CIS-WQMS

Background of the Project

Due to dynamic climate change and ecosystem disruption caused by human developmental activities, access to instant drinkable water is a major global crisis. Through real-time, intuitive, and innovative monitoring of water quality by leveraging the sophistication of disruptive technologies, this menace can be curtailed drastically. The project proposed a novel approach that blends artificial intelligence and edge computing capabilities to provide instant monitoring and prediction of a given sample of water based on given parameter thresholds as stipulated by the World Health Organization. 

System Component

The proposed design comprises the front-end and back-end, which form the software and hardware architectures. The hardware consists of actuators, sensors, and controllers that were connected remotely over wireless networks. The software consists of applications that are connected with artificial intelligence (AI) models for intelligent predictions. Eight ensemble learning models are considered for the front-end edge devices to meet the requirement of tiny machine learning (ML), while the back end has a self-supervised learning (SSL) model. The dataset comprises various features of the five parameters for determining water quality; conductivity, turbidity, oxygen, pH, and temperature, which are among the sensors included in the sensor module. Databases are used to store the data that the sensors have collected. A mobile AI-powered interactive app is developed to evaluate the water quality instantaneously based on sensor measurements. 

Project Outcome

The simulation results assert the DT model as the most suitable model for resource-efficient, cost-effective, reliable, and connected intelligence-based underlying prediction models in edge devices for real-time water quality monitoring and prediction with a precision of 99.36%, a sensitivity of 99.54%, an accuracy of 99.46%, rationality of 99.45% 0.0214 prediction error, and 0.0989 intra-reliability. 

Acknowledgment

The project was supported by the Priority Research Centers Program through NRF funded by MEST (2018R1A6A1A03024003) and the Grand Information Technology Research Center support program (IITP-2023-2020-0-01612) supervised by the IITP by MSIT, Korea.