Optical detection of contamination event in water distribution system using online Bayesian method with UV–Vis spectrometry
| Autor: | Guangxin Zhang, Pingjie Huang, Hang Yin, Dibo Hou, Qiaojun Yu, Jie Yu, Ke Wang |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
0303 health sciences
Schedule Computer science business.industry Process Chemistry and Technology 010401 analytical chemistry Bayesian probability Supervised learning Message passing Pattern recognition Contamination 01 natural sciences 0104 chemical sciences Computer Science Applications Analytical Chemistry Set (abstract data type) 03 medical and health sciences ComputingMethodologies_PATTERNRECOGNITION Probability distribution Artificial intelligence business Spectroscopy Software 030304 developmental biology Event (probability theory) |
| Zdroj: | Chemometrics and Intelligent Laboratory Systems. 191:168-174 |
| ISSN: | 0169-7439 |
| Popis: | The detection of contamination events in water distribution systems remains a major concern to public health. However, much of the contaminant detection methods are supervised learning which cannot adapt to the complex environment in practical applications. In this work, a contaminant detection method using ultraviolet–visible spectroscopy technique was developed to achieve a goal of real-time detection on online acquisition of absorbance spectra. This method combined the advantages of probability distribution, message-passing algorithm and Bayesian theory. Message-passing algorithm was used to achieve a goal of online detection on noisy UV–Vis spectra signals. The proposed Bayesian algorithm organized the message passing schedule and helped to extract sequential patterns for event classification, which can avoid extreme conclusions. In addition, parameters were set up based on reasonable prior, by which the detection model was dynamically updated. Pilot scale experiment was conducted for long-term online monitoring of the water distribution system. And the experiment results showed improved performances in its ability to detect contamination events with higher probabilities, compared to previous studies. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect