Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Israel → Algorithms for Early Contamination Detection in Drinking Water from Varying Water Sources

Technion - Israel Institute of Technology (2020)

Algorithms for Early Contamination Detection in Drinking Water from Varying Water Sources

Asheri Tehila

Titre : Algorithms for Early Contamination Detection in Drinking Water from Varying Water Sources

Auteur : Asheri Tehila

Etablissement de soutenance : Technion - Israel Institute of Technology

Grade : Doctor of Philosophy (PhD) 2020

Résumé
A safe and reliable supply of water is a main goal of any water supply company. Event Detection Systems (EDS), can rapidly detect contamination events, enabling a quick respond to mitigate water supply interruptions, and health hazards. An effective EDS should respond quickly to abnormality and obtain high sensitivity to diverse contaminations, and low probability for false alarms. It should also be reliable and affordable. Combining online water quality (WQ) sensors, already installed in the supply system, with interpretation algorithms, can provide a very effective means to meet these challenges.

WQ sensors measure physicochemical water parameters, such as, residual chlorine, electrical conductivity, pH, and UV (Ultra-Violet) absorbance. Detecting contamination events in these measurements is complicated, since they may vary significantly due to operational causes and water sources variability, and data might be imprecise or erroneous. Therefore, automated, computerized water monitoring EDSs’ inevitably produce false alerts, and there is a tradeoff between reducing invalid alerts and maintaining high sensitivity.

This research consists of development and validation of effective, reliable and applicable algorithms for early detection (ED) of contaminations in drinking water (DW) from one or more sources, using data from WQ sensors. This was done in two layers. First, reliable datasets of WQ measurements for background water from different sources, with and without several contaminants, are obtained. Then these data are used to develop and validate ED algorithms, as surrogates for direct measurements of contaminants.

Specifically, anomaly detection in UV-absorbance spectra as means for contamination detection is presented. The detection is based on an algorithm, which is based on a new affinity measure - the Fitness - which combines the Euclidian distance and the correlation matrix. If the sources of water in the system network are known, characterized, and there is no mixing between them, detecting water contamination is done by using the Fitness measure directly. When the water sources are known, characterized, and are mixed with unknown, varying proportions, the detection mechanism is based on the Fitness measure and a contamination signal amplification, Support Vector Machine classifier and a sequence analysis.

A second ED algorithm, the Kahal-Sahar algorithm, has also been developed, utilizing WQ measurements of standard physicochemical parameters. The core of the algorithm is a novel measure, autoconnection, which evaluates the similarity between two consecutive time windows, cut from a sensor’s time series of real data. The algorithm high performance, together with its simplicity, adjustability, ease of implementation and low computational complexity - make it a valuable addition to water monitoring systems.

Testing the performance of the two ED algorithms showed that processing physicochemical WQ measurements to detect anomalies, can serve as effective EDSs’ for DW contaminations. The algorithms are highly specific, without compromising their sensitivity, merits which comply with the basic criteria of reliability and applicability, demanded from an EDS. The methodologies behind the two algorithms can potentially be implemented in various applications which aims at detecting anomalies.

Présentation

Page publiée le 22 décembre 2022