טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentDaphna Yogev
SubjectDomestic Water End Users Classification as Potential
Method for Leakage Detection
DepartmentDepartment of Civil and Environmental Engineering
Supervisor Professor Fishbain Barak
Full Thesis textFull thesis text - English Version


Abstract

Water is a resource that affects every facet of life. It is one of the most important resources to human kind, but it is also a depleting necessity. Any water distribution system is bound to have some level of leakage. Home leakage may be of small scale, easy to miss and costly in the long run. Developing an automated system that monitors the consumption of end users (e.g., washing machine, dishwasher, sink or shower) can make these leaks noticeable and treatable. The aim of this research is to devise an algorithm that can classify water consumption according to the domestic end users. This classification algorithm is to be used for leak detection.

The data for this research consisted of flow rates recordings obtained from a single flow meter mounted on a domestic water system. The recordings were processed into separated events and later classified. This research investigated classification using different Machine Learning (ML) methods. The final algorithm devised, used Support Vector Machine (SVM) as a classification tool. This classification algorithm achieved a 92% success rate.

This research also proposes a leak detection method, utilizing the aforementioned classification.

Good classification results, along with the potential of leak detection method, make this research a good starting point for a sophisticated leak detection system that can be mounted on any domestic water system. This system would be easy to apply to other needs such as water theft detection or leakage in larger scale water systems. Further development of such a system will require additional measurements in more homes as well as considering using flow meters with greater accuracy and sensitivity.