Smart Urban Water Supply Systems (Smart UWSS)

Urban water supply systems (UWSS) are the lifeline of 3 billion people globally; however, these vital systems are aging and fraught with deficiencies and inefficiencies. In Hong Kong (HK), UWSS, comprised of thousands of devices and about 7,800km of conduits. They leak water at an annual cost of more than HK$1 billion. Pipe failures can paralyze businesses and cause devastating urban floods. Worldwide, UWSS are challenged by urban growth and climate change. And yet current methods to diagnose leakage and defects in complex underground UWSS networks are deficient.

Water infrastructure has been highlighted as a critical issue nationally—in the 2011 “No. 1 document” issued by the Chinese Central Government; the 2011 “Green Quality Living in Greater Pearl River Delta” study, and the 2008 “Total Water Management” and 2015 “Water Intelligent Network (WIN)” policy of HK government. Many other countries have critical needs for massive UWSS upgrades (e.g., the American Water Works Association estimates that US$250 billion is needed in the USA alone). Indeed, HK has committed HK$20 billion to the rehabilitation and replacement of its water supply infrastructure. However, with the effort of HK$20 billion, it is anticipated that leakage can only be reduced from 25% to 15% at best.

The design and management of UWSS is currently limited by the range and resolution of data collection in the relatively inaccessible buried pipelines. Current methods fail to provide the diagnostic resolution needed for many practical problems. We propose a comprehensive theme-based research program involving theoretical, laboratory and field studies to develop a new diagnostic paradigm for water supply network monitoring and fault detection. The findings will enable timely detection of UWSS system defects and proactive mitigation measures. We have assembled an internationally-recognized and cross-disciplinary research team to create the next generation of UWSS—a truly Smart UWSS.

We propose to study the sensing of actively generated waves that travel at high speed (km/s) in the fluid in the pipe and to electronically capture wave echoes. The resulting data will be processed with advanced transient-based inverse methods and algorithms to pinpoint and characterize leaks, blockages and weak pipes. The theories will be evaluated in a field test bed in HK; a general pilot-scale demonstration experimental test bed will be developed for testing of hydraulic transient behavior for UWSS.

This research will be conducted in close collaboration with the HK Water Supplies Department (WSD) and will support HK’s WIN vision. The findings will crucially contribute to the sustainable development of HK through water conservation via locally developed innovation and technology.The project has successfully secured a funding of HK$33.225M from the HK Research Grant Council (RGC)'s Theme-based Research Scheme. Another HK$7M is obtained from the matching fund in the Hong Kong University of Science and Technology (HKUST).

Examples of Pipe Defects: Leakage and Blockage

Examples of Pipe Leakage

Examples of Pipe Blockages

Background

This section is organized as follows. PART A: project’s background is given and the existing defect detection methods and their limitations are summarized. PART B: the proposed Smart UWSS is introduced followed by the 4 research tasks that will be required to accomplish it.

PART A: Existing defect detection methods and their limitations

UWSS are large and complex. In HK, there are ~7,800 km of pressurized pipes, and thousands of pumps, valves and other devices. these systems are generally buried beneath streets, they are relatively inaccessible. Significant portions are decades old and are plagued with both obvious and hidden faults. In 2010, the South China Morning Post reported “A spate of water main bursts…left some wondering if Hong Kong is sitting on a ticking time bomb of ageing water pipes. The most recent burst…left thousands … without fresh water for 12 hours.” The HK WSD reported finding 11,028 leaks and 257 bursts for the year 2013.

Aging water infrastructure is a worldwide concern. An estimated US$334 billion over 30 years is required to address aging water infrastructure in the USA and $1 trillion if expansion is included. Recognizing the importance and urgency, the National Science Foundation provided US$18.5 million as initial funding with more to come pending review to a research project “Re-inventing America's urban water infrastructure (ReNUWIt)” (http://www.renuwit.org/). Four hundred cities in China have serious water scarcity problems; investments of ~$USD 650 billion are planned by 2020. HK has committed HK$20 billion to the rehabilitation and replacement of its water supply infrastructure.

The global state of water supply systems suggests an urgent need for more accurate, detailed, timely and cost-effective diagnostics. The proposed research promises to revolutionize UWSS condition assessment and defect detection. Just as medical advances have resulted in significant improvements in quality of life, we envision similar gains from research advances in the diagnosis of UWSS.

