Title: Evaluation of Hydraulic Transients and Damage Detection in Water System under a Disaster Event

Size: px
Start display at page:

Download "Title: Evaluation of Hydraulic Transients and Damage Detection in Water System under a Disaster Event"

Transcription

1 Cover page Title: Evaluation of Hydraulic Transients and Damage Detection in Water System under a Disaster Event Authors: Masanobu Shinozuka (Contact person) UCI Distinguished Professor and Chair Department of Civil and Environmental Engineering University of California, Irvine E-4150 Engineering Gateway Irvine, CA Phone (949) Office Fax: (949) Fax shino@uci.edu Xuejiang Dong (co-author) Research Associate Department of Civil and Environmental Engineering University of California, Irvine E-4315 Engineering Gateway Irvine, CA Phone (949) Office Fax: (949) dongx@uci.edu 1

2 Evaluation of Hydraulic Transients and Damage Detection in Water System under Disaster Events Masanobu Shinozuka 1, and Xuejiang Dong 2 ABSTRACT This study explores the methods of rapidly detecting and locating the damage in a water delivery system taking advantage of sharply transient change in hydraulic parameters such as water head and flow rate under disaster events. In addition, we also considered the detection of equipment malfunction within the system that can cause similar and often more serious transient states. For this purpose, we used computer code HAMMER by Heasted Methods [1] in an ARC/GIS platform so that the inventory, operational, and management features can be integrated into the transient analysis all at the outset. We envision that the emerging sensor and data transmission technology will make it possible to monitor, process, and analyze the key hydraulic data in real-time, and extract the signature (such as maximum water head gradient) which can most effectively identify the severity and location of damage and malfunction. The proposed technology will serve as a next generation of the SCADA (Supervisory Control and Data Acquisition) system. Current generation of SCADA system that the utility industry deploys primarily for the purpose of system operation, not for rapid response to acute transients resulting from severe damage sustained by pipes, sudden stoppage of pump operation, and the like. We demonstrate by numerical simulation that the local water head gradient, for example, can serve as key signature, and that the source of damage/malfunction is at the joint closest to this point of maximum gradient. This research is also consistent with the national effort in enhancing the level of homeland security as described in the 2003 Academy Press publication Making the Nation More Secure [2] which identified future development of SCADA as one of the most critical agenda items for enhanced national security. INTRODUCTION Urban water delivery network systems, particularly the underground components, can be damaged due to earthquakes, severely cold weather, heavy traffic loads on the ground surface, and other causes. In all these situations, the damage cannot be detected and located easily, especially immediately after the damaging event. In recent years, real-time or near real-time damage assessment and diagnosis of buried pipelines has attracted much attention from researchers focusing on early detection of the damage severity and location. However, due to large size of network and complex nature of the physics that affect the pipe damage, particularly under seismic waves, such detection still remains difficult to achieve. As a possible solution, time histories of these and other parameters can be simulated with the aid of computer codes that are capable of transient analysis and allow creation of new nodes at the location of pipe damage. In addition, we also considered the 1 UCI Distinguished Professor and Chair of Department of Civil and Environmental Engineering, University of California, Irvine, E-4150 Engineering Gateway, Irvine, CA Research Associate, Department of Civil and Environmental Engineering, University of California, Irvine, E-4315 Engineering Gateway, Irvine, CA

