当前位置:首页 >> 电力/水利 >>

Experimental study of pipeline leak detection basedon


STRUCTURAL CONTROL AND HEALTH MONITORING

Struct. Control Health Monit. 2015; 22:799–812 Published online 14 November 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/stc.1718

Experimental study of pipeline leak detection based on hoop strain measurement
Zi-guang Jia1,2, Liang Ren1,*,?, Hong-nan Li1, Siu-Chun Ho2 and Gang-bing Song1,2
1

School of Civil and Hydraulic Engineering, Dalian University of Technology, Dalian, Liaoning, 116024, China 2 Department of Mechanical Engineering, University of Houston, Houston, TX 77204, USA

SUMMARY Pipelines are widely used for the transport of a large variety of ?uids, such as natural gas, across long distances. While pipelines provide a convenient mode of transportation of ?uids, their safe usage is one of the foremost concerns especially if their contents are harmful to the environment or if the hosting area is prone to third-party intrusions. Thus, rapid detection and localization of pipeline leakage is paramount to the minimization of damage brought to the environment and stakeholders in the event of an unexpected leakage. In this work, a novel hoop strain based negative pressure wave (NPW) approach was used to detect and localize pipeline leakages in a 180 ft PVC pipeline equipped with ?ve manually controllable leakage points. Using the new approach, both the arrival time of the NPW and the energy attenuation pro?le of the NPW can be used to detect and localize leakages with higher accuracy and in a wider variety of situations. The time of arrival approach allowed accurate (within 7.33% error) and repeatable localization of the leakage points; however, using the energy attenuation of the NPW, leakages with low leakage rates (<5 lpm) can also be detected, albeit at the cost of inaccuracies at the inlet and outlet of the pipeline (8.3% error in the ends vs. 4.3% error in the body). Copyright ? 2014 John Wiley & Sons, Ltd. Received 31 March 2014; Revised 1 July 2014; Accepted 8 October 2014
KEY WORDS: leak localization; negative pressure wave (NPW); energy attenuation; hoop strain; ?ber Bragg grating (FBG); pipeline monitoring

1. INTRODUCTION Pipelines are used as one of the most practical and economically effective modes of transport for large volumes of ?ammable and potentially dangerous substances, such as natural gas, for which automotive transportation is often impractical [1]. A pipeline leak can cause both a costly loss of product as well as serious environmental damage if the leak is not stopped in time [2]. Therefore, it is important for companies and the government hosting the pipeline to be able to quickly detect leakages to mitigate problems and avoid potentially crippling costs. For a leak detection method to be effective, the monitoring system must accurately and precisely detect the location of potential leakage points. For long pipelines, accuracy of the system plays an important role in the response time of repair crews. For example, a location accuracy of 3% compared with 2% can make a difference of a half a mile for a 50-mile pipeline, which can signi?cantly stall efforts involved in visually locating and repairing the leak. The negative pressure wave (NPW) technique has risen as among the most effective, prevalent leak detection methods and has already been widely applied in practical pipeline monitoring in combination with other signal processing techniques and mathematic analyses [3–6]. The basic principle of NPW
*Correspondence to: Liang Ren, School of Civil and Hydraulic Engineering, Dalian University of Technology, Dalian, Liaoning, 116024, China. ? E-mail: renliang@dlut.edu.cn

Copyright ? 2014 John Wiley & Sons, Ltd.

800

Z.-G. JIA ET AL.

states that when a leak occurs, a rarefaction wave propagates from the leakage point toward both ends of the pipeline. The NPW is relatively constant and can be detected with pressure sensors because of the pressure drop due to the leakage. With two sensors, one placed upstream and the other downstream to the leak, the leakage point can be pinpointed within certain accuracy. A drawback, however, is that the energy of NPW attenuates while propagating along the pipeline, thus placing a spacing requirement on sensors [7]. For the traditional NPW technique, the pressure sensors are mounted only at the inlet and outlet for a speci?c section of the pipeline. Therefore, the necessary condition for a leakage to be detectable is that the detected pressure change must be large enough to register above the noise and precision levels of the pressure sensor. In some cases, the pressure drop induced by the pipeline leak cannot be detected when the energy of NPW attenuates too much as it travels through long distances of pipeline, which is an obvious limitation of this traditional method with using pressure sensors. To address the limitations of the traditional NPW method as mentioned earlier, this paper describes a new approach based on pipeline hoop strain variation measurement to detect and locate pipeline leakages. The reduction in hoop strain due to pressure drop is monitored by strain gages mounted on the surface of the pipeline. Changes in the gauge readings can indicate leakage, and by processing the strain difference among two or more hoop strain sensors, the site of the leakage can be calculated. On the other hand, the ?ber Bragg grating (FBG) [8,9], with its advantages of immunity to electromagnetic interference, high accuracy and multiplexibility is chosen as the ideal sensor for measuring strain variations on pipelines. Electrical strain gauges, on the other hand, will require much power and cabling as pipeline distance increases. Therefore, as described in this paper, an investigation using FBGs to detect pipeline leakage based on hoop strain measurement was conducted.

