Beruflich Dokumente
Kultur Dokumente
state equation in terms of a drop pressure head between 2.2 Leak modeling
the pipeline ends, and also it can be computed via sensors
measurements. Finally for each possible node, a residual A leak that appears at position zl ∈ (0, L) can be handled
is generated. The leak-node is identified when the residual as a new boundary condition in (1) and (2). A leak’s
is the smallest. Notice that, this algorithm is limited to flow √model is derived from Torricelli’s equation: Ql =
the knowledge of the pipeline configuration a priori and Cd Al 2gHl , in which Cd is the discharge coefficient, Al
it becomes specially important for water pipeline systems is the leak cross-section area, Hl is the pressure head at
installed underground because uncertainties in the leak lo- the leak point and Ql is√the flow rate at the leak orifice.
cation provided by others methods as in Delgado-Aguiñaga Now by using λ = Cd Al 2g. The Outflow through a hole
and Begovich (2017) are not longer important. can be modeled as:
This work continues as follows: In Section 2 dynamics of
a fluid inside a pipeline with leaks is presented. Section Ql = λ H l (6)
3 presents the steady-state algorithm. Section 4 provides In order to obtain a spatial-discretized approximation from
some results on the basis of simulations. Finally Section 5 (1) and (2), the Finite-Difference method is used due to
concludes this paper. its simplicity and low-cost computational effort. Also, by
considering boundary conditions (5), a finite-dimensional
approximation with n sections can be written as Chaudhry
2. PIPELINE DYNAMICS (2014):
The nonlinear model is derived considering the following −gA
Q̇i = (Hi+1 − Hi ) − µi Qi |Qi |; ∀i = 1, ..., n (7)
conditions: the fluid is slightly compressible, the pipeline ∆zi
is straight, without any fitting and any slope, the duct
−b2
wall is slightly deformable, the convective velocity changes Ḣi+1 = (Qi+1 − Qi ); ∀i = 1, ..., n − 1 (8)
are negligible, the cross section area of the pipe and the gA∆zi
fluid density are constant. Then, the nonlinear Hyperbolic where ∆zi represents the length of the i section, i.e.,
Partial Differential Equations which describe the fluid ∆zi = zi − zi−1 , with z0 = 0 and zn = L the external
dynamics in a pipeline are as in Chaudhry (2014): positions and n − 1 interior nodes which can represent
Momentum Equation n − 1 leak occurrences through (6). Now, let us consider
∂Q(z, t) ∂H(z, t) the single leak case, by applying (7) in (8) with (n = 2),
+ gA + µQ(z, t)|Q(z, t)| = 0 (1) the following state space model can be obtained:
∂t ∂z
−gA
Continuity Equation Q̇1 = ∆z1 (H2 − Hin ) − µ√
1 Q1 |Q1 |
−b2
gA∆z1 (Q2 − Q1 + λ H2 )
∂H(z, t) b2 ∂Q(z, t) Ḣ2 = (9)
+ =0 (2) −gA
∂t gA ∂z Q̇2 = ∆z2 (Hout − H2 ) − µ2 Q2 |Q2 |
where H(z, t) is the pressure head [m], Q(z, t) is the flow where ∆z1 = zl (zl denotes the leak position) and ∆z2 =
rate [m3 /s], t is the time coordinate [s], z is the length L − zl . In the following section a steady-state expression
coordinate [m], g is the gravity acceleration [m/s2 ], A is is derived from (9) in order to design the leak diagnosis
the cross-section area [m2 ], b is the pressure wave speed system.
in the fluid [m/s] and µ = f (Q)/2DA, with f (Q) is the
friction factor coefficient depending on flow rate and D is
the inner diameter [m]. In this work, the variable z ∈ [0, L] 3. STEADY STATE ALGORITHM
and L is the pipeline length [m].
The idea to detect a leak is as follows: once that
|Qin (t) − Qout (t)| > β a leak is detected, for any β [m3 /s]
2.1 Friction modeling
chosen small enough such that it provides a good trade-off
between leak detection (leaks around 4% of nominal flow)
Friction coefficient is computed by means of the Swamee- and avoiding false alarms caused by noisy measurements.
Jain approximation as Swamee and Jain (1976): On the other hand, the idea to isolate a leak is as follows:
0.25 Let us take advantage of that each possible leak can appear
f (Q) = ε 5.74 2 (3) at any junction between two consecutive pipe sections,
[log10 ( 3.7D + Re0.9 )]
possibly being of the same size and no matter if they are
where ε is the roughness of the pipe [m] and Re is the curved, in accordance with the experience of the staff of
Reynolds number computed as follows: SIAPA.
