*4.3. Architecture of Underground WSN Systems for Pipelines*

Different networks and communication protocols have been used to monitor the underground and aboveground water pipelines for the transmission and propagation of sensor data to end users. A WSN consists of various units that work together to gather desired data in a specific environment where communication can be established over a wireless channel, including sensing, computing, and communication devices [64,65]. There are three tiers—(1) sensor tier; (2) master-node tier; and (3) end user—in any underground WSN system for pipelines, as depicted in Figure 4. In the sensor tier, sensors are placed along the pipeline network to wirelessly transmit assigned data to the master node, which subsequently transmits data to the end user(s). The main distinction between underground and terrestrial WSN systems is the communication medium (soil or air) [66].

erosion and ultimately results in sinkholes.

*4.3. Architecture of Underground WSN Systems for Pipelines*

BenSaleh et al. [58]

BenSaleh et al. [63]

Developing energy-efficient nodes with adequate sleep and wake-up mechanisms.

Develop a robust and reliable system, which is costeffective, scalable, and customizable in future.

Tariq et al. [59] Improvement in the communication radius efficiently. Computer Sciences 2013

From Table 4, it can be concluded that all suggestions for future research are related to the fields of computer science, electrical, electronics, and telecommunication sectors. Therefore, it can be considered that all researchers, in their respective review articles, aimed to present the challenges encountered in the field of WSN, sensor-deployment strategies, and communication radius of the system. However, in the fields of soil mechanics, civil infrastructure, and geology, the after effects of leakage on the surrounding soil must be considered. As mentioned earlier, leakage can lead to soil

Different networks and communication protocols have been used to monitor the underground and aboveground water pipelines for the transmission and propagation of sensor data to end users. A WSN consists of various units that work together to gather desired data in a specific environment where communication can be established over a wireless channel, including sensing, computing, and communication devices [64,65]. There are three tiers—(1) sensor tier; (2) master-node tier; and (3) end user—in any underground WSN system for pipelines, as depicted in Figure 4. In the sensor tier, sensors are placed along the pipeline network to wirelessly transmit assigned data to the master node,

Electronics and

Electronics and

Communication <sup>2015</sup>

Communication <sup>2013</sup>

**Figure 4.** Typical WSN architecture for underground water pipeline monitoring. **Figure 4.** Typical WSN architecture for underground water pipeline monitoring.

Many wireless sensor communication protocols and interfaces, such as Wi-Fi, ZigBee, LoRa, and Bluetooth, have been used to overcome multiple transducer connectivity problems [67,68]. Table 5 explains the characteristics of commonly used communication protocols along with their properties such as speed, range, frequency, and limitations. Bluetooth is applicable for communication over a short range of 1–100 m with better data transmission speed in comparison to ZigBee and GPRS. Many wireless sensor communication protocols and interfaces, such as Wi-Fi, ZigBee, LoRa, and Bluetooth, have been used to overcome multiple transducer connectivity problems [67,68]. Table 5 explains the characteristics of commonly used communication protocols along with their properties such as speed, range, frequency, and limitations. Bluetooth is applicable for communication over a short range of 1–100 m with better data transmission speed in comparison to ZigBee and GPRS. Certain communication protocols such as GPRS and 3G possess low data transfer speed but their range is up to 10 km. Some communication protocols such as 5G, 4G, 3G, and LTE provide a somewhat longer communication range with high data-transmission speed, whilst requiring a large spectrum fee [69,70]. It is, therefore, essential to select a communication protocol based on the end application and availability of required resources to achieve desired results of cost-effectiveness, reliability, and low power consumption. The communication protocols listed in Table 5 are in order of increasing range of different protocols.

**Table 5.** Communication protocols and their properties [69,70].


*4.4. Water Pipeline Leakage Monitoring Methods Based on WSN System*

The loss of water is one of the most critical issues, particularly in urban localities. In an urban framework, the occurrence of any severe pipeline leakage issue such as water supply interruption and traffic flow interruption may intensify, thereby resulting in an increase in maintenance costs. Therefore, the implementation of a robust, reliable, and energy-efficient WSN-based system is imperative to overcome water loss issues and after effects of leakage [71]. Ng et al. [72] proposed a vibration

sensor-based sound variation detection device that records suspicious water leakage sounds and compares them with normal pipeline sounds, thereby identifying the symptoms of water leakage. The device was used on two types of pipelines—metal and PVC—but not concrete ones. The said device has a disadvantage, that is it only functions properly when installed near a leak point, and it is immensely difficult to predict the exact location of leakage. It is also not clear to judge the utility of the device when on concrete pipes.

