LEAK ANALYSIS IN PIPES USING TRANSIENTS



WS2

Leak Analysis in Pipes Using TransientS

B. Brunone and M. Ferrante

Department of Civil and Environmental Engineering, University of Perugia

L. Ubertini

Department of Hydraulics and Transportation, University of Rome “La Sapienza”

1. INTRODUCTION

Many problems face engineers in the design and maintenance of water distribution networks. These problems can be grouped into two types: network characteristic definition problems, and problems associated with computer hydraulic modeling. The network characteristic definition problems involve physical parameters such as unknown pipe and valve layouts, unauthorized users of water, and pipes deteriorating and forming leaks. Computer hydraulic modeling can pose uncertainty, not necessarily with the accuracy of the software package, but rather with the confidence in parameters used in the models. Some of the largest issues in water distribution are calibration of models and detecting leaks.

It has been estimated that, as an example, within the Asia-Pacific region, as well as in most European countries, at least 25 % of the water supply is being lost from water distribution networks. This is very costly for both the water utility who directly loses money and the consumer who is indirectly penalized through increases in water rates. There is no completely satisfactory method for finding leaks or unauthorized use of water at present without costly meter reading and exact quantification of flows. Although there have been methods in the past that can quantify and locate leaks, they have only really been for single long pipelines rather than networked systems.

The aim of the present paper is to show that leak detection and evaluation in pipe systems can be reached by means of unsteady-state tests that minimize both time consuming and costs for Water Supply Companies as well as minimize the risks for the integrity of conduits.

In the paper, first the traditional techniques for leak detection are described; second, principles on which the proposed technique is based are analyzed; third numerical and experimental evidence of the effects of leaks on pressure transient are shown.

2. “TRADITIONAL” TECHNIQUES FOR LEAK DETECTION

The detection of leaks in pipe networks is an important issue. The presence of leaks in a network will lower the hydraulic efficiency of that network. In most countries water is a finite resource and leaks deplete that store. This usually results in a higher cost of water to the population. The detection of leaks can also find unauthorized use of water whereby people are stealing this valuable resource. In pipe networks that transport potentially hazardous fluids, the detection of leaks can stop or minimize any contamination of the environment. Leaks can be a precursor to failure of pipes if they are subjected to extra pressure events. There are many different techniques to prove the existence and location of leaks, with most coming from the field of oil and gas where the consequences of leaks are the most severe.

2.1 Hydrostatic Method

The traditional method of determining and locating leaks is to test each length of pipe line as it is layed. This involves pressurizing the pipe with a low-cost liquid, usually water. Pressure sensors are then positioned in the pipe to see if there is any reduction of pressure that could mean the presence of a leak. This is a good technique because leaks can be found when the pipe has just been placed and the leaking pipe can be easily replaced with a non-faulty member. The only problem with this method is that it cannot easily be done with existing pipe networks, due to the fact that part of the network has to be shut down. Another traditional technique is just to measure the inflow and compare it with the outflow. If there are any differences there must be leaks. It should be said that this is a very cost efficient way to show the existence of leaks. A drawback with this is that the number and location of leaks are unknown, thus tracking them down can be a problem.

2.2 Acoustical Method

This procedure uses the correlation between acoustic velocity and pressure in a liquid filled pipeline to detect leaks. This method has been used to test the integrity of pipelines in the oil and gas field. The manner in which this technique is set up is to have at one end of the pipeline a pressure gauge and a few meters apart from it a sound transmitter. The sound transmitter generates signals in the range of 2 to 1,000 Hz with an amplitude of about 1 kPa. The signal travels down the pipe then the transducer detects and records the echo. Using the time difference between the outgoing and incoming signals, the acoustic velocity can be calculated. Readings of both pressure and acoustic velocity are then taken at about half hour intervals so a graph can be drawn up. If there are no leaks the acoustic velocity curve will mimic the pressure curve, if there is a leak the acoustic velocity curve will link, and not correlate after the leak occurs. The leakage rate can also be estimated based on the difference between the acoustic velocity and the pressure and the measured pressure change/volume change.