VD To do that, one can build a model by using only two
Re = (4) sections (n = 2 by using (7) and (8)) in order to represent
v(t)
the single leak case. In this way, one can introduce a
where v(t) =ce(a/Tk (t)) is the kinematic viscosity [m2 /s], new pipe section variable as follows: ∆ζi = ζi − ζ0 for
for some a and c being constant. Tk (t) the water temper- i = 1, ..., n − 1, ζi = zi and ζ0 = z0 fixed, see Fig. 1. Note
ature in [o K] assumed to be also constant. V is the flow that in this new representation the section size ∆ζi divides
velocity [m/s]. Finally, boundary conditions for 1 and 2 the pipeline only in two sections and its size depends on the
are here considered to be: index i. This allows to represent all possible leak positions
H(z = 0, t) = Hin (t) in the pipeline (sweeping all possibilities).
(5)
H(z = L, t) = Hout (t)
403
2018 IFAC MICNON
404
Guadalajara, Mexico, June 20-22, 2018 A. Lizarraga-Raygoza et al. / IFAC PapersOnLine 51-13 (2018) 402–407
4. SIMULATION RESULTS
404
2018 IFAC MICNON
Guadalajara, Mexico, June 20-22, 2018 A. Lizarraga-Raygoza et al. / IFAC PapersOnLine 51-13 (2018) 402–407 405
20
0.2
18 Hin
Pressure head [m]
16 Hout
0
14
12
0 50 100
10
Time [s]
8 Fig. 5. Errors e∆Hi , case of a leak at ∆zl ≈ 17 [m].
0 50 100
Time [s] Notice that all the errors are closer to zero before the leak
occurs t < 45 [s] which can be explained as follows: since
Fig. 2. Hin , Hout , case 1. there is no leak then Q1 = Q2 is satisfied and therefore
µ1 = µ2 as well. Now, by rearranging (11) it results:
µ1 Q21 ∆ζi
e∆Hi ≈ (∆H) − gA (∆ζi + (L − ∆ζi )). As a result
×10−3 the ∆ζi terms are annulled and no matter which one is
9.4
chosen the error will always be the same.
Qin
9.2
Flow rate [m 3/s]
e2 ∆H
∆ζ1 71.6391
10.6 ˆ 1
∆H ∆ζ2 2.7967
ˆ 2 ∆ζ3 43.1083
∆H
Pressure head [m]
10.4 ˆ 3
∆H
ˆ
∆H 20 ×10−3
10.2 18 Hin 9.2 Qin
Pressure head [m]
Hout
Flow rate [m 3/s]
10 16 Qout
14 9
9.8 12
8.8
10
9.6
0 50 100 8 8.6
Time [s] 0 50 100 0 50 100
Time [s] Time [s]
Fig. 4. Drop pressure head between the pipeline ends, case
of a leak at ∆zl ≈ 17 [m]. Fig. 6. Hin , Hout , case 2. Fig. 7. Qin , Qout , case 2.
405
2018 IFAC MICNON
406
Guadalajara, Mexico, June 20-22, 2018 A. Lizarraga-Raygoza et al. / IFAC PapersOnLine 51-13 (2018) 402–407
10.4 20 ×10−3
9.2
ˆ 1
∆H 18 Hin Qin
10.2 9
ˆ 3
∆H 14
ˆ
∆H
12 8.8
10
10
8.6
8
9.8 0 50 100 0 50 100
Time [s] Time [s]
9.6 Fig. 10. Hin , Hout , case 3. Fig. 11. Qin , Qout , case 3.
0 50 100
Time [s]
10.2
Fig. 8. Drop pressure head between the pipeline ends, case
0.4 9.8 ˆ 1
∆H
|e2 ∆H1 | ˆ 2
∆H
0.3 |e2 ∆H2 | 9.6 ˆ 3
∆H
Pressure head [m]
|e2 ∆H3 | ˆ
∆H
0.2 9.4
0 50 100
Time [s]
0.1
Fig. 12. Drop pressure head between the pipeline ends,
case of a leak at ∆zl ≈ 61 [m].
