Water pollution recognition with the electronic nose KAMINA

Abstract

The Karlsruhe Micronose KAMINA was tested in some model applications concerning water pollution screening. The headspace of water samples polluted with chloroform (as a representative for chloro-organic solvents) or ammonia (representing faecal contamination) was investigated simulating stagnant and flowing waters. Discrimination and recognition of tap water, ammonia contaminated water and chloroform contaminated water were successful. From 12 repeated headspace measurements, a reliable LDA-model was obtained (cross-validation with leave-one-out method). A flexible data processing method was developed to adapt the LDA-model to changing background water odors for on-site field tests of water quality. Including only two additional field measurements of flowing water into the laboratory data evaluation model was enough to correctly classify contaminants in flowing waters. In both cases response times of 1–3 s and recovery times below 1 min were found ( t 90). Detection limits were about 1 mppm (ppm by mass of the contaminant in water). Signal processing was developed which considers a floating background of water odor.