Four things—cost efficiency, reliability, power consumption, and sensor placement across the pipeline network—must be considered regarding monitoring underground sewer/water networks and sinkholes using WSN. Cattani et al. [73] proposed a method called ADIGE based on the long-range LoRa WSN technology to reduce the number of gateways uploading data gathered from wireless sensors. The proposed ADIGE system comprised two parts—(1) sensing, controlling, and collection of data; and (2) data management, including data fusion and analysis—as depicted in Figure 5a. This method was designed to achieve high reliability and energy efficiency along with wider coverage of water pipeline monitoring using a large number of low-cost sensors for collecting more real-time information at multiple locations instead of using lesser numbers of expensive sensors installed at fewer locations. However, using the WSN system to accurately identify leakage locations inside a long-running water pipeline and the effects caused by leakage remains a major challenge for researchers. *Sustainability* **2019**, *11*, x FOR PEER REVIEW 12 of 25

**Figure 5.** Leakage monitoring methods using different combinations and algorithms for leak detection and localization: (**a**) System combination of method based on the long-range LoRa WSN technology; (**b**) different types of methods and algorithms used for leak detection and localization. **Figure 5.** Leakage monitoring methods using different combinations and algorithms for leak detection and localization: (**a**) System combination of method based on the long-range LoRa WSN technology;(**b**) different types of methods and algorithms used for leak detection and localization.

Karray et al. [74] also proposed a method to identify leakage locations within a pipeline network. This method involved the use of a combination of leakage detection algorithms, localization Karray et al. [74] also proposed a method to identify leakage locations within a pipeline network. This method involved the use of a combination of leakage detection algorithms, localization algorithms,

algorithms, and system-on-chip (SoC) architecture. This solution was based on two methods, as described in Figure 5b. Instead of using multiple algorithms at once to compromise battery life, the

m<sup>3</sup> volume of water was used to supply water using 1 hp power supplied by the laboratory test bed motor. A force-resistive sensor (FSR) was used to measure the pressure within the pipelines. The said method was proposed for use in long-distance pipelines installed on the ground but was not suitable

Sun et al. in [75] suggested a new solution for water pipeline monitoring, leakage, and burst detection. Real-time leakage and burst detection were set as the primary objectives of this method. Considering the propagation of sensor signal data in soil, a WSN system based on magnetic induction (MI) was proposed. The sensors were distributed into two layers—(1) the hub layer and (2) in-soil sensor layer. In the hub layer, pressure and acoustic sensors were deployed at checkpoints and pump stations inside pipelines to measure pressure and vibration changes caused by leakage, as depicted in Figure 6. After measuring the values of pressure and vibration, data were wirelessly sent to the administration center using MI channels. Whereas in the in-soil layer, sensors were deployed along the pipeline to measure different soil parameters such as temperature and humidity. This was accomplished by rolling MI relay coils around the pipes. Beyond the system architecture and

for use with underground pipelines.

and system-on-chip (SoC) architecture. This solution was based on two methods, as described in Figure 5b. Instead of using multiple algorithms at once to compromise battery life, the author preferred using single algorithm—the Kalman Filter (KF)—to conserve energy. The method used for this purpose was referred to as energy-aware reconfigurable sensor node for water pipeline monitoring (EARNPIPE). A plastic pipeline capable of supporting pressures of up to 25 bar and 1000 m<sup>3</sup> volume of water was used to supply water using 1 hp power supplied by the laboratory test bed motor. A force-resistive sensor (FSR) was used to measure the pressure within the pipelines. The said method was proposed for use in long-distance pipelines installed on the ground but was not suitable for use with underground pipelines.