2. LEAK DETECTION THEORY BASED ON HOOP STRAIN MEASUREMENT 2.1. Traditional NPW technique In certain situations, such as third-party intrusion, the pipeline leakage occurs as a sudden event. As the contents suddenly escape from inside the pipe, the instantaneous pressure drop due to the loss of ?uids causes a decompression wave to propagate toward both ends of the pipeline [10]. This decompression wave is referred to as the NPW, which travels through the contents of the pipe. In order to detect the wave, the initial pressure of the pipe prior to leakage must ?rst be measured as a standard for comparison. In the traditional method, the pressure drop due to NPW is measured by two pressure sensors installed in the upstream and downstream sections of the pipeline, usually near the pump stations. The difference in time it takes for the NPW to reach the sensors, along with the estimated wave propagation velocity, is used to project the location of the leakage point. Figure 1 illustrates the working principle of this method. Referring to Figure 1, the distance of the leak point to the upstream pressure sensor is L 1 ? ?μ ? V ? ? μ ? V ? Δt ? L 2μ (1)

where L1 denotes the distance between the upstream end and the leak point, μ denotes the propagation velocity of the NPW, V denotes the ?ow velocity and Δt denotes the arrival time difference of the wave

Figure 1. Illustration of pipeline leakage and NPW propagation.
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

PIPELINE LEAK DETECTION BASED ON HOOP STRAIN MEASUREMENT

801

to the upstream and downstream sensors. Usually, the NPW velocity μ is much higher than the ?ow velocity V, and thus if V is ignored, the formula can be simpli?ed to become L ? μΔ t (2) 2 The effectiveness of the traditional method is dependent on several parameters, one of which is the actual distance between the sensors and the leakage. The longer the pipeline, the longer the NPW must travel before reaching both sensors and the higher chance that the energy of the NPW will attenuate below the sensor threshold levels, thus becoming undetectable. Until now, no solution for leakage detection has been proposed in the literature for cases in which the NPW attenuates below the sensor threshold. The closest related research is the calculation of the smallest detectable leakage ?ow rate for the traditional method [7]. L1 ? 2.2. Energy attenuation of NPW As the NPW propagates along the pipeline, the energy of the wave attenuates, and with suf?cient pipeline length, the propagation energy will fall below the detection abilities of traditional pressure sensor based techniques. The attenuation of energy can be modeled on the basis of the various parameters of the pipeline as well as the pressure drop due to the leak. The pressure drop at the position which is l away from the leak position while taking into consideration the energy attenuation can be expressed as [7]   fQ l (3) ΔPl ? ΔP0 1 ? μA ρD where ΔP0 is the pressure drop at the leakage point, f denotes the friction factor, Q denotes the mass ?ow rate, ρ denotes the ?uid density and D denotes pipeline diameter. From this equation, it can be noted that the maximum pressure drop occurs at the leakage point and that the attenuation of the NPW is proportional to the distance between the leak point and the reference point. These variables work together to in?uence the rate of attenuation. Consider the attenuation coef?cient of the NPW, which relates the preceding variables to pressure drop per distance along the pipe and can be de?ned as follows: C NPW ? fQ μA ρD (4)

