Monitoring and Forecasting Water Leakage in Aqueduct Networks Using Intelligent Sensors and AI/ML.

1. Introduction

Water leakage in aqueduct systems is a significant issue that leads to water loss, higher operational costs, and potential damage to infrastructure. Efficiently detecting and forecasting such leaks in large, connected grids of aqueducts is vital for maintaining the integrity and functionality of the water distribution system. A solution using a combination of intelligent sensing technology and AI/ML algorithms can be used to early detect the leakage of water. The sensor LNC (Leak Noise Correlator System) is used for real-time monitoring of water leaks. AI/ML techniques are then employed to detect patterns and forecast potential leaks, allowing for proactive maintenance. The GSM based SCADA/IoT system will be employed to collect the data recorded through sensors.

2. Monitoring Water Leakage Using Intelligent Sensors

The Leak Noise Correlator (LNC) system uses advanced acoustic sensors to detect leaks in underground pipelines by capturing sound and vibration data. The system consists of piezo-ceramic sensors that are placed at key points in the pipeline, such as valves or hydrants. These sensors detect noise frequencies caused by water escaping through leaks.

The collected data is transmitted wirelessly to a base unit and processed via a mobile app. The app uses signal processing algorithms to analyze the sound levels and frequencies recorded at different sensor locations to precisely pinpoint the leak.

Fig. 1: LNC Sensor

Fig. 2: Sensors deployment showing on GIS map

Fig. 3: Deployed Sensors

Fig. 4: Water leakage position

System Features

  • Frequency Range: The system filters noise between 20 Hz to 20,000 Hz, targeting the frequency range of typical water leaks.

  • Wireless Communication: Sensors communicate with the base unit and a private Wi-Fi network (802.11g)/GSM network, ensuring reliable data transmission.

  • GIS Integration: The LNC system integrates with GIS maps, allowing users to visualize the locations of leaks and related pipeline data (e.g., material, length).

  • Multiple Sensors: It supports up to 4 sensors, allowing simultaneous monitoring of multiple sections of the aqueduct grid.

  • Environment-Resistant: Sensors are IP68-rated (waterproof and submersible), while the base unit is IP67-rated (weatherproof), ensuring the system’s durability in harsh conditions.

  • A single full charge lasts up to 6 months.

3. Data Connectivity/Transmission:

The GSM and Wi-Fi modules in a Leak Noise Correlator (LNC) sensor, like the WaterPoint LNC GIS Integrated Kit, enable remote data transfer to a server by providing different methods of connectivity. Here’s how each can work to transfer data from the sensor to a server:

  • Wi-Fi Connectivity:

  1. The LNC sensor, equipped with a Wi-Fi module, can connect to a local Wi-Fi network.

  2. Once connected, it transmits leak data (e.g., sound recordings, GPS coordinates, or correlated leak locations) over the internet to a designated server.

  3. The server-side application could use secure protocols (like HTTPS) to receive data from the LNC sensor, where it’s then processed and stored for further analysis or display on a dashboard.

  4. Wi-Fi is ideal for urban areas or locations with reliable wireless network coverage, as it provides high-speed data transfer.

  • GSM (Cellular Network) Connectivity:

  1. The GSM module enables the LNC sensor to use cellular networks (2G, 3G, 4G, or even 5G) to send data to the server.

  2. When the sensor detects a leak, it encodes the data and sends it as a data packet over the cellular network to a preconfigured IP address or server endpoint.

  3. This method is advantageous in rural or remote areas where Wi-Fi coverage is limited, as GSM coverage is generally more widespread.

  4. Data transfer via GSM may also be secured using cellular encryption and VPN connections, ensuring data integrity and confidentiality.

4. Cost Estimation for a Small Aqueduct Network

The cost breakdown for implementing the LNC system in a small aqueduct network is shown below. This estimate applies to a single pipe with a maximum length of 1 km and no bends. If bends are present, the separation distance between installations will decrease.

The Total Cost: For a basic 2-sensor system with iPad and GIS integration, the total cost is now approximately 12,345 USD.

Note:

· Some cost-effective sensors are also available, and based on the analysis of severity and criticality, they can be installed to make the entire system more cost-effective.

· The installation cost will be @20-25% extra based on the location of site.

· For connectivity through GSM, device and data charges are additional, based on the number of sensors with GSM transceivers installed. Monthly data prices vary by country.

5. Utilizing AI/ML for Leak Detection and Forecasting

Data Processing for AI/ML Models

The LNC system generates large volumes of sensor data, including sound levels, frequencies, and vibration patterns. AI/ML models can be applied to this data to enhance leak detection and provide predictive insights. Key data processing steps include:

  • Preprocessing: Cleaning and filtering the data to remove irrelevant noise.

  • Feature Extraction: Identifying key parameters such as peak frequency, vibration duration, and sound amplitude to feed into the AI/ML models.

Leak Detection with AI/ML

  • Anomaly Detection: Machine learning models such as Isolation Forests or Autoencoders can identify abnormal patterns in the sensor data, suggesting the presence of a leak.

  • Supervised Learning: Models like Random Forests or Neural Networks can be trained to classify leak events using historical labeled data (e.g., leak vs. no leak).

Forecasting Future Leaks

AI/ML techniques can also be applied for time-series forecasting to predict future leaks based on historical trends:

  • Time Series Forecasting: Models like LSTM (Long Short-Term Memory) or ARIMA can be used to analyze historical leak data and forecast potential leaks in the future.

  • Predictive Maintenance: By predicting the likelihood of leaks, AI/ML systems allow for proactive maintenance, reducing downtime and repair costs.

6. Conclusion

Combining the LNC Leak Noise Correlator system with AI/ML algorithms offers a powerful solution for detecting and forecasting water leaks in aqueduct networks. The LNC system provides real-time, wireless monitoring, while AI/ML techniques enable proactive maintenance by identifying patterns and forecasting future leaks. This integrated approach significantly improves operational efficiency, reduces water loss, and lowers repair costs, making it a cost-effective and scalable solution for both small and large aqueduct grids.