Sun et al. in [75] suggested a new solution for water pipeline monitoring, leakage, and burst detection. Real-time leakage and burst detection were set as the primary objectives of this method. Considering the propagation of sensor signal data in soil, a WSN system based on magnetic induction (MI) was proposed. The sensors were distributed into two layers—(1) the hub layer and (2) in-soil sensor layer. In the hub layer, pressure and acoustic sensors were deployed at checkpoints and pump stations inside pipelines to measure pressure and vibration changes caused by leakage, as depicted in Figure 6. After measuring the values of pressure and vibration, data were wirelessly sent to the administration center using MI channels. Whereas in the in-soil layer, sensors were deployed along the pipeline to measure different soil parameters such as temperature and humidity. This was accomplished by rolling MI relay coils around the pipes. Beyond the system architecture and framework presented in this article, more work regarding evaluating the system performance is required to be performed by deploying MI-based WSN for underground pipeline monitoring (MISE-PIPE) in real-life applications. *Sustainability* **2019**, *11*, x FOR PEER REVIEW 13 of 25 framework presented in this article, more work regarding evaluating the system performance is required to be performed by deploying MI-based WSN for underground pipeline monitoring (MISE-PIPE) in real-life applications.

**Figure 6.** System architecture of magnetic induction (MI)-based WSN for underground pipeline **Figure 6.** System architecture of magnetic induction (MI)-based WSN for underground pipeline monitoring (MISE-PIPE) [75].

#### monitoring (MISE-PIPE) [75]. *4.5. Sewer Pipeline Leakage Monitoring Methods Based on WSN Systems*

applicable to water, sewer, and pipelines [77].

the condition and performance of pipelines [77].

*4.5. Sewer Pipeline Leakage Monitoring Methods Based on WSN Systems* Leakage in sewer pipelines needs to be considered as a potential source of underground cavity creation and sinkholes in urban areas. According to Kuwano et al. [76], the sewer pipe mains buried underground for over 25 years demonstrate a remarkable increase in cracks and defects. Researchers and scientists have contributed toward the development of means to address issues related to cracking, leakage, and bursting of sewer pipelines. Kim et al. [77] proposed a novel RFID-based autonomous mobile device monitoring system for pipelines, called RAMP. The mobile robotic device comprising sensors (visual-, chemical-, pressure-, and SONAR-sensing) was capable of moving inside the pipelines along the direction of fluid flow whilst monitoring the presence of any defect on its way and localizing the same. The primary function of the robotic agent was pipeline inspection. However, Leakage in sewer pipelines needs to be considered as a potential source of underground cavity creation and sinkholes in urban areas. According to Kuwano et al. [76], the sewer pipe mains buried underground for over 25 years demonstrate a remarkable increase in cracks and defects. Researchers and scientists have contributed toward the development of means to address issues related to cracking, leakage, and bursting of sewer pipelines. Kim et al. [77] proposed a novel RFID-based autonomous mobile device monitoring system for pipelines, called RAMP. The mobile robotic device comprising sensors (visual-, chemical-, pressure-, and SONAR-sensing) was capable of moving inside the pipelines along the direction of fluid flow whilst monitoring the presence of any defect on its way and localizing the same. The primary function of the robotic agent was pipeline inspection. However, the system requires a structure with improved fluid resistance and enhanced mobility of the robotic agent because

Similarly*,* PIPENET is a system proposed for pipelines with large diameters, such as sewer and drainage pipelines, and comprises a WSN-based system for underground pipeline leakage detection and localization [78]. For real-time monitoring, the system was deployed in collaboration with Boston Water and Sewer Communication (BWSC), USA. Their deployment primarily focused on two critical applications—(1) sewer collector's water level monitoring and (2) hydraulic and water quality monitoring [79]. The method aimed to detect and localize the presence of any leakage or burst within the pipeline by collecting the hydraulic and vibration data at a high sampling rate using different sensing parameters—flow, pH, vibration, and pressure—and data collection was performed for an extended period of over 22 months in the city of Boston, USA. The experiment was performed both inside a laboratory and in the field to compare the experimental results and develop an algorithm for leakage detection and localization. Some limitations of this method include false alarms and low energy efficiency. Considering sewer pipelines, robot-sensing devices have been preferred to monitor

Most robots used for the inspection of sewer pipelines moved along one direction (straight). In contrast, KANTARO was an innovative, fast, and robust sensing device intended for use in sewer-

the system requires a structure with improved fluid resistance and enhanced mobility of the robotic agent because at present its usage is only limited to inside straight pipelines. Additionally, to achieve at present its usage is only limited to inside straight pipelines. Additionally, to achieve better results, energy efficiency and sensor power must be improved. Such methods are equally applicable to water, sewer, and pipelines [77].