where A denotes the section area of the pipeline. From the preceding equation, it can be seen that for a pipeline section with speci?c diameter, ?uid density, friction factor and velocity of NPW, a larger ?ow rate and thus a larger pressure difference, ΔPf, will cause more rapid NPW energy attenuation. However, it should be noted that the calculated value of CNPW may not be exact as some of the parameters are dif?cult to determine and the pressure difference ΔPf is variable. Under the right conditions, the pressure drop due to the NPW will be decreased to the point where the pressure sensors will not be able to detect any change and thus cannot provide any warnings. However, no solution was presented for extremely low rates because of the limitation of the traditional NPW technique. One way to rectify this problem is to minimize the length that the NPW must travel before reaching the sensors. As will be discussed in the next section, this can be achieved by installing additional sensors along the pipeline and by measuring hoop strain instead of pressure. In our new energy attenuation approach, multiple FBGs will be bonded along the pipeline so that the energy attenuation pro?le can be measured, with the resolution being affected by the number of sensors used. 2.3. Hoop strain theory As a hollow cylindrical object such as a pipe is pressurized internally whether due to a liquid or gas, a force begins to act outwardly upon the inner surface area of the pipe (Figure 2). One of the key parameters that change because of the introduction of the pressure is the circumferential or hoop stress of the pipe wall. The hoop stress is normal to both the axial and radial stress vectors.
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

802

Z.-G. JIA ET AL.

Figure 2. Pipe cross-section diagram under internal pressure.

If a pressurized pipeline experiences changes in internal pressure, for example, because of leakage, the pipeline will strain along the circumferential direction. The circumferential strain or the hoop strain of the pipe can be obtained by [11] ε? σ pR ? Ep Ep δ (5)

where σ is the hoop stress, δ is the pipe thickness, R is the inner radius and EP denotes the Young’s modulus of the pipe. For the equation to hold, δ must be much less than R. The axial stress is ignored because the pipe can be regarded to be in?nitely long. During a pipeline leak, the hoop strain will change proportionally to the sudden pressure drop according to Equation (5). If the hoop strain can be measured in real time, the leak can be detected in the form of unexpected hoop strain variation. The principle of hoop strain variation can be used to monitor and localize pipeline leakage points. 2.4. Leakage detection using hoop strain variations From the aforementioned introduction to NPW techniques, we now propose a hoop strain measurement based pipeline leak detection method that compensates for the limited sensitivity of traditional pressure sensor based techniques. The principle of the new method is illustrated in Figure 3. As shown in Figure 3, a series of hoop strain sensors are installed along the length of the pipeline to provide real-time monitoring of the hoop strain variation. During normal operation conditions, the hoop strain along the pipeline ?uctuates within an expected range as the pipeline is pressurized with a relatively stable pressure distribution. A drop in the hoop strain measurement would indicate a potential leakage problem due to the loss of hoop strain that is concurrent with pipeline depressurization. On the basis of the variations of the different hoop strain sensors, the leakage point can be localized using the following steps.

Figure 3. Principle of leak detection based on hoop strain measurement.
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

PIPELINE LEAK DETECTION BASED ON HOOP STRAIN MEASUREMENT

803

Firstly, referring to Figure 3, the hoop strain sensors S1 at the upstream and Sn at the downstream are utilized as the pressure sensors to determine the arrival time difference of the NPW. The basic algorithm to localize the leak position is the same as the traditional NPW technique based on pressure sensors, and thus the leak position can be calculated using Equation (2). However, this method requires that the hoop strain decrease due to the leak at position x exceed the sensor detection threshold even after the attenuation from long-distance propagation. On the other hand, an advantage of this approach over the traditional pressure sensor based NPW technique is that multiple hoop strain sensors can be used to determine the NPW arrival time. Note that theoretically the arrival time difference between any two hoop strain sensors can be entered into Equation (2) to compute the leakage position. However, the accuracy cannot be guaranteed if the two sensors are close (e.g. Sm and Sm+1 in Figure 3) because of the potential sampling rate limitations of the data acquisition equipment used for the hoop strain sensors. Secondly, as discussed earlier, the pressure drop induced by NPW may be undetectable by all sensors after severe attenuation from long-distance propagation, and thus S1 and Sn cannot be solely used to determine the arrival time of NPW. Consider a case in which the hoop strain decrease can only be detected by the sensors from Sa to Sb, and all other sensors are unable to detect the leakage due to the limitations of the data acquisition equipment. In this situation, the procedure to locate the leak point by using NPW energy attenuation is as follows: 1. From Equations (3)-(5), the hoop strain variation during leakage can be expressed as Equation (6), where Li denotes the distance between the hoop strain sensor Si and the original point. εi ? ΔPi R ΔP0 ?1 ? C NPW jLi ? xj?R ? Ep δ Ep δ (6)

