been used
Conservation of mass or continuity principal
The work-energy principal
The relation between fluid friction and energy dissipation
Simulation and modeling has tried to optimize and
monitoring data ‘s
3. Main goal of simulation
Distribute water from reservoir to customers in looped
network system through economic pipeline with desired
pressure
4. Structure of the modeling
Population projection
Estimate water demand
Capacity of storage tank
Build up the model
Setting the input data
Running the model
Analyzing data’s
5. Population projection
Projected population requires certain information on:
1. Historic population counts
2. Birth
3. Deaths and other rates which affect population change.
Using various mathematical method to estimate future
population in sample area
Budapest historic data's have been used
constant growth rate method is selected for this project
Considering percent of growth for target area
Budapest population in 2010 1721556
Budapest population in 2030 1916540
6. Water demand
Considering Budapest water demand trend during last years
Correlation coefficient between population and water
demand
Positive correlation between 1991 to 2006 and negative one
between 2007 to 2010 which mean by increasing population
the water demand has decreased.
Considering water demand dependency to temperature
Estimating the months which have highest consumption in
summer and winter both from 1990 to 2010
Estimating water demand by Q(Daily) or Q(hourly) and
Q(yearly)
8. Estimating capacity of storage tank
Estimating demand for fire works
126 lit per person per day
Calculation R (ratio between production and consumption)
Which is 0.8
Capacity of storage = R.Q(daily)+fire demand
Capacity of storage tank=193200 m^3/hr in 2011
Estimating capacity of storage tank for sample area by
evaluation population and water demand, which is around
8300 m^3
9. Build up the model
Effective area: 525.16 sq Km
Population density in 2011: 3301.3 per sq Km
Population density in 2025: 3649.4 per sq
Km
13. Running the first model
Getting the results for the first model
such as pressure, velocity in each link, flow type and friction
loss
Define a new goal, increasing pressure in network nodes
without pump station
Lead to low energy consumption
Definition of extra loop , extra mass from the nodes with
high pressure to nodes which have lower pressure by
considering allowable velocity in each link
Pressure in each node decrease in upper hand nodes but in
farthest nodes is around 2.8 bar
Change the pipe diameter s to get better results for velocity
15. Comparison between the results
Pressure has increased in all branches
Velocity has decreased but still in allowable limit
Total pressure drop has decreased
18. Daily and night demand
Finally daily and night demand with 80% and 20% of
maximum demand has been calculated for the model
Overload has been researched for the model with 1.2% of
maximum demand without extra loop
Vacuum has occurred in some nodes (farthest ones)
Vacuum can occur as a results of intense fire fighting
Recalculating the model with extra loop shows different
results
22. conclusion
simulation with pump station has done
The results has been reported in major project report
Comparison between real case in Budapest with simulation
has been reported
In real Budapest network each loop has been connected to
each other by higher diameter pipe
23. Final project
Leak phenomena and leak detection
45 million cubic meters are lost daily through water
leakage in the distribution networks which is enough
to serve nearly 200 million people
EPANET software has been used to simulate and
monitoring data ‘s for leakage simulation
Support vector machine (SVM) has been used to
analyze monitored data’s and report results
Summery of leak detection has been reported ,Such as
acoustic method, Computer base method and etc…
24. Leak definition
Simulate leak as an orifice area
Define emitter coefficient
Q : flow rate P internal pressure γ unit weight of
water P internal pressure Cd discharge coefficient
Above formula leads to emitter coefficient for
simulation
0.5 is Pressure exponent for whole loop
27. Demand multiplier
24 hour hydraulic time step
Define demand multiplier based on real data’s for various
time step during 24 hour of a day
28. Emitter coefficient Selection
Examine computer based method , I tried to use variety
of leak flows rates between 0.87 m^3/hr to 8.76 m^3/hr
By assuming 0.5 pressure exponent
Emitter coefficient 0.01 to 0.1 by step 0.01
Monitoring pressure and flow rate on the nodes lower and
upper hand of candidate leak node E
Select radius around leak node by around 1200 meter
radius
Monitoring pressure and flow rate at three time step
according to maximum and minimum demand which are
8:00 am 14:00 pm and 21:00 pm
29. Table for leak flow rate and orifice
area and emitter coefficient
32. Table of flow rate 2
Form data’s it can be interpreted that percentage
differences has increased by leak rate and it has been
increased slightly when distance has closed to leak point
33. Pressure monitoring
Pressure difference has shown with different emitter
coefficient’s in three hydraulic time steps
Normally pressure has decreased through upper to
lower nodes from leak condition node (E)
To show better results , difference percentages has been
shown ,which can help to interpret better to leak node
The results shown that, pressure difference percentage
has increased from upper nodes to leak node (E) and
then slightly remain constant by lower hand nodes
35. Leak location estimation
Finding correlation coefficient between flow rate or
pressure , and distance to leak node (E)
Show direction of relationship (+1 or -1)
Finding regression line ( the best fit of data’s on scatter
plot)
Finally finding Standard Error of Estimate
Define radios around leak node (E) with radios around 1
km
43. Interpreting results
As can be seen from the above tables , percent of
differences in flow rate cases have much higher values
in compare with pressure cases
Standard error of estimate have higher value in flow
rate analysis than pressure one , moreover , in pressure
case at 8:00 am it has lowest value than other two , due
to higher demand at this time
Leak predicted distance , has much lower error with
pressure data’s especially at 8:00 am in compare with
flow rate data’s
Flow rate in links may give better results in leak finding
and pressure monitoring have better results in leak
location estimation
44. Leak flow rate and orifice area
Correlation between emitter coefficient and orifice area
has been calculated
Leak flow rate , correlation with emitter coefficient’s and
orifice area (mm^2) have been reported
45. Modeling breakage in pipe line
Breakage is a fail in pipe line which doesn’t lead liquid to
lower hand side of failure point ,but in pipes with higher
diameter ,bigger orifice area leads to higher rate of
leakages and its physically equal to breakage
Detection of breakage is nearly easy due to the fact that
there are some out signs
Sometimes there is ness on the breakage location when
pipe size and flow rate is small
In modeling , we want to know how much pressure will
drop in other links or branches
Find a way to reduce the radius of breakage effect to other
links and branches
48. Simulation of Breakage
in the first model the extra loop is out of service and out
pressure happen with 1 bar
Comparison with the case when extra loop is in service