Similarly, PIPENET is a system proposed for pipelines with large diameters, such as sewer and drainage pipelines, and comprises a WSN-based system for underground pipeline leakage detection and localization [78]. For real-time monitoring, the system was deployed in collaboration with Boston Water and Sewer Communication (BWSC), USA. Their deployment primarily focused on two critical applications—(1) sewer collector's water level monitoring and (2) hydraulic and water quality monitoring [79]. The method aimed to detect and localize the presence of any leakage or burst within the pipeline by collecting the hydraulic and vibration data at a high sampling rate using different sensing parameters—flow, pH, vibration, and pressure—and data collection was performed for an extended period of over 22 months in the city of Boston, USA. The experiment was performed both inside a laboratory and in the field to compare the experimental results and develop an algorithm for leakage detection and localization. Some limitations of this method include false alarms and low energy efficiency. Considering sewer pipelines, robot-sensing devices have been preferred to monitor the condition and performance of pipelines [77].

Most robots used for the inspection of sewer pipelines moved along one direction (straight). In contrast, KANTARO was an innovative, fast, and robust sensing device intended for use in sewer-pipeline inspection. The device could move in a straight path as well as bend around the curves. For this, a particular patented mechanism called the "Sewer Inspection Robot (nSIR) Mechanism" was developed [80]. KANTARO was equipped with a fisheye camera mounted to detect any damage or blockage within the sewer pipeline network. It could only fit in pipelines with internal diameters in the range of 200–300 mm and included lithium-polymer batteries for power supply to motors, sensors, computer systems, and underground wireless communication modules.

#### *4.6. Pipeline Leakage Monitoring Methods Based on Computer Vision and Image Processing*

Other such approach involved the use of closed-circuit television (CCTV) image processing by concerned researchers for the detection of defects and damages in sewer pipelines. Manual interpretation of images and videos has previously been used, and those methods were observed to be more time consuming, labor intensive, and the results obtained were deemed to be less reliable and possibly inaccurate. In contrast, automated defect detection methods have been proposed by researchers using artificial intelligence (AI) and CCTV image and video processing techniques, such as the deep learning technique called the faster region based convolutional neural network (faster R-CNN) [81]. The faster R-CNN model has been demonstrated to detect defects in sewer pipelines much faster with higher accuracy.

However, the faster R-CNN approach only works on static images. Similarly, the authors in [17] used an automated approach to detect cracks, deformations, joint displacements, and settled deposits in sewer pipelines. Their approach was based on image processing and mathematical formulations to analyze the output obtained from CCTV images. To validate the performance of the proposed method, the authors in [17] constructed a confusion matrix. The results observed for the accuracy of crack, displaced-joint, and settled-deposit detection were found to 74%, 65%, and 54%, respectively, which could be improved by increasing the number of images captured. A drawback of the proposed methodology was that the method relied heavily on the expertise and experience of operators. From the literature overviewed in earlier sections it is evident that over time, technologies have developed from the visual inspection of sewer/water pipelines to the application of wired sensors, wireless sensors, IoT, big data, and AI technologies, as illustrated in Figure 7.

methods over time.

faster with higher accuracy.

sensors, IoT, big data, and AI technologies, as illustrated in Figure 7.

pipeline inspection. The device could move in a straight path as well as bend around the curves. For this, a particular patented mechanism called the "Sewer Inspection Robot (nSIR) Mechanism" was developed [80]. KANTARO was equipped with a fisheye camera mounted to detect any damage or blockage within the sewer pipeline network. It could only fit in pipelines with internal diameters in the range of 200–300 mm and included lithium-polymer batteries for power supply to motors,