Established Methods of Defect Detection and Condition Assessment & their Limitations: Currently there are two classes of commercially available technologies. Class 1 comprises intrusive methods, where excavation is required to expose the pipe in order to either (i) cut a section for inspection; or (ii) to create an insertion point for a CCTV camera, a Smart Ball or the Sahara system. These methods are costly, disruptive, time consuming and labor intensive. Class 2 allows more normal operation and includes visual inspection, moisture detecting sensors, ground penetrating radar, and acoustic correlators . Although less disruptive, these methods tend to be short ranged, sensitive to noise, expensive, and often require traffic diversion. Despite the use of best-in-class commercially available technologies for leakage and fault reduction in the HK$20 billion UWSS rehabilitation program, it is anticipated that leakage can be reduced from 25% to 15% at best. All current technologies provide only limited spatial and temporal snapshots of the system state, and hence cannot adequately address the aging UWSS problem. As an example, on March 1, 2011 a 450 mm pipe burst in Wong Nai Chung Road caused severe flooding and extensive property damage, disrupting the lives of thousands of people and many businesses for 16 hours. Significantly, this pipe had been identified as fault free prior to the incident and the supervisory control and data acquisition (SCADA) system did not immediately detect the burst. Even after notification, 6 additional hohttps://www.puretechltd.comurs were needed to pinpoint and isolate the fault. This and similar failures clearly indicate the need of a long-term research program to diagnose pipe systems in real time.

LFW – An Emerging Method of Defect Detection and Condition Assessment: A class of methods that uses responsive transducers (usually sampling at several hundred Hz) to measure the pressure wave propagation in pipelines has emerged in recent years. As the waves travel, the change in waveform can be correlated with system features such as blocks and leaks – monitoring the pressure waves at key points in the UWSS offers a powerful non-intrusive method that can detect system defects over long ranges. Many theoretical studies have been conducted on these transient-based methods of fault detection and have culminated in a number of successful field trials by our team members - in Italy (Brunone B and Meniconi S), Canada and Mexico (Karney BW) and New Zealand (NZ) and China (Lee PJ). Other investigators (e.g., Covas D, Stephens ML, Allen M) have also had some success using low-frequency waves (LFW). In fact, LFW form the basis of the WaterWise@SG project in Singapore and the North Point trial project in HK.

Limitations of LFW: Field applications confirm that LFW techniques are technically feasible and demonstrate that UWSS faults are associated with characteristic pressure signatures. Nevertheless, the experience to date indicates a high probability of false positives and negatives in the predictions and an inability to associate features in wave signals to their origin. For example, the real-time system in HK shows sudden pressure fluctuations of the order of 15 m (i.e., 1.5 atmospheric pressure) on a daily basis, but its origin has yet to be determined (WSD, private communication).In addition, simulated sudden bursts can only be located to within ~45m even though the sensor was only 20 m away from the fault. This low spatial resolution of LFW was also identified in HK (simulated burst could only be located to within 42 m; WSD, private communication), Italy (leaks could only be located to within 800 m) and others. Moreover, existing LFW methods reveal significant gaps in knowledge of the accurate representation of system devices and interpretation of system response. Both the Singapore and HK systems are passive in that they have no mechanism to inject waves to probe the system. Such systems target faults that generate sudden large transients (e.g., bursts), but not blockages, pre-existing leaks and poorly performing devices, and small cracks. For example, the acoustic noise that emanates from a leak is local (i.e., does not propagate far in the fluid); thus, it cannot be detected by passive measurements of pressure signals. By contrast, the Smart UWSS envisioned here would use active wave probing, where it has been shown that wave reflections from leaks and other imperfections propagate far (~kilometers).   

A pipe system can be viewed as essentially a confined space with a longitudinal dimension much larger than a highly bounded lateral dimension. The pipes serve the function of flow conveyance and are inherently passive. However, dynamic controls are created at the ends of pipe mains over long distances – at pumps, valves, storage elements, and demand connections – the 3D turbulent flow behavior (e.g., pressure – flow) of these devices are inherently complex and boundary conditions must be introduced to describe their role in pressure wave propagation. These devices range from a simple isolation valve (that is normally either open or closed) to an active air valve with transient two-phase flow character, to an operating pump station. The dynamic response of the pipe system to any excitation is created, maintained and modified by actions at boundaries. Yet, LFW methodologies use models that originate in steady state theory, where inertial effects are neglected and their parameters are associated with steady turbulent flows, to represent these important actions. In this proposal, particular attention is paid to the study of critical boundary elements and the proper formulation of boundary conditions in the theoretical model (Tasks 1, 2, 4).