The advantage in using this technique is that there are no temperature measurements needed, not like other hydrostatic methods (especially for gas pipelines) where multiple temperature readings are needed to check if the pressure drops can be attributed to temperature drops. Also the equipment can be set up at just one end of the pipe meaning, that data telemetry is not needed for very long pipes and other pipes do not need to be excavated to install sensors. Disadvantages of this technique are few, but some are very important. One of the most important is the presence of air in a pipeline. This reduces the bulk modulus and density of the liquid and can cause large reductions in the acoustic velocity. Another is the amount of suspended solids in the liquid. Suspended solids will increase the density of the liquid and cause a decrease in the acoustic velocity. These decreases in the acoustic velocity can be initially interpreted as leaks. The last problem is noise that can be detected by the transducer which may come from the environment or electrical equipment. While the method detects the presence of a leak, the location cannot be determined. This is not a problem when testing the integrity of pipes or connections, but is a drawback if the leak is to be repaired.

2.3 Inverse Steady-State Method

The use of inverse steady state analysis to find leaks was carried out by Pudar and Liggett (1992). The data consisted of flows and pressures at various positions in the network. The approach behind the analysis was to set up the steady state equations and include a leak area at nodes. An orifice equation was applied to calculate flow due to pressure loss. An objective function was then created which was the sum of the squares of the differences in measurable properties such as pressures and flows. The transition then followed a minimization process to fit the leak areas to the measured values. Pudar and Liggett concluded that inverse steady state leak detection would only be commercially possible if there was massive data for a system.

3. THE PROPOSED TECHNIQUE

The technique here proposed for leak detection is part of the class of the Transient Analysis Methods which are based on the properties of transmission and reflection of pressure waves (Brunone 1989, 1999; Brunone and Ferrante 1999; Jönsson 1994, 1995; Liggett and Chen, 1994).

When a pressure wave due to a change of the flow propagates in an intact pipe, it is modified by friction forces which cause the progressive damping and smoothing of pressure peaks (e.g. Brunone et al. 1995). Devices, changes in pipe characteristics, sediment or air accumulation, and leaks all abruptly affect the characteristics of the pressure wave (Brunone 1989, 1999; Brunone and Ferrante 1999; Jönsson 1994, 1995). Specifically, a leak has three effects on the resultant pressure graph. The first effect is when the initial pressure wave reaches the leak. It will produce a reflection which can be seen when it arrives back at the section where the pressure wave has been determined. The amplitude of the reflected wave increases with increasing the dimension of the leak. The time difference between the initial transient wave and the reflected wave can then be measured and the location down the pipe can be calculated by the following equation:

[pic] (1)

where: s( = distance to the leak from the section where the pressure wave has been produced,

(T( = time difference between the initial transient wave and the reflected wave,

a = wave propagation velocity.

Moreover, by considering partial reflection of pressure waves which takes place at a leak, the discharge behavior of the leak itself can be assessed.

The second effect is the distortion in the oscillatory peaks and troughs that occur after the main pressure rise has occurred.

The third and last effect due to the leakage is a reduction in the amplitude of the oscillatory pattern after the main pressure rise has occurred.

Thus, for a given change of the flow, the presence of a leak can be detected by analyzing the pressure signal, that is the pressure time-history at a section of the pipe. Of course, the larger and the faster the flow change which gives rise to transient, and then the larger and the steeper the pressure wave, the better the leak detection. The measured pressure signal can be compared with the predictions of a numerical model for the same transient assuming the pipe to be intact. Alternatively, experimental values previously obtained, such as when the pipe was just built, can be taken into account.

4. NUMERICAL AND EXPERIMENTAL EVIDENCE OF THE EFFECTS OF LEAKS ON PRESSURE TRANSIENT

Numerical modeling

Firstly, the possibility of locating a leak on the basis of transient tests is shown by discussing the results of numerical simulations; details of the numerical model are given in (Brunone 1989, 1999) as well as in Appendix I.

Example refers to an iron outfall pipe with diameter D = 300 mm, length L = 1000 m, characteristic roughness ( = 0.4 mm (Colebrook-White equation), a = 1250 m/s. Transients are caused by an increase [pic]of the piezometric head in the initial section created, for example, by a pressurized tank truck. The pressure signal, hm, is shown (Fig. 1) at the measurement section (s = sm = 50 m) downstream of the controlled valve whose action gives rise to the transient ([pic]= 5 m). In the graph, the bold line represents an intact pipe, whereas the thin lines refer to a damaged pipe where, at the section [pic]= 400 m, a leak of different discharge behavior is located: [pic]= 20, 80 and 180 mm2, respectively, with [pic]and[pic]being the coefficient of discharge and the area of the leak. Furthermore, the dimensionless time ( = t/( is considered, with t being the time since the beginning of the valve motion and ( = 2L/a. Fig. 1 shows that for the intact pipe, hm is practically constant and equal to [pic]in the time interval [(1, (3], with (1 = sm/(2L), the time at which the pressure wave approaches section m, and (3 = 1 - sm/L, the time when the pressure wave, reflected back at the end diffuser, arrives at section m. For a damaged pipe, by contrast, at (2 = s( /L - (1, a reduction y( of hm occurs, an effect which increases with increasing [pic].