Other such approach involved the use of closed-circuit television (CCTV) image processing by concerned researchers for the detection of defects and damages in sewer pipelines. Manual interpretation of images and videos has previously been used, and those methods were observed to be more time consuming, labor intensive, and the results obtained were deemed to be less reliable and possibly inaccurate. In contrast, automated defect detection methods have been proposed by researchers using artificial intelligence (AI) and CCTV image and video processing techniques, such as the deep learning technique called the faster region based convolutional neural network (faster R-CNN) [81]. The faster R-CNN model has been demonstrated to detect defects in sewer pipelines much

However, the faster R-CNN approach only works on static images. Similarly, the authors in [83] used an automated approach to detect cracks, deformations, joint displacements, and settled deposits in sewer pipelines. Their approach was based on image processing and mathematical formulations to analyze the output obtained from CCTV images. To validate the performance of the proposed method, the authors in [17] constructed a confusion matrix. The results observed for the accuracy of crack, displaced-joint, and settled-deposit detection were found to 74%, 65%, and 54%, respectively, which could be improved by increasing the number of images captured. A drawback of the proposed methodology was that the method relied heavily on the expertise and experience of operators. From the literature overviewed in earlier sections it is evident that over time, technologies have developed

sensors, computer systems, and underground wireless communication modules.

*4.6. Pipeline Leakage Monitoring Methods Based on Computer Vision and Image Processing*

**Figure 7.** Illustration of technological developments in underground pipeline network inspection **Figure 7.** Illustration of technological developments in underground pipeline network inspection methods over time.

Using WSN-based systems, previously discussed sewer and water pipeline surveillance models (Sections 4.4 and 4.5) are summarized in Table 6. This table provides a detailed perspective of WSNbased sewer and water pipeline leakage monitoring methods in terms of the protocols used for wireless communication between the transmitter and receiver, type of wireless sensors used (noninvasive or invasive sensors, static or mobile sensors, etc.), field of application, and water or sewer pipeline network. Table 6 makes it easier to determine the most feasible method to monitor water and/or sewer pipelines under certain conditions. Meanwhile, it also helps to select the most feasible methods, which can be further modified for the application of sinkhole due to leakage. For example, Sun et al. [75] used sensors both inside and outside the pipelines to detect various factors related to leakage. In such methods, more soil property measuring sensors can be added to further collect data Using WSN-based systems, previously discussed sewer and water pipeline surveillance models (Sections 4.4 and 4.5) are summarized in Table 6. This table provides a detailed perspective of WSN-based sewer and water pipeline leakage monitoring methods in terms of the protocols used for wireless communication between the transmitter and receiver, type of wireless sensors used (non-invasive or invasive sensors, static or mobile sensors, etc.), field of application, and water or sewer pipeline network. Table 6 makes it easier to determine the most feasible method to monitor water and/or sewer pipelines under certain conditions. Meanwhile, it also helps to select the most feasible methods, which can be further modified for the application of sinkhole due to leakage. For example, Sun et al. [75] used sensors both inside and outside the pipelines to detect various factors related to leakage. In such methods, more soil property measuring sensors can be added to further collect data related to the factors (soil moisture, density, temperature, overburden, and porosity, among others) that contribute to the occurrence of sinkholes.


**Table 6.** Comparison of various methods for leakage detection in water and sewer pipelines.

<sup>1</sup> NIS: Non-invasive sensors (sensors placed outside a pipeline), <sup>2</sup> IS: Invasive sensors (sensors placed inside a pipeline).

#### *4.7. Sinkhole Monitoring and Detection Methods*

A record of the past 60 years of Gauteng, South Africa, reports the occurrence of more than 3000 sinkholes, including natural and human-induced ones [93]. The Maryland geological survey reported a total of 139 sinkhole occurrences, 51 of which occurred naturally, while the remainder were the human-induced type [94]. According to the British Geological Survey (BGS) (a British government institute), 10% of all sinkholes occur as a result of leakage in underground pipelines [82]. As discussed in previous sections (Sections 4.4–4.6), various methods are adopted for leakage monitoring. However, sinkholes due to leakage require proper concentration as they result in considerable damage. According to the US Geological Survey, nearly \$300 million per year is spent on the reconstruction of, and compensation for, damage caused by sinkholes to the infrastructure [95]. The frequency of sinkhole occurrence has increased considerably in South Korea, since 2010, mainly because of underground construction in urban areas. An account of the total occurrence of sinkholes in Seoul, South Korea, between January and July 2014 reported that ruptured sewer pipelines resulted in approximately 85% of all man-made sinkholes, 18% were triggered by excavation during construction activities, and the remaining 3% were caused by leakage in freshwater pipelines [96].