2 Two adjacent maximum values εm and εm+1 in the detectable hoop strain variation series εa to εb should be found so that the following requirement is met  εm ≥ εi ; i ∈ ?a; m? εm?1 ≥ εi ; i ∈ ?m ? 1; b? (7)

3 Therefore, the hoop strain series are divided into two groups by εm and εm+1, and thus two straight lines LU and LD can be ?tted. The ?tted lines can be expressed by Equation (8), where εmax is an unknown quantity, and the hoop strain variation at the leak point and α1, α2, Δ1 and Δ2 are all parameters obtained by ?tting LU and LD  LU : εi ? α1 Li ? ?εmax ? α1 x? ? α1 Li ? Δ1 ; i ∈ ?a; m? LD : εi ? ?α2 Li ? ?εmax ? α2 x? ? ?α2 Li ? Δ2 ; i ∈ ?m ? 1; b? (8)

4 Finally, the intersection point of the ?tted straight lines can be used to obtain the leak position x and the pressure drop at the leak point ΔP0 (Equations (9) and (10)). Δ2 ? Δ1 Δ2 α1 ? Δ1 α2 ?Li ; εi ?intersection ? ; α1 ? α2 α1 ? α2 " # Δ2 ? Δ1 Ep δ ?x; ΔP0 ? ? ; 1 ?Δ1 α2 α1 ? α2 R Δ 2 α α1 ?α2   (9)

(10)

From the Equation (10), it can be seen that the located leak position x is only dependent on parameters obtained from the process of ?tting the data instead of physical parameters of the pipeline (e.g. pipeline diameter, material property and NPW energy attenuation coef?cient). This is an advantage over traditional NPW technique in which the NPW velocity must be known. On the other hand, in some situations, the attenuation of the NPW may be low enough as to cause an indiscernible
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

804

Z.-G. JIA ET AL.

hoop strain variation measurement among the sensors. In such situations, it would be recommended to use the traditional method instead. The aforementioned hoop strain based leakage detection method requires the use of multiple strain sensors along a pipeline. While conventional electrical strain gauges can be used to gather the necessary data, ?ber optic strain gauges are recommended to be used instead because of their ability to be multiplexed into one line. Fiber optic sensors would help avoid the scenario in which a cumbersome amount of electrical power and data transmission wires are used to operate multiple electrical strain gauges along a lengthy pipeline. In the following experimental veri?cation of the hoop strain based method, FBG sensors were used to measure the hoop strains of a model pipeline.

3. EXPERIMENTAL SETUP OF MODEL PIPELINE LEAK DETECTION 3.1. Preliminary tests for investigating the in?uence of pipeline bends Prior to applying the leak detection method to a full model pipeline, certain fundamental aspects that can affect the outcome of the ?nal results must be investigated. One of the important aspects is the effect of the bending connectors. The energy of the NPW attenuates according to the concepts of friction loss in pipes, and head loss can be calculated through the Darcy equation. According to the Working Guide to Pump and Pumping Stations calculations and simulations [12], the minor losses associated with ?ttings, such as elbow connectors, can be addressed by the concept of the equivalent length method. The frictional loss hf-?tting in a ?tting is hf ? ?tting = K(V2/2g), where K is a dimensionless resistance coef?cient and V is the velocity head. This equation is also sometimes referred to as the head loss coef?cient of energy loss coef?cient. For example, the resistance coef?cient K for 0.75 in. standard 90° and 45° elbow is 0.75 and 0.4, respectively, according to the aforementioned working guide. Thus, the in?uence of each elbow connector to NPW energy attenuation is considered to be independent of each other and can be linearly accumulated along the length of the pipeline. Combined with the basic relationship between the hoop strain and the inner pressure, the effect of the number of bends on hoop strain is considered to be linear and accumulative. However, a more accurate and effective way to investigate the energy attenuation of NPW due to ?ttings is through experimental study. Thus, a set of preliminary experiments were carried out, and their purpose was twofold. First, the velocity and energy attenuation coef?cient of a NPW on a straight pipeline section needs to be calibrated. Second, the effects of pipeline elbow connectors (e.g. 45° and 90°) on the NPW also requires investigation, as the full model pipeline as well as real in-?eld pipelines will have similar connectors to allow changing in the pipeline direction. Data from the tests will be used to compensate for the bending of the full model pipeline. As shown in Figure 4, three PVC pipeline sections (0.75 in. diameter, 0.125 in. wall thickness) were constructed for the preliminary test. Commercially available PVC cement was used to connect pipe sections. For each pipeline, two bare FBGs (stripped to the core) were installed at the upstream and downstream ends (SA and SB). Using FBG sensors, the drop in hoop strain as measured instead of pressure drop. Two manually controllable leakage valves were also installed at the two ends (LA and LB). The Micron Optics sm130 was used to interrogate the FBG sensors. The calibration test was performed to provide information about the base velocity and energy attenuation of NPW in air-pressurized PVC pipes with the aforementioned dimensions. Pipeline 1 (single straight section) was selected for the calibration test. The inlet and outlet were kept closed during the test as the ?ow rate had negligible effect on NPW velocity. The NPW velocity μ was obtained by manually opening a valve (to simulate a leak) after pressurization and using the traditional time-difference method to analyze the resulting NPW. This test was repeated 20 times each for pipelines 1–3 under different internal pressures and ?ow rates. The resulting NPW velocity from the tests was used in the subsequent full model test. The second part of the preliminary test was to examine the effects of pipe bending on the NPW energy attenuation coef?cient. As described by Equation (4), the ?ow rate of the air within the pipeline will in?uence the coef?cient of energy attenuation, CNPW. Thus, a steady, controlled ?ow rate caused by a 1 psi pressure differential between the inlet and outlet was kept during the energy attenuation test. Higher pressure differentials led to unsteady ?ow rates after leakage was initiated with the present
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

