A Non-Intrusive Water Consumption Monitoring System

Water is an essential resource for humans as it is used in many activities for both leisure and hygiene. However, the technology available in monitoring water consumption is limited to the traditional flowmeter. Households and small buildings rely only on the end-of-month billing by the water distributor. This study presents a water monitoring system that uses a pressure sensor which is a non-intrusive method of determining water activity. Aside from calculating the volume of water consumed, the system implements fixture recognition using machine learning as its main feature. This provides more information to users allowing them to identify appliances or fixtures that consume a lot of water. Multiple test sites were used with varying pipe networks from building restrooms to houses to see its viability. Results show that the fixture recognition, using preprocessing techniques, improved in performance with accuracy 88 %, precision of 91 %, recall at 88 %, leading to an f1-score of 87 %.