Laboratory Tests

With respect to the case of an intact pipe, numerical simulations confirm that, when a pressure wave arrives at the orifice node, a part of it is transmitted and the other part is reflected back, with the reflection phenomenon depending on the characteristics of the leak. Therefore, as already mentioned, the pressure signal conveys information about the state of the system: not only the presence of a leakage, but also its location and size.

An experimental apparatus has been designed and constructed to investigate unsteady-state flow processes in pressurized pipes in the Laboratory of Hydraulics of the Department of Water and Structural Engineering of the University of Perugia, Italy. The apparatus consists of (Fig. 2):

i) upstream pressurized tank (air vessel) in which the pressure can be held nearly constant; in fact, during tests the speed of two submerged pumps, drawing water from the recycling reservoir, can be varied;

ii) polyethylene pipe, 352 m in length of internal diameter 93.8 mm and wall thickness of 8.1 mm; the pipe is arranged in concentric circles with bends having a minimum radius equal to 1.5 m and is almost horizontal, except for the last short part;

iii) downstream hand operated quick-closure ball valve that discharges into a free surface tank.

In order to simulate a leak, a device with a variable diameter orifice at its wall has been used (Fig.2). Pressure has been measured at the upstream tank and at the downstream end section of the pipe (the measurement section in Fig. 2). Pressure transducers are of the strain-gauge type, with a recording range of 0 to 60 m of water, an accuracy of 0.5% of the full-range scale and a time response of 50 ms. Output signals from pressure transducers are read directly into a PC at a rate of 130 Hz.

Transient events have been caused by quick closure of the end valve. The device simulating the leak has been placed at a distance s( = 128.12 m from the end section of the pipe. In tests, different values of the orifice diameter have been considered in order to evaluate the effects of the dimension of the leak on the characteristics of the pressure signal, for a given shape of the leak. Furthermore, some tests have concerned an intact pipe with the orifice simulating the leak closed by a disk. As a result of the pressure signal analysis extended to all tests, the mean value of the velocity of the water hammer wave a = 359.9 m/s has been estimated; small differences in the a value from test to test can be explained with the existence of small trapped air volumes.

To allow detailed discussion, only four of the experimental results from the laboratory apparatus are chosen as entirely typical of the remainder. The related traces of pressure signal, hm, observed at the measurement section are shown in Figs. 3 to 6, where the reference datum for hm is the floor of the laboratory and t is time since data acquisition.

Transient of Fig. 3 corresponds to the intact-pipe condition. Thus, it can be assumed as a reference for no-leak conditions during transients.

Transients of Figs. 4 to 6 correspond to damaged-pipe conditions. In these plots it is evident that the pressure wave reflected at the leak modifies the pressure signal giving rise to a dent. Furthermore, the presence of the leak causes a faster damping of the pressure signal.

In Table I, the diameter d of the orifice simulating the leak along with the experimental values of (hm, y( and (t( are given, where (hm = maximum value of the overpressure due to the closure of the valve; y( = piezometric head reduction (dent) due to the leak; (t( = t1 – t0, with t0 = time at which closure begins and t1 = time when the reflected wave arrives at the measurement section. As can be seen from Table I, in order to make easier the evaluation of the effects of the leak, the value of (hm is almost the same for the chosen transients. As expected, for a given (hm, the larger d, the larger the dent y(.

Table I. Characteristics of transients

|Test |d |(hm |y( |(t( |

|(no.) |(mm) |(m) |(m) |(s) |

|1 |--- |16.75 |--- |--- |

|2 |4.9 |15.70 |1.04 |0.728 |

|3 |9.9 |15.05 |2.49 |0.727 |

|4 |15.9 |15.97 |4.47 |0.719 |

On the basis of the knowledge of (t( and a, the calculated value of leak location, [pic], can be obtained through Eq. (1). In Table II, the values of [pic] along with those of the relevant percentage error, (( , in the leak location:

[pic] (2)

are shown.

