Any method that can proactively act to moderate factors such as pressure and changes in the flow rate can make the system safe in real time, and less prone to pipeline failures. Until recently, electromechanical controls have been utilized. The advent of new technologies like the Internet of Things (IoT), machine learning (ML), and deep learning (DL) have revolutionized the control system for water supply to a great extent.
The IoT is a system of interrelated sensing devices with unique identifiers so that each system can be directly connected to a network and be able to transfer data over the network without requiring human-to-human or human-to-computer interaction. IoT devices share the data they collect through an IoT gateway or other edge device where the data is either sent to cloud servers before analyzing or is analyzed locally.
Sometimes these devices interact with other related devices and act on the information they get from one another. These devices do most of the work without human intervention, although people can interact with the devices; for instance, to set them up, give them instructions, or access the data and even link to remote stations such as hydropower plants, cross-country oil piping systems, and nuclear power plants. These systems can be effectively used for managing the system in real-time. A few examples are provided here.
The IoT and ML/DL for efficient control of the water supply system over undulating terrain
A network of pipes for water supply is spread over an undulating terrain with many valves. An undulating terrain makes the actual pressure inside the pipeline vary drastically because of elevation changes and varying demands from different portions. If proper balancing has not been carried out by controlling the valves in real time, connections in certain high-elevation areas will be deprived of supply due to inadequate pressure. If the system is equipped with IoT-enabled pressure sensors, this data can be received in real-time by a processing center.
Partial closing of valves that supply water to low-lying areas can bring back the pressure in high-elevation areas and avoid such interruption. The partial shutting of the valve as to avoid any disruptions or pipeline failures can be carried out by employing either a system using a rule-based approach or a system based on ML/DL. Machine learning and deep learning systems try to achieve the capability of the human brain in making decisions, of course in a limited manner, by using the concept of learning from examples.
The IoT and ML/DL for efficient control of the water hammer in the pipeline
The IoT and ML/DL can also be used for controlling the water hammer in a piping system. The data regarding pressure fluctuations at salient points corresponding to the change in flow rate can be detected by IoT sensors in real time and can be transferred to the processing center. using the concept of learning from examples.
The processing center can take corrective measures in real-time. For example, at the onset of pressure rise, the bypass valves can be operated before the surge reaches critical locations. Also, the system can send instructions to the surge protection devices so they are actuated to control the surge. As previously mentioned, the control system can be either rule-based or ML/DL-based.
Thus, a sustainable trouble-free hydraulic system can be established without human intervention. Importantly, the transient analysis should be conducted in the same system initially to generate the data that will be used to train the ML/DL-based control system for better response. Many commercial software platforms are available for conducting transient analysis in piping systems.
The IoT and DL for detecting leaks in a supply system
Leaks in a pipeline system can change the pressure at salient points and can lead to pipeline failures if not tackled. The change in the pattern of pressure at these points can be utilized for detecting the pressure variation. IoT can record the pressure and transfer the data in real-time to a data processing center. The advent of ML/DL has opened an avenue for detecting pattern changes, and this technique can be used for identifying the presence and location of leaks from the change in the pattern of pressure at salient points. Figure 15 depicts IoT and ML/DL-based architecture.
Belsito et al. (1998) and Barradass et al. (2009) detected the location and size of the leaks in a pipeline by using an artificial neural network (ANN) (a deep learning structure). Bohorquez et al. (2020) presented an innovative transient-based technique that used ANN to identify topological elements such as junctions in water pipeline networks and the characteristics of leaks. In this technique, the pressure head data of consequent transient events are required for the training and testing of the ANN and are obtained from numerical models of transient flow.