0
0.5
0 50 100
|e3 ∆H1 |
Time [s] 0.4 |e3 ∆H2 |
Pressure head [m]
406
2018 IFAC MICNON
Guadalajara, Mexico, June 20-22, 2018A. Lizarraga-Raygoza et al. / IFAC PapersOnLine 51-13 (2018) 402–407 407
heart of this proposition. It could be applied in real cases Navarro, A., Begovich, O., Sánchez, J., and Besançon,
since the configuration of pipelines is often known, that is, G. (2017). Real-time leak isolation based on
normally those systems are made by using same-size pipe state estimation with fitting loss coefficient calibra-
sections. Although this algorithm is simpler than others tion in a plastic pipeline. Asian Journal of Con-
already available in the literature, its main advantage is trol, 19(1), 255–265. doi:10.1002/asjc.1362. URL
that the leak position is given without uncertainties, since http://dx.doi.org/10.1002/asjc.1362. Asjc.1362.
the identified-leak-node has a location known a priori. This Ostapkowicz, P. (2014). Leakage detection from liquid
is of particular interest for pipelines placed underground transmission pipelines using improved pressure wave
since other approaches could estimate the leak position technique. Eksploatacja i Niezawodność Maintenance
but with uncertainties which could generate extra costs and Reliability, 16, 9–16.
and waste of time in the repair process. In addition this Swamee, P. and Jain, A.K. (1976). Explicit equations for
method is not limited to deal with pipelines made of a pipeflow problems. Journal of the hydraulics division,
large number of sections. Finally applications with real 102(5).
databases will be part of future developments. Verde, C. and Torres, L. (2015). Referenced model-
based observers for locating leaks in a branched
ACKNOWLEDGEMENTS pipeline. IFAC-Papers on-line, 48(21), 1066 – 1071.
doi:http://dx.doi.org/10.1016/j.ifacol.2015.09.668. 9th
The first author would like to thank to CONACYT for the IFAC Symposium on Fault Detection, Supervision and
scholarship granted during his M.Sc. studies. Safety for Technical Processes SAFEPROCESS 2015,
Paris, 2-4 September 2015.
REFERENCES wei Liu, C., xing Li, Y., kun Yan, Y., tao Fu, J., and gian
Zhang, Y. (2015). A new leak location method based on
Begovich, O., Pizano-Moreno, A., and Besançon, G. leakage acoustic waves for oil and gas pipelines. Journal
(2012). Online implementation of a leak isolation al- of Loss Prevention in the Process Industries, 35, 236 –
gorithm in a plastic pipeline prototype. Latin American 246. doi:http://dx.doi.org/10.1016/j.jlp.2015.05.006.
Applied Research, 42(2), 131–140. Zhang, T., Tan, Y., Zhang, X., and Zhao, J. (2015).
Begovich, O. and Valdovinos-Villalobos, G. (2010). DSP A novel hybrid technique for leak detection and lo-
application of a water-leak detection and isolation al- cation in straight pipelines. Journal of Loss Preven-
gorithm. In 7th International Conference on Electrical tion in the Process Industries, 35, 157 – 168. doi:
Engineering Computing Science and Automatic Control http://dx.doi.org/10.1016/j.jlp.2015.04.012.
(CCE), 2010, 93–98. IEEE.
Billmann, L. and Isermann, R. (1987). Leak detection
methods for pipelines. Automatica, 23(3), 381–385.
Brunone, B. and Ferrante, M. (2001). Detecting leaks in
pressurised pipes by means of transients. Journal of
Hydraulic Research, 39(5), 539–547.
Carvajal-Rubio, J., Begovich, O., and Sánchez-Torres, J.
(2015). Real-time leak detection and isolation in plastic
pipelines with equivalent control based observers. In
12th International Conference on Electrical Engineer-
ing, Computing Science and Automatic Control (CCE),
2015, 1–6. IEEE.
Chaudhry, M.H. (2014). Applied Hydraulic Transients.
Springer-Verlag New York.
Delgado-Aguiñaga, J., Begovich, O., and Besançon, G.
(2016). Exact-differentiation-based leak detection and
isolation in a plastic pipeline under temperature varia-
tions. Journal of Process Control, 42, 114 – 124. doi:
https://doi.org/10.1016/j.jprocont.2016.04.005.
Delgado-Aguiñaga, J. and Begovich, O. (2017). Modeling
and Monitoring of Pipelines and Networks, chapter
Water leak diagnosis in pressurized pipelines: A real case
study, 235–262. Springer International Publishing.
Delgado-Aguiñaga, J., Besançon, G., and Begovich, O.
(2015). Leak isolation based on extended Kalman filter
in a plastic pipeline under temperature variations with
real-data validation. In 23th Mediterranean Conference
on Control and Automation (MED), 2015, 316–321.
IEEE.
Li, S., Song, Y., and Zhou, G. (2018). Leak detec-
tion of water distribution pipeline subject to failure
of socket joint based on acoustic emission and pat-
tern recognition. Measurement, 115, 39 – 44. doi:
https://doi.org/10.1016/j.measurement.2017.10.021.
407