PART B: The Proposed System

The effective solution of the Smart UWSS problem requires a two-order-of-magnitude increase in the speed, range and accuracy with which UWSS faults can be pinpointed and rectified – the goal of this project. Indeed, it is evident that massive improvements in the management of UWSS are needed to safeguard health and sustainability issues brought about by rapid urbanization (in China alone ~200 million more people are expected to move to cities by 2030). We envisage a smart UWSS infrastructure as illustrated in the figure below. Currently, there is no system capable of producing reliable high-resolution “images” (we use “images” to mean that the Smart UWSS’s output will provide accurate identification & localization of anomalies) of the pipe system.

The proposed system is firmly founded on wave theory which is widely used to probe and characterize various media and to convey information in various applications that include non-destructive material testing, medical diagnostics, and underwater communications. The spatial range and spatial resolution of the probing waves are two essential properties underlying the system design. The spatial range, R, of waves travelling within the fluid in the pipe flow is of the order of a²/f²DVF^0.5, where a = wave speed, f = frequency, D=pipe diameter, and V is the average fluid velocity; F is a damping coefficient. The spatial resolution, l, is of the order of a/f. It is clear that as an increase in the wave frequency is accompanied by a reduction in range and increase in resolution, and vice-versa. Using typical values of a, D, V and F, LFW have spatial resolution in the order of 50-100 m and range of the order of ~10 km. Therefore, LFW have a wide range but relatively low spatial resolution; high frequency waves (HFW) have the opposite attributes. The proposed work represents the marriage of the relative strengths of both LFW and HFW techniques.

Currently LFW techniques are limited in several ways (see paragraph “Limitations of LFW”). With better device characterization (Tasks 1, 2), better signal design, conditioning and analysis (Tasks 1, 3), LFW methods can be usefully deployed on restrictive parts of the system such as within District Metering Area (DMA) boundaries - the approach is ideally suited for obtaining rapid but blurred images of a large part of the pipe system, where zones that contain potential problems are delineated for further investigation (Task 4).

On the other hand, the HFW (10 to 40 kHz) methods, developed in Tasks 1, 2 have the potential to provide high resolution images when applied to problematic zones identified by LFW (Task 4). Such waves have resolution of the order of centimeters to meters and are less susceptible to interference from the ever-present system noise (frequencies << 1 kHz). Their range is of the order of 100 m to a few kilometres which is sufficient to investigate the zones that are deemed problematic by LFW. This HFW approach is supported by our preliminary numerical and experimental research. In particular, the use of the HFW method in pipe diagnostics has been successfully tested in proof-of-concept field trials across 30 sites in NZ and China in collaboration with a leading water utility company. Preliminary results showed that HFW can be transmitted across distances of over 300 m. A comprehensive theoretical and experimental understanding of HFW propagation in a turbulent pipe flow will be developed in this research (Tasks 1, 2) – which would lead to paradigm shifts in real-time pipeline condition assessment (Tasks 3, 4). HFW will be generated by compact, piezoelectric actuators that result in no water loss or system disruption (Tasks 1, 3).

In addition to their imaging capability, HFW are also essential to the novel concepts of both acoustic communication between in-pipe sensors and executing hydraulic controls (Task 3). Existing systems (e.g., WaterWise@SG) use a network of integrated multi-sensor probes to acquire and transmit data in real-time through wireless sensing nodes, where the connection of the sensing nodes to the pipes require specific access points. Issues of inaccessibility, cost and security would not favor the creation of access points (WSD, personal communication). With the use of HFW we can explore the use of sensors that reside inside the pipe and communicate acoustically with existing receiver systems or base-stations (e.g., SCADA sites) where it can be then relayed in real-time using existing wireless connections. Acoustic communication inside the pipe is selected because it also uses HFW as opposed to EM signals which are too limited in range and require excessive energy for transmission in water. Kokossalakis G provides initial tests that point to the feasibility of using HFW for in-pipe communications. The extensive knowledge and experience of our team members Yang TC and Zhang X in the field of acoustic communications in water will be actively leveraged.

This research aims to resolve a number of fundamental issues that include: What are the distinct signatures of defects such as leaks, blockages (air and solid) and wall-thinning that can be used to detect defects using waves (Task 1, 2, 4)? What is the impact of different hydraulic elements on both LFW and HFW propagation – including effects of multi-path, nonlinearities, turbulence, noise, and system topology (Tasks 1, 2)? What is the spatial range of LFW&HFW and which frequency bands are best for in-pipe communications (Tasks 1, 2, 3)? What is the optimal pipe flow diagnostic strategy in terms of sensor locations and signal design (Tasks 3, 4)? What regularization strategy should be used to identify defects and assess the condition of the pipe system (Task 4)? Under what flow conditions should the system be probed to detect small (difficult) anomalies (Task 4)?