Sinkholes occurs on roads, highways, and railways owing to the infiltration of leaking underground pipelines or rainwater. These cavities can be detected beneath the surfaces using different monitoring and detection methods. Although many sinkhole detection methods have been developed, it is still difficult to predict the formation of sinkholes. There are alternative methods to predict when and where sinkholes will form. Researchers have drawn maps for sinkhole-prone regions that predict risk based on the ground composition or other local sinkholes; however, this still does not provide a definite answer.

#### 4.7.1. Sinkhole Monitoring Methods Using Image Processing and Radar

Many researchers have contributed toward the prediction of the occurrence of natural sinkholes via the use of diverse technologies, such as the trenching method along with ground penetration radar (GPR), electrical resistivity tomography (ERT), high-precision leveling [97], Brillouin optical fiber sensor [98], high-precision differential leveling [99], laser imaging detection and ranging (LIDAR), and interferometric synthetic aperture radar (InSAR) [100]. The combined use of the geographic information system (GIS) and analytical hierarchical process (AHP) has previously been proposed to identify hazard zones in Kuala Lumpur and Am pang Jaya of Malaysia that are susceptible to the occurrence of natural sinkholes on the basis of five criteria—lithology, groundwater level decline, soil types, land use, and proximity to groundwater wells [4]. However, the human-made (owing to leakage) requires proper concentration.

As in the UK, the British Geological Survey used digital map data for geohazard monitoring in urban areas. This digital map data can be accessed from GIS [101]. High-resolution ground-based InSAR (GB-InSAR) and LiDAR are methods available for sinkhole monitoring. These technologies use GIS data for sinkhole mapping and are particularly effective for sinkhole tracking [100]. However, it is challenging to discern sinkholes or subsurface features based on GIS data or satellite data [102]. Soil erosion gradually leads to collapse, and the application of GB-InSAR, LiDAR, scanning, and photogrammetry methods have to scan and analyze different areas continuously to observe the changes over time, which can be expensive and time-consuming. Therefore, the current practices in urban areas for monitoring and detection of the sinkhole are limited.

Similarly, in some developed countries, ground penetration radar (GPR) is currently used for sinkhole monitoring and detection, especially in urban areas. The results of a GPR rely on the reflection of a high frequency (25–1000 MHz) electromagnetic (EM) pulse from subsurface contacts and other anomalies such as boulders and cavities. [103]. GPR uses radar pulsations to predict the changes in the subsurface. This method uses EM radiation of the radio spectrum and detects the reflected signals from the subsurface structures [104]. However, there are limitations associated with GPR, as noise effects the magnetic techniques, and the presence of clayey soils affects the transmission of EM signals from GPR [105]. Previously, the application of GPR was excluded from the investigation in Nigeria owing to the presence of clay [106].

Other methods include geomorphological mapping, borehole drilling, and air photo interpretation. Table 7 lists the methods currently available to address challenges and obtain solutions for sinkholes; their advantages and limitations are also discussed.


#### 4.7.2. Human-Induced Sinkhole Monitoring Methods Using WSN

Soil strata and profile vary from for different regions and are characterized based on the unique geomorphological and hydrological conditions, and pipelines are not limited to areas with specific geology. If we consider the subsurface of any urban area, miles of water and sewer pipelines are passing through different soil profiles. Therefore, the effect of water leakage from sewer or water pipelines will be different for different subsurface soil strata.