3 detection of equipment malfunction within the system that can cause similar and often more serious transient states. It is envisioned that, in the very near future, the current generation of SCADA system that the utility industry deploys primarily for the purpose of system operation, not for rapid response to acute transients resulting from severe damage sustained by pipes, sudden stoppage of pump operation, and the like. We demonstrate by numerical simulation that the local water head gradient, for example, can serve as key signature, and that the source of damage/malfunction is at the joint closest to this point of maximum gradient. Hence, given adequately dense sensor network data, we can reduce the problem to that of finding numerically and in real-time the point of maximum gradient from the optimal 3D surface fitted to the data. The magnitude of the gradient is expected to relate closely to the severity of the anomaly that caused the transient and this relationship will be the subject of future study. Also, it is the matter of future study to develop, by means of analytical simulation, first principle procedure of detection and associated algorithm. All these, together with field experiment and verification, lay the foundation for establishment of an optimal strategy toward cost-effective monitoring and emergency response taking into consideration, among other things, the type, number and location of sensors and nodes to be installed. Regional water utilities (e.g. Irvine Ranch Water District) will participate in this research providing technical data so that their system can be used as test-bed. The localization technology described above can also be applied to other lifeline systems such as power and transportation networks with appropriate modifications. The essence is the integration of a dense sensor network with rapid and robust data transmission capability, with the software capable of recognition of damage unique for each lifeline system. This research is also useful in enhancing the level of national security as described in Making the Nation More Secure (2003). This publication identified future development of SCADA as one of the most critical agenda items for enhancement of national security. In the last ten years or so, many researchers attempted to develop real-time damage and assessment diagnosis techniques for buried pipelines subjected to earthquake ground motion. Some researchers focused on establishing the relationship between damage ratio (breaks per unit length of pipe) and ground motion, taking the soil condition into consideration (e.g., Nishio [3], Tanaka[4], Yamazaki [5] ). Eguchi [6] put forward a method in which nominal damage estimated through some earthquake parameters is updated gradually based on the collection of post-earthquake observation information. In addition, Takadea and Ogawa [7] discussed seismic monitoring and real-time damage assessment, and Shinozuka et al [8] developed a methodology to detect the damage location and severity with the aid of neural network methods, and applied the method to a water network that consisted of 31 nodes and 50 pipes under the assumption of equilibrium flow. However, real water networks consist of a much larger number of nodes and links in a more complex topology and the system will produce a transient state of behavior when it suffers from damage. In 2004, Shinozuka and Dong developed a GIS-based methodology in which the correlation analysis was used to determine the damaged pipes using MLGW water system [9]. The present study demonstrates a method where the transient analysis is carried out between two equilibrium states of the water flow before and after the damage, which leads us to a rational, and cost-effective damage identification procedure for water delivery system. 3

4 DAMAGE DETECTION AND LOCALIZATION OF PIPE NETWORK Hydraulic Transients A hydraulic transient represents a temporary flow, pressure, and other hydraulic conditions that control a hydraulic system, in this case, a water delivery system, between an original (first) steady state and final (second) steady state the system achieves after the effect of the disturbance that caused such a transient is absorbed into the second state. The disturbance includes such events as a valve closure or opening, a pump stopping or restarting depending on power supply, and pipe damage or break leading to water leakage. The transient can produce a significant change in water head and pipe pressure. In fact, as described in more detail in what follows, it is envisioned that the sudden change of such pressure will generate a measurable pressure wave and used for detection and localization of pipe damage due to seismic forces. If the magnitude of this transient pressure is beyond the resistant capacity of system components, their failure can induce disastrous effects on the water system [1] where it is also suggested that these effects include (1) high or low transient pressure which results in pipe burst or collapse, (2) high transient flow which can loosen the deposits and rust and thus degrade water quality, (3) high transient forces on pipe bends and other fittings, which can cause joints to move, and (4) column separation and vibration which may cause pipes to rupture, and flanged pipes and fittings (elbows and bends) to dislodge. Therefore, it is more than prudent to simulate the transient behavior of the water system under various adverse scenarios in order to assess the magnitude of their impacts. In this study, the industry-grade computer code HAMMER developed for the transient analysis of hydraulic systems by the Haesead Methods is employed to generate time histories of key hydraulic parameters (primarily water head and flow rate). The analysis is carried out for a hydraulic system as shown in Figure 1 (intact water system) which appears in HAMMER User s Guide [1]. This water system consists of two reservoirs, one pump, one valve, thirty-eight nodes and 54 pipe links. In the ensuing analysis, we consider a case in which pipe break is assumed to occur at the mid point of link 111 in which case, a node is created and labeled node J10 as shown in Figure 2. The physical parameters of nodes and pipes are listed in Tables 1 and 2 where the following comments apply. 1. Pipe or link has two ends or nodes. One is defined as from node and the other as to node, which are labeled as FNode and TNode in Table2. 2. In Figure 1, connection node represents a junction at which two or more links meet. Consumption node means that the node from which water is supplied to the user at a specified level of flow rate and water head. 3. In Figure 2, the damaged network with pipe break location + at J10 (in Table 1 in bold and italic) is shown. In this paper, the damage is modeled as an orifice. The original pipe P111 (in Figure 1) is divided into two pipes P11 and P12 (in Figure 2, P11 and P12 in bold and italic in Table 2). 4. The water head is measured in meter and the water flow rate is measured in cubic meter per second (cms). For flow rate, the positive value means the water flows in the direction from FNode (from node of pipe) to TNode (to node of pipe) while the negative value means the flow in opposite direction (from TNode to FNode). 4