PIPELINE LEAK DETECTION BASED ON HOOP STRAIN MEASUREMENT

805

Figure 4. Three types of model pipelines with same length were used for the preliminary tests.

pipeline con?guration. On the basis of the hoop strain measurements made by the FBGs, SA and SB during the leakage can be used to calculate CNPW through the following substitution of Equations (4) and (5) into Equation (3):   fQ ΔPA R Δ P 1 ? l 0 A μA ρD εA 1 ? C NPW lA E δ  ? ? PpB R ? (11) fQ εB ΔE 1 ? C NPW lB ΔP0 1 ? lB δ
p

μA ρD

In which CNPW can be expressed as follows
εA ?1 C NPW ? εAεB εB lB ? lA

(12)

where εA, εB, lA and lB denote the hoop strain variation caused by the leak and the distance between the leak position and the sensors SA and SB, respectively. The same measurements and analyses were made on pipelines 2 and 3 in order to examine how bending in the pipeline in?uences NPW energy attenuation. As will be seen in the following section, the energy attenuation results from pipeline 3 (90° elbows) will be used to compensate for the bending in the full model pipeline. The second part of the preliminary test was repeated 20 times for each pipeline. 3.2. Full model pipeline test with multiple leakage points The full model pipeline consisted of a series of PVC pipe sections connected together to form a pipeline with a total length of 180 ft. Because of the long length, the pipeline was bent at 30 ft straight sections using ten 90° elbow connectors and ?ve 1 ft pipe sections. Sixteen bare FBG sensors were installed at 10 ft intervals from each other. Five manually controllable leakage valves were installed every 30 ft along the pipeline starting at 15 ft of the ?rst FBG sensor (Figure 5 and Table I)
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

806

Z.-G. JIA ET AL.

Figure 5. Diagram of model pipeline with multiple simulative leak points. Table I. Dimensions of the preliminary pipelines. Pipeline 1 Total length (ft) Number of connectors Section length 30 0 30 Pipeline 2 (45°) 30 20 1 Pipeline 3 (90°) 30 20 1

Four tests were conducted on the full model pipeline to verify the traditional time-difference based method in localizing leaks. Using a pressure regulator and ?ow meter, the internal pressure and leakage rate were controlled to set levels during the tests (Table II). The calibrated NPW velocity and the NPW arrival time at the FBG hoop strain sensors S1 and S16 located at the opposite ends of the pipeline (Figure 5) were used to estimate (Equation (2)) the location of the leakage point.

Table II. Parameters for NPW arrival time-difference test for full model pipeline. Internal pressure (psi) Test Test Test Test 1 2 3 4 40 40 20 20 Leakage rate (lpm) 100 40 100 40
Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

Copyright ? 2014 John Wiley & Sons, Ltd.

