Pipeline networks are commonly used to transport water, oils and gases over long distances in cities, housing estates and industrial areas. While some pipelines are subject to faults such as weld defects that are caused by a variety of reasons which include poor quality of pipe materials and cracking due to strain [1], most pipelines are openly exposed to environmental conditions, such as rain and floods, and damage due to human error and vandalism, as well as unintended damage due to construction and development activities. For overground pipelines, although structural failures, such as cracks and leakages can be identified visually, often, these failures can only be detected at their critical stages when they become disruptive. For buried, underground and submarine pipelines, where visual inspection is not possible, inspection tools, which are either human-operated [2,3] or automated [4,5], are used.

Human-operated inspection tools are often inefficient since intensive human participation is required in order to inspect relatively long distances of pipelines daily. Therefore, the employment of automated inspection tools has become increasingly popular. Prior to 2010, growing popularity in the use of ultrasonic-based inspection methods was observed, where research in the optimisation of the geometrical design of ultrasonic-phased arrays for guided wave inspection was actively conducted [6,7,8]. Since 2010, there has been a shift of interest from ultrasonic transducers to acoustic emission (AE) sensing methods [9,10,11]. There has also been a growing interest among researchers in inspection methods based on the analysis of hydraulic parameters such as pressure and flow rate [12,13,14]. At the same time, the employment of magnetic flux leakage (MFL) sensing for pipeline failure detection has become increasingly relevant due to its applicability across various types of pipeline failures [15,16,17].

Technologies such as ground-penetrating radar (GPR) [2], infrared thermography [18] and impact echo (IE) [19] are widely employed in the industry, especially in human-operated inspection tools. However, the dimensions, designs and operational requirements of the sensing devices for these technologies have constrained them from being adopted in remote and automated pipeline monitoring systems [20]. In conjunction with the extensive implementation of Industry 4.0 [21,22,23], sensors, such as ultrasonic, acoustic, hydraulic and Hall effect sensors, have been retrofitted in the form of wireless sensor networks in existing pipeline networks [24,25,26]. These sensors are often small in size, inexpensive and can be easily interfaced with embedded systems. These technologies became increasingly relevant with the deployment of autonomous robots for the direct measurement of the magnitudes of defects in pipeline networks [27,28]. The emergence of unmanned aerial vehicles, better known as drones, for the detection of surface defects of pipelines has overcome the limitations of remote monitoring, at the same time reducing the workload required to monitor the integrity of pipelines in large plants [3].

Since 2010, many researchers have been focusing on developing efficient pre-processing and pipeline failure categorisation techniques for data or signals collected from sensors by employing machine learning methods [29,30] suited to embedded platforms. The pre-processing of data or signals using such methods as Kalman filter [13] and wavelet transform algorithms [31] helps to increase the reliability of failure categorisation techniques through the removal of noise and the enhancement of quality. There has also been an increasing interest in the image reconstruction of the in-pipe environment, where detection methods, such as process tomography [32,33], are becoming more mainstream. As of today, many innovative wireless sensor networks for pipeline systems emphasize the failure response time, efficiency of computation and the reliability of the communication systems used [14,34,35,36,37,38,39,40]. By having a combination of physical pipeline networks, sensing capabilities and computational elements, wireless sensor networks for the detection of defects in pipelines are essentially part of the family of cyber-physical systems.

2. Pipeline Failure Detection Methods