Water leak detection with data analytics

• Main Author: Cheong, Yun Cai
• Other Authors: Soong Boon Hee
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/149539
• Institution: Nanyang Technological University

In a water distribution network, there may be unintended water losses caused by wear and tear or incidents, resulting in pipe bursts or pipe leakages. Such cases are undesirable as it leads to financial losses due to water wastage, interruption to water services, waste of natural resources and even health concerns due to water contamination. Thus, it is crucial to rectify pipe bursts or pipe leakages promptly. While pipe bursts are typically more prominent, pipe leakages may not necessarily be as easily detectable. The purpose of this project is thus to develop an algorithm that can detect water leaks in small pipes. This is achieved by analyzing the data extracted from the water meter sensors and the use of machine learning to identify water leakages. This report presents the data analytics and machine learning conducted on the given data set from a foreign worker’s dormitory in Singapore. The data set consists of .csv files with water pressure and volume data collected daily at certain intervals over a period of 75 days. Data analysis on the data set were conducted via Microsoft Excel first before moving on to Jupyter Notebook for a more in-depth analysis. Python libraries such as Matplotlib, Numpy, and Scikit-learn were then utilized to map out various plots to illustrate the data set, giving a better understanding and useful insights with regards to the recorded water pressures. Data cleaning was then conducted before executing the machine learning portion of this project. Through unsupervised learning, the K-means algorithm utilized in this project clusters the data set and is thus able to detect water leakage based on certain assumptions. Further refinements for future developments were recommended in the last section of this report.