A leakage in the sewer or water pipeline results in the change in underground soil and other various problems mentioned above. Over the past few years, numerous cases of sinkhole creation owing to the rapid deterioration of underground sewer and water pipelines have been reported. To overcome this issue, researchers have developed sensor-based pipe safety units to detect leakage by analyzing water leaks and pipeline behavior. Pipeline data are collected by the use of smart sensors and WSN [3]. However, the system is still being developed, and the primary purpose of the program is to develop a sinkhole risk index (SRI) via the real-time monitoring of underground construction activities. This is an individual research endeavor aimed at the prevention of sinkhole creation due to leakages or ruptures in underground sewer pipelines owing to human interference. In the said study, experiments were conducted on a buried pipeline, exposed pipeline, exposed acrylic pipeline, and test-object pipe. During the experiments, the concerned researchers used cables for the connection between sensors.

However, previously, researchers have suggested several methods for the study of the mechanism of natural sinkholes by considering different factors that influence the occurrence of natural sinkholes. Therefore, different numerical, mathematical, and experimental models have been developed. Each researcher in their experimental model considered different sinkhole mitigating factors and soil types to understand the mechanism.

Researchers considered sea side soil (mudflat, sea clay) as a soil type, and subsurface void growth, ground water level, salt dissolution, and overburden pressure as the factors responsible for sinkhole or subsidence in their experiments [119,120] as the researchers aimed to understand the mechanism of sinkhole along the seaside. Therefore, this sinkhole mechanism cannot be true for other cases, as other soil types need to be considered (clay, sand, bedrock, carbonates, etc.) and other sinkhole mitigating factors (size of cavity, subsurface soil properties, aquifer, etc.). Similarly, Tao et al. [121] also performed

a numerical simulation to understand the mechanism of natural sinkholes owing to changes in ground water table. Sand and clay were used as the soil profiles for the sinkhole model. However, similar to previous contributions, this study was just limited to understanding the natural sinkholes mechanism. In addition, it must be improved to increase the simulation capabilities.

#### **5. Discussion and Conclusions**

Numerous conventional techniques have been used over the years for the accurate detection of water and sewer pipeline leakage and prevention of sinkhole creation. However, there are a few limitations in the execution of conventional techniques such as high cost, need for large workforce and experts, long execution times, and unreliable results. Over the years, conventional methods have been replaced by methods based on IoT, WSN, smart sensors, and AI; these methods are more manageable and reliable.

This article focuses on research contributions regarding the use of WSN for water and sewer pipeline leakage and sinkhole monitoring and detection, thereby providing methods to prevent the creation of sinkholes in urban areas. The proposed review was performed following a two-pronged methodology—patent analysis and literature review of methods used for monitoring underground pipeline leakage and sinkhole creation using WSN-based systems. Both patent analysis and literature review demonstrated a lack of research on the prevention or monitoring of sinkhole creation caused by leakage in water and/or sewer pipelines.

The development of modern technology has led to the development of improved water and sewer leakage monitoring systems. However, modern, technology-based monitoring systems require efficient communication technologies such as 5 G in combination with WSN to create an advanced infrastructure such as underground fluid transportation networks. Over the last decade, new wireless communication networks have been developed to serve a range of new applications and deployment scenarios. Similarly, smart cities are being developed, where WSN and 5 G are considered to be one of the key enabling technologies that will provide smart solutions to smart cities to improve the overall quality of life [122].

#### *5.1. Review Findings*

The results of a preliminary search of patent databases demonstrate that a majority of patents published in the past 18 years were related to overcoming leakage problems associated with water and sewer pipelines. However, they lack in terms of suggesting measures to counter the after effects—sinkholes and ground subsidence—of leaked water and sewer pipelines. Additionally, existing patents for sinkhole monitoring and detection face several limitations. For instance, the "sinkhole detector" can only detect sinkholes near the installed device [45]. Similarly, the patent named "device for detecting changes in underground medium" explored the use of ultrasonic sensors; however, the propagation of ultrasonic signals can be easily affected by the soil medium, traffic-induced vibrations, and sound effects generated by different means of transportations or other sources.

Similarly, the literature review conducted in this study revealed that majority of researchers are concerned with overcoming the water losses caused by leakages in sewer and water pipelines. Particularly, a burst or leak of a sewer pipeline can endanger human lives as well as damage nearby infrastructures such as railways and highways; this is because most cases of human-induced sinkholes and ground subsidence are caused by leakages in sewer pipelines [123]. Compared to naturally occurring sinkholes, human-induced sinkholes incur less damage to infrastructures and human lives and this is why they have not been considered a significant issue by researchers in the past. However, over the last few years, sinkholes caused by leakage in water and sewer pipelines have become a severe problem across the world. Natural, as well as human-induced, sinkholes cannot be directly detected, and in this regard, the development of SRI or a sinkhole risk model is considered likely to play the role of a bridge between leaky pipelines and sinkhole detection [3].

However, no prominent studies have yet been found concerning the monitoring, detection, prevention, and localization of human-induced sinkholes. There exist many factors—soil type, topography of the area under consideration for sinkhole evaluation, history of sinkhole occurrence, recharge-area category, thickness and depth of underground cavity, groundwater table, and public safety—that need to be considered during sinkhole risk or risk index evaluation [124]. Similarly, other factors such as temperature, moisture content, and porous water pressure of soil, which may also change owing to a leakage or rupture in underground water or sewer pipelines, require proper attention.

#### *5.2. Future Research Directions*

After a critical analysis of various methodologies used by researchers previously in the domain of water and sewer pipeline leakage and sinkhole induction, the authors believe that there are certain major challenges that must be addressed on priority. Sinkhole formation, or collapse of the ground caused by a rupture or leakage in underground sewer and water pipeline, requires considerable attention. Similarly, there exists an urgent need for the development of a human-induced SRI to determine the magnitude of the occurrence of sinkholes induced by underground sewer and water pipeline leakage.

The authors advocate toward utilizing a set of technological tools to assess the occurrence of sinkholes by prioritizing the critical factors associated with them. State-of-the-art technologies such as AI and WSN can be used in combination to monitor and access the geographical changes and the locations that are at a higher risk of sinkhole formation owing to leaked underground pipelines. Image processing techniques based on AI platforms can be implemented to monitor disruptions at both the ground surface level and the underground pipeline level. Combining these techniques with WSN, which uses sensors to determine soil properties such as moisture, density, porosity, pH, temperature, and bearing capacity, can help to gain a better understanding of the processes at this geographical location.

As WSN was previously used to analyze the physical properties (vibration, flow rate, sound, and so on) of pipelines and their flow in order to monitor leakage, it did not concentrate on the properties of the soil profile, which can change after the interaction of soil with water leaked from the sewer/water pipelines. Thus, the systematic approach proposed in this paper can pave the path for future research in this area. In the future, such technological tools can be used in smart cities to overcome the issues of pipeline leakage and sinkhole formation, where radio technologies such as 5 G enables smart city networks to support interconnected infrastructure elements and to manage big data from existing smart infrastructures [122].

An evaluation of previous research publications in this area clearly shows that a majority of the methods used are based on laboratory experiments and the findings and proposals appear to be difficult for practical application. Therefore, to achieve significant improvement, we suggest implementing collaborative exercises between research institutes and water supply agencies.

#### *5.3. Conclusions*

Based on the analysis reported in this study, it can be concluded that the key focus of research in existing studies has been concerned with (1) the use and enhancement of wireless sensor networking systems; (2) types and quality of sensors; and (3) improvement of hardware and software systems adopted for underground water and sewer pipeline monitoring. Although soil properties, such as bearing capacity, water content, soil density, air voids, pH level, ground subsidence, and others, change owing to pipeline leakage, these have seldom been considered in previous investigations. Nonetheless, these parameters are critical with regard to triggering sinkhole formation and ground subsidence owing to leakage in water and sewer pipelines. This article presents a state-of-the-art review of different methods for monitoring sinkhole and underground water and sewer pipeline leakage and their effects whilst considering the use of applications based on IoT and WSN systems over the past two decades. This review would be of interest to researchers intending to work in this area as it serves as a platform to direct future studies whilst accounting for challenges likely to be encountered during the same.

**Author Contributions:** H.A. and J.C. wrote the original draft supervised by J.C.

**Funding:** This work was supported and funded by the Basic Science Research Program through the National Research Foundation of Korea (NRF) and funded by the Ministry of Education (2016R1A6A1A03012812 and 2017R1D1A1B03036200).

**Conflicts of Interest:** The authors declare no conflict of interest.
