Intelligent Classifiers on the Construction of Pollution Biosensors Based on Bivalves Behavior

Abstract

The aquatic environment is subject to a series of contaminants resulting from anthropogenic activity which may compromise biota health and the quality of water resources. There is an imperative need for cost-effective, accurate and online solutions for monitoring aquatic environments. The present work proposes the construction of aquatic pollution biosensors for detecting both petrochemical and anti-fouling paint compounds based on the behavioral analysis of Perna perna mussels through multiple classifier systems. Networks of mussels instrumented with Hall effect sensors and magnets were exposed to Water-Accommodated Fraction of diesel and micro-particles of anti-fouling paint. The hourly behavioral parameters average amplitude, transition frequency and amount of motion reversals were used to infer the contamination status (polluted or not) through voting classifiers. Results presented high accuracy (95.8%) in predicting diesel pollution and non-pollution while lack of data and intrinsic characteristics of anti-fouling paints provided less significant results for detecting its compounds. This paper has demonstrated a promising use of artificial intelligence in the construction of aquatic pollution biosensors using behavioral analysis of bivalves mollusks.