5 Table 1 Nodal Parameters LABEL CATEGORY ELEVATION LABEL CATEGORY ELEVATION J1 Junction 412 J23 Junction 396 J2 Junction 395 J24 Junction 397 J3 Junction 395 J25 Junction 408 J4 Junction 386 J26 Junction 390 J5 Junction 380 J27 Consumption 395 J6 Junction 420 J28 Junction 396 J7 Junction 395 J29 Consumption 396 J8 Junction 395 J30 Junction 396 J9 Junction 395 J31 Junction 396 J10 Orifice 395 J32 Consumption 397 J11 Junction 410 J33 Junction 410 J12 Junction 420 J34 Consumption 420 J13 Consumption 390 J35 Junction 372 J14 Junction 396 J36 Junction 360 J15 Junction 397 J37 Consumption 355 J16 Junction 397 PJ1 Junction 363 J17 Junction 380 PJ2 Junction 363 J18 Junction 420 PMP1 Pump 363 J19 Junction 435 Res1 Reservoir 383 J20 Consumption 410 Res2 Reservoir 456 J21 Consumption 385 VLV1 Valve 395 J22 Junction 395 Table 2 Pipe Parameters Pipe Length Dia FNode_Head TNode_Head Flow Label FNode TNode (m) (mm) D-W (m) (m) (cms) P1 PJ2 J P2 J1 J P3 J2 J P4 J3 J P5 J4 J P6 J5 J P7 J6 Res P8 J2 J P9 J8 J P10 J9 J P11 J9 J P12 J10 J P13 J12 J P14 J13 J P15 J9 J

6 Table 2 Pipe Parameters (Cont d) P16 J15 J P17 J11 J P18 J15 J P19 J16 J P20 J17 J P21 J17 J P22 J14 J P23 J18 J P24 J19 J P25 J15 J P26 J23 J P27 J19 J P28 J20 J P29 J16 J P30 J25 J P31 J21 J P32 J22 J P33 J22 J P34 J23 J P35 J23 J P36 J23 J P37 J24 J P38 J27 J P39 J28 J P40 J28 J P41 J29 J P42 J31 J P43 J30 J P44 J31 J P45 J32 J P46 J25 J P47 J33 J P48 J35 J P49 J35 J P50 J36 J PMP1D PMP1 PJ PMP1U PJ1 PMP PS1 Res1 PJ PVD VLV1 J PVU J7 VLV

7 P10 Pipe #10 J9 Joint #9 VLV1 Valve #1 PMP1 Pump #1 Figure 1 Water Delivery System P10 Pipe #10 J9 Joint #9 VLV1 Valve #1 PMP1 Pump #1 Figure 2 Damaged Water Delivery System (P111 Damaged) 7

8 Four event scenarios are considered and modeled as follows. 1. pipe P111between Joints 9 and 11 (initially not damaged) suffers a break in the middle at time t=5 sec. We create Joint 10 at this mid point servings as an orifice through which the water leaks as shown in Figure 2. Pipes 11 and 12 are also created in place of P111 as a part of the network. No repair is made. 2. The same damage scenario as (1) above, but repair is made at t= 15 sec* (*unrealistic, analytical convenience only) 3. At t= 5 sec, the pump stops due to loss of power. The pump is not restarted. 4. The same power loss as (3) above, but at t=20 sec,* operator restarts it activating emergency power. HAMMER code can provide a wide range of system performance information under these scenarios. However, only time histories of water head and flow rate are plotted here. The water head histories at J9, J11, J13 and J20 (the last two are consumption nodes) are plotted for these four scenarios as shown in Figures 3-6. The water head time histories show that significant hydraulic transients can occur at the nodes which are close to damage location, while the nodes far away from the damaged pipe experience less prominent transient behavior. This is obviously what we expect and yet these quantitative results are the pillars of the framework for the proposed advanced SCADA system. Damage Detection A method of damage detection and localization, including the identification of malfunctioned equipment, is described here that is based on the comparison of the hydraulic parameters (the water head in this case and flow rate will also be considered in future study) before and after the event. For the primary purpose of a rapid detection and localization, it is most effective to catch the sign of change at the outset of the event. Fortunately for a sudden change such as a pipe break and pump stoppage, the response of the network is rapid particularly in the neighborhood of the source, as demonstrated in Figures 3-6. This suggests that some measurable signatures that indicate the rapidity of this change can be used for this purpose. One convenient quantity that serves this purpose is the water head gradient as defined below. H 2 H1 D = (2) t2 t1 Here H 2 and H 1 are the water head of a node at the time t 2, t 1 respectively and t 2 -t 1 =0.2 second in this study. During the steady state normal operation, D is usually very small. In this paper, the water head gradient measured at the observation nodes are integrated into the GIS platform for real-time visualization and for other advantages. In fact, the time histories of water head is generated and shown in Figure 3-6 under the scenarios. Figure 7 and 8 are the water head gradient distribution under event scenario 1 and 3 respectively. They indicate that the point of highest value of D as the location where damage occurred. 8

9 (a) J9 (b) J11 (c) J13 (d) J20 Figure 3 Nodal Water Head Time Histories under Event Scenario 1 (a) J9 (b) J11 (c) J13 (d) J20 Figure 4 Nodal Water Head Time Histories under Event Scenario 2 9

10 (a) J9 (b) J11 (c)j13 (d) J20 Figure 5 Nodal Water Head Time Histories under Event Scenario 3 (a) J9 (b) J11 (c) J13 (d) J20 Figure 6 Nodal Water Head Time Histories under Event Scenario 4 10

11 Figure 7 Distribution of Water Head Gradient due to a Pipe Break Figure 8 Distribution of Water Head Gradient due to Pump Losing Power 11

12 Development of Advanced SCADA Damage localization in utility networks was studied in the context of SCADA [5,6,7]. At this time, we plan to instrument Irvine Ranch Water Distribution s network using MEMS sensors at possible locations (e.g. surface of pipes at manhole and hydrant, [8] ) to observe sudden changes in the water pressure. This will not require invasive procedures and still identify the extent and location of the damage. The result of transient simulation analysis using a virtual network (Fig.1) demonstrated sharper changes in the pressure (Figs. 3-6) at the joints closer to the location of a pipe break. These changes observed at a large number of sensor sites through the SCADA consisting of the MEMS (Micro Electro Mechanical Systems) sensors can be used to identify the location and extent of the damage. The pressure and flow rate changes themselves are not necessarily the quantities to be monitored, but it is the sudden vibration activity that is induced by propagating waves originating from the point of pipe damage. In this context, acoustic emission techniques could prove to be useful depending of pipe material. Clearly, monitoring of other facilities in the water network systems such as central buildings and filtration plants must also be integrated in the SCADA system. We use again the Irvine Ranch Water District s network as test-bed for this purpose as we make progress. CONCLUSION AND FUTURE RESEARCH The purpose of this study was to develop a methodology to identify the location and determine the severity of damage in a water delivery system by monitoring water pressure on-line at densely installed sensor locations within the system. Water head gradient is introduced as index and used as a key parameter to locate damaged pipe when the water system exhibit acute transient behavior. Currently, only water head is considered for the stated purpose. Actually, the flow rate change also provides useful information and will be used to enhance the accuracy of the identification result for this purpose. The method, which is based on on-line water pressure variation before and after water system damage or malfunction, will be integrated with GIS-based SCADA system to provide a practical real-time damage identification method and a decision support tool for the effective disaster response. The future study will be focused on (1) optimizing the number of monitoring stations with careful selection of their locations, and (2) refining the methodology to achieve more accurate results, particularly with the effective usage of water flow information and (3) applying this method to regional water utilities (e.g. Irvine Ranch Water District). ACKNOWLEDGEMENTS This work was supported under National Science Foundation Grant CMS REFERENCES [1] HAMMER User s Guide (2003), HAESTAD Press, CT, USA, [2] National Academies (2003) Making the Nation More Secure, National Academy Press [3] Nishio, N. (1994) "Damage Ratio Prediction for Buried Pipelines Based on the Deformability of Pipelines and the Nonuniformity of Ground," J. of Pressure Vessel Technology, ASME, 116, [4] Tanaka, S., Shinozuka, M., and Hwang, H. H. M. (1993) "LIFELINE-W(2) User's Guide: a Program for Connectivity and Flow Analysis of a Water Delivery System Under Intact and Seismically Damaged Conditions." Technical Report, Princeton University 12

13 [5] Yamazaki, F., Katayama, T., and Yoshikawa, Y. (1994) "On-Line Damage Assessment of City Gas Networks Based on Dense Earthquake Monitoring." Proc. 5th U.S. Conf. on Earthq. Engrg., EERI, Vol.4, [6] Eguchi, R. T., Chrostowski, J. D., and Tillman, C. W., (1994) "Early Post-Earthquake Damage Detection for Lifeline System." EQE Research Report prepared for National Science Foundation [7] Shinozuka M., Liang J.w. and Feng, Q., M., (2005). Use of SCADA for Damage Detection of Water Delivery Systems, Journal of Engineering Mechanics, ASCE, March,. [8] Takada, S., and Ogawa, Y. (1994) "Seismic Monitoring and Real Time Damage Estimation for Lifelines." Proceedings of the 4th U.S. Conference on Lifeline Earthquake Engineering, ASCE, [9] Dong,X.J. and Shinozuka, M. (2004) GIS-Based Seismic Damage Localization for Water Supply Systems Proceedings of the 13 rd World Conference on Earthquake Engineering, Vancouver, B.C., Canada, Paper No