Table II. Leak location on the basis of test results

|Test |[pic] |[pic] |

|(no.) |(m) |(%) |

|1 |--- |--- |

|2 |131.00 |2.25 |

|3 |130.82 |2.11 |

|4 |129.55 |1.12 |

Basing on Table II, it can be seen that for all considered cases the error in leak location is acceptable. Moreover, for given (hm and shape of the leak, the smaller d the greater (( because of uncertainty in determining the actual time at which the reflected wave reaches the measurement section. In other words, the larger (hm, and then y(, the more evident the discontinuity in the pressure signal due to the reflected wave.

On the basis of the value of y(, the discharge behavior of the leak can be determined using inverse transient calculations (Brunone 1999).

Results of both numerical simulation and laboratory tests carried out on a long pipe confirm the reliability and validity of the proposed technique, at least when a simple pipe system is considered. The transient tests proposed do not require complex equipment to be carried out, in fact pressure measurement at only one section is needed, and do not give rise to dangerous over-pressures (about 10 – 15 m of water column). The state of the pipe can then be determined periodically by comparing the present pressure signal with the one recorded, as an example, just after the construction corresponding to an intact pipe.

REFERENCES

Brunone, B. (1989). “A technique for leak detection in wastewater outfalls. Numerical experiments”, Congress on ‘Wastewater Outfalls’, Ischia, Italy, Univ. of Naples, D.223-236 (in Italian).

Brunone, B. (1999). “A transient test-based technique for leak detection in outfall pipes”, J. Water Resour. Plng. and Mgmt., ASCE (in press).

Brunone, B., and Ferrante, M. (1999). “On leak detection in single pipes using unsteady-state tests”, In M.H. Hamza (Editor): Modeling and Simulation, IASTED ACTA PRESS, Anaheim, California, 268-272.

Brunone, B., Golia, U.M., and Greco, M. (1995). “Effects of two-dimensionality on pipe transients modeling”, J. Hydr. Engrg., ASCE, 121(12), 906-912.

Jönsson, L. (1994). “Detection of leaks using pressure transients”, Report, Univ. of Lund, Sweden.

Jönsson, L. (1995). “Leak detection in pipelines using hydraulic transients. Laboratory measurements”, Report, Univ. of Lund, Sweden.

Liggett, J. A., and Chen, L. (1994). “Inverse transient analysis in pipe networks”, J. Hydr. Engrg, ASCE, 120(8), 934 -955.

Pudar, R.S., and Liggett, J.A. (1992). “Leaks in pipe networks”, J. Hydr. Engrg, ASCE, 118(7), 1031-1046.

Wylie, E. B., and Streeter, V. L. (1993). “Fluid transients in systems”, Prentice Hall, Englewood Cliffs, N.J.

APPENDIX I – NUMERICAL SIMULATION OF TRANSIENTS IN DAMAGED PIPES

When numerically modeling unsteady-state flow in pressurized pipes (Wylie and Streeter 1993), the instantaneous value at time t of the discharge through a leak, [pic], can be visualized as occurring through a lateral orifice. Thus [pic] is assumed to be given by the traditional equation:

[pic] (A1)

in which [pic]and[pic]are the coefficient of discharge and the area of the orifice, respectively; g is the acceleration due to gravity; [pic] and [pic]are the piezometric head and the elevation at the orifice node, respectively.

To obtain the time-history of [pic] and [pic], the differential equations for transient flow can be used. When the method of characteristics is considered, within a finite-difference approximation, at the orifice node, along with at any interior computational node, the following equations can be written:

[pic] (A2)

[pic] (A3)

for the reach length (s upstream (along the C+ characteristic line) and downstream (along the C- characteristic line) the orifice node, respectively. In (A2) and (A3), the condition of a common piezometric head, [pic], at the orifice node is assumed and [pic]([pic]) is the upstream (downstream) discharge and s = axial co-ordinate, with s = 0 at the initial section of the pipe. In (A2) and (A3), coefficients are defined as: [pic]; [pic]; [pic]; [pic]; where [pic] and [pic] ([pic] and [pic] ) are the piezometric head and the discharge at the node upstream (downstream) the orifice node at time [pic], respectively; [pic]= time step; [pic] is the pipe characteristic impedance; A = cross-sectional area of the pipe; [pic] is the pipe resistance coefficient; ( = Darcy-Weisbach friction factor and D = pipe internal diameter.

Furthermore, at any time t the continuity equation must be satisfied at the orifice node:

[pic] (A4)

The boundary conditions to be satisfied both at the upstream and downstream end section of the pipe depend on the characteristics of the induced action, which gives rise to he pressure wave, along with on those of the pipe.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download