PIPELINE LEAK DETECTION BASED ON HOOP STRAIN MEASUREMENT

807

Using the energy attenuation method used previously in the preliminary tests, pipeline leakage localization was performed on the full model pipeline. A pressure differential of 1 psi was also applied between the inlet and outlet of the pipeline prior to starting any leaks. Through the hoop strain measurements of the FBGs and the linear ?tting method as described by Equations (6)-(10), the leakage position was localized. In order to understand the characteristics of the NPW for a straight pipeline of the same length (180 ft), the results from pipeline 3 was used to compensate for the 90° bends of the full model pipeline. Essentially, the data from pipeline 3 will allow the unraveling of the bended full model pipeline into a hypothetical straight pipeline of the same total length.

4. RESULT DISCUSSION 4.1. Preliminary calibration for NPW velocity A typical hoop strain reading from the FBG sensors is shown in Figure 6. As shown in the ?gure, the hoop strain as measured by the FBG sensors (SA and SB) reduced upon the arrival of the NPW to their location. In particular, a LabVIEW based algorithm was used to obtain the arrival time of the NPW to the sensor locations. In the algorithm, if the hoop strain signal exceeds a threshold of about 5 με for over 1 ms (10 samples at 1000 Hz), then a leak is likely to have occurred. On the basis of the difference in time in which the signal crossed the threshold for each sensor, the arrival time of the NPW to the sensors (SA and SB) can be determined. The FBG closer to the leakage point (both LA and LB were used) experienced the hoop strain sooner than the other FBG. On the basis of the arrival time difference of the NPW and the distance of the sensors from the leakage point, the NPW velocity μ can be estimated through Equation (2). This calibration test was repeated 20 times (10 times each for LA and LB) to obtain more reliable results. Figure 7 presents the estimated NPW velocity obtained from the preliminary tests (20 tests for each pipeline). As can be seen in Figure 7, the estimated velocity was similar across the three pipelines, and thus no correlation can be observed between NPW velocity and the amount of bending in the pipeline. The average NPW velocity estimated from the tests was 285.75 ± 23.6 m/s (937.5 ± 77.5 ft/s, with a coef?cient of variation of 8.3%), which is close to the theoretical 300 m/s NPW velocity reported in the literature [13]. 4.2. Leak localization based on NPW arrival time difference Upon comparing the experimental localization results (see the four conditions of Section 3.3) and the actual distances of the leakage positions, an average error of 7.33% was found (Figure 8). However, it should be noted that the error is a function of the ratio between the length of the pipeline and the sampling frequency of the interrogator. At 1000 Hz sampling frequency, the resolution was about

Figure 6. Typical hoop strain variations induced by leak.
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

808

Z.-G. JIA ET AL.

Figure 7. Calibrated NPW velocity for all three model pipelines.

Figure 8. Leak position determination results of model pipeline with multi leak valves.

1 ft/s assuming the velocity of the NPW is about 1000 ft/s. If applied to the much longer in-?eld pipelines that span for many miles, the error would become much smaller as to be negligible. 4.3. Test of CNPW and in?uence of bending connectors The effects of NPW energy attenuation on the hoops strain at the locations of the FBG sensors in the preliminary tests are shown in Figure 9. A difference in the returning baseline level can be observed for the two sensor locations. Because the sensing point SA was nearer to leakage point LA, the energy of NPW as SA attenuated less than in SB. Therefore, the difference of hoop strain variation can be applied to determine both the value of CNPW (for pipeline 1) and the NPW energy attenuation due to bending (for pipelines 2 and 3). A lower sampling rate (10 vs. 1000 Hz) was used instead to stabilize the signal, contributing to the difference between Figures 9 and 6. Because of the small leakage rate, the NPW arrival time difference cannot be distinguished from the data. As can be seen from Figure 9(a), εA = 5.02 με and εB = 3.24 με (Δε = 1.78 με), corresponding to CNPW = 0.0112/ft through Equation (12). The value of the CNPW changed when the leakage rate or the internal pressure of the pipeline was adjusted; however, for a longer pipeline, additional FBGs may be used, which allows the use of the hoop strain ?tting method described in Section 2.2. As the method does not rely on CNPW, unexpected changes in the leakage rate or internal pressure would not affect the accuracy of localization.
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

PIPELINE LEAK DETECTION BASED ON HOOP STRAIN MEASUREMENT

809

Figure 9. Example NPW energy attenuation test results from the three model pipelines.

The primary difference among the three preliminary test pipelines was the number and degree of bends in the pipeline, while the total length remained the same. Thus, the differences in hoop strain difference as seen in Figure 9(a)–(c) can be attributed to the presence of the connectors. Using pipeline 1 (Figure 9(a)) as the baseline, the NPW energy attenuation due to bending can be determined by subtracting the strain difference seen in pipelines 2 and 3 from pipeline 1 and dividing by the number of connections. Therefore, δNPW-45° = (Δ2 ? Δ1)/20 bends = 0.0215 με/bend and δNPW-90° = (Δ3 ? Δ1)/20 bends = 0.039 με/bend, where δNPW-45° and δNPW-90° represent the energy attenuation
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

810

Z.-G. JIA ET AL.

factor of the 45° and 90° connectors, respectively. While only several conditions (internal pressure and leakage rates) were examined, the results remained stable and were used in the localization of leakage positions for the full model pipeline (δNPW-90° for the 90° bends of the full model pipeline). The strain difference between SA and SB for each of the 20 tests across the three pipelines are shown in Figure 10, and as can be seen in Table III, the average strain difference is greatest in pipeline 3 and lowest in pipeline 1. Taking into consideration the standard deviation and coef?cient of variation, a mild amount of overlap in the strain difference is present across the three pipelines. The overlap can be attributed to the dif?culty in exactly regulating the inlet to outlet pressure difference, which affects the NPW energy attenuation. 4.4. Leak localization based on NPW energy attenuation Figure 11 shows a typical hoop strain variation across the full model pipeline at the locations of the FBGs due to a leakage at L2. The line ?tting process involves ?rst ?nding the location of the highest hoop strain change, eliminating readings below 0.5 με (strain measurements below that absolute value are unstable), and compensating for the effects of the 90 connectors (adding δNPW-90°). The slope of the lines LU and LD are based on the gradual changes of hoop strain across the pipeline, and the intersection of the lines indicate the estimated leakage position. It should be noted that strain measurements made below the absolute value of 0.5 με were unstable, potentially because of secondary effects caused by the leakage. Measurements made below that value were required to be cut off, thus suggesting that the detection system is currently limited to localizing leakages that cause a strain of more than 0.5 με. The ?tting process was performed for leakages from L1 to L5, and the results are listed in Table IV. For L2–L4, the estimated location of the leakage point showed high accuracy (4.3 to 5.3% average error) with high rates of success (17 to 18 out of 20 times). However, for L1 and L5, the rates of success were much lower (11 to 12 out of 20 times), and the severity of error (7.2 to 8.6% average error) was relatively larger, potentially because of the proximity of the leak and the sensors to the inlet and outlet, thus involving additional boundary conditions that affected energy attenuation behavior of the NPW.

Figure 10. Strain differences due to NPW energy attenuation in pipelines 1–3.

Table III. Average strain difference due to NPW energy attenuation in pipelines 1–3 along with additional statistical data. Pipeline 1 (straight) Average (με) Standard deviation (με) Coef?cient of variation (%)
Copyright ? 2014 John Wiley & Sons, Ltd.

Pipeline 2 (45°) 2.25 0.18 8.2

Pipeline 3 (90°) 2.56 0.22 8.5

1.74 0.18 10.5

Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

PIPELINE LEAK DETECTION BASED ON HOOP STRAIN MEASUREMENT

811

Figure 11. L2 leak localization based on NPW energy attenuation.

Table IV. Results of leakage localization based on NPW energy attenuation for full model pipeline. Leakage location L1 L2 L3 L4 L5 Times repeated 20 20 20 20 20 Times succeeded 12 17 18 18 11 Average error (when succeeded) 8.6% 5.3% 4.3% 4.7% 7.2%

It should be noted that the general reason for the failure of L1 and L5 is the stability problem associated with the experimental setup for simulating leakages. In this experiment, the model pipeline is simply supported and the structure is ?exible. The leakage process is simulated by opening the valves manually which will cause minor vibrations to occasionally propagate from the leakage valve toward the sensing points. In this situation, the hoop strain variation induced by the vibration is mixed with the original leakage signal. On the other hand, because of the low pressures (<40 psi), the in?uence of the vibration was larger than if seen in real, in-?eld pipelines because of higher working pressures and a more rigid structure supporting condition. On the other hand, a possible future work for this hoop strain based leakage investigation is to accomplish leakage signal identi?cation from complex disturbances including structural vibration and other environmental disturbances. Generally, pipelines used in the ?eld are much longer, thus considerably reducing the effects of boundary conditions at the inlet and outlet, and should raise the success of the detection system to the levels seen for L2–L4. It is worth mentioning that in this scenario, the hoop strain variation was quite small at the inlet and outlet because of the low leakage rate (less than 5 lpm), which would have prevented the use of the traditional time-differenced based method to localize leakage. On the other hand, using information from additional FBG sensors and using the ?tting method, the leakage was localized despite the inhibiting similarity (arrival times cannot be distinguished) between the arrival times.

5. CONCLUSION A hoop strain based pipeline leakage detection and localization method was presented in this paper. The method consists of the time of arrival approach and the energy attenuation approach. Both methods were successfully demonstrated using a 180 ft PVC pipeline with ?ve manually controllable leakage points. As ?ber optic sensors are more suitable for usage in lengthy structures, FBG sensors were used to measure hoop strain at regular points along the pipeline. During the tests, it was found
Copyright ? 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc

812

Z.-G. JIA ET AL.

that the time of arrival approach allowed accurate and repeatable localization of the leakage points; however, using the energy attenuation of the NPW, leakages with low ?ow rates can also be detected, albeit at the cost of inaccuracies at the inlet and outlet of the pipeline. The work detailed in this paper opens the door to many potential avenues of research for leakage detection in pipelines. A possible next step for this research is the transition from the leakage detection and localization of single pipelines to the more general case of branched network of pipelines. Additional examples of future work are the testing of different pipeline materials and even the use of water as the internal contents of the pipelines.
ACKNOWLEDGEMENTS

This work was supported by the Science Fund for Creative Research Groups of the National Natural Science Foundation of China (No. 51121005), the Natural Science Foundation of China (No. 51108059), the Special Fund for Basic Research on Scienti?c Instruments of the National Natural Science Foundation of China (No. 51327003), the Special Project of China Earthquake Administration (No. 2015419014) and the China Scholarship Council (No. 201206060081). These grants are greatly appreciated.
REFERENCES 1. Papadakis GA. Assessment of requirements on safety management systems in EU regulations for the control of major hazard pipelines. Journal of Hazardous Materials 2000; 78(1):63–89. 2. Sandberg C et al.. The application of a continuous leak detection system to pipelines and associated equipment. Industry Applications, IEEE Transactions on 1989; 25(5):906–909. 3. Ellul IR. Advances in pipeline leak detection techniques. Pipes and Pipelines International. Journal of Hazardous Materials 1989; 34(3):7–12. 4. Hu J, Zhang L, Liang W. Detection of small leakage from long transportation pipeline with complex noise. Journal of Loss Prevention in the Process Industries 2011; 24(4):449–457. 5. Yang Z, Xiong Z, Shao M. A new method of leak location for the natural gas pipeline based on wavelet analysis. Energy 2010; 35(9):3814–3820. 6. Lei N, Juncheng J, Yong P. Leak location of pipelines based on transient model and PSO-SVM. Journal of Loss Prevention in the Process Industries 2013; 26(6):1085–1093. 7. Ge C, Wang G, Ye H. Analysis of the smallest detectable leakage ?ow rate of negative pressure wave-based leak detection systems for liquid pipelines. Computers & Chemical Engineering 2008; 32(8):1669–1680. 8. Li H, Li D, Song G. Recent applications of ?ber optic sensors to health monitoring in civil engineering. Engineering Structures 2004; 26(11):1647–1657. 9. Majumder M et al. Fibre Bragg gratings in structural health monitoring—present status and applications. Sensors and Actuators A: Physical 2008; 147(1):150–164. 10. Xiaohua Z. Judgment of leakage position in oil transfer line by negative-pressure wave principle. Chemical Engineering Design Communications 2006; 32(4):38–39. 11. Liang R et al. Design and experimental study on FBG hoop-strain sensor in pipeline monitoring. Optical Fiber Technology 2014; 20:15–23. 12. Menon E S, Menon P S. Chapter 4—Pressure loss through piping systems. In Working Guide to Pumps and Pumping Stations, Menon E S, Menon P S (eds). Gulf Professional Publishing: Boston, 2010; 69–111 13. Benkherouf A, Allidina YA. Leak detection and location in gas pipelines. IEE Proceedings-D 1988; 135:142–148.

Copyright ? 2014 John Wiley & Sons, Ltd.

Struct. Control Health Monit. 2015; 22:799–812 DOI: 10.1002/stc


赞助商链接
相关文章:
更多相关标签: