Abstract
The paper describes a simple and lowβcost multisensing instrument based on neural network to monitor multiple environmentally significant metals in solution by implementing the potentiometric stripping analysis technique. Our analyzer consists of a measuring unit, implemented on a lapβtop/notebook computer equipped with a simple electronic circuitry, and a reusable sensor, based on glassyβcarbon technology. The current system has been characterized with synthetic samples containing different concentrations of lead, copper, cadmium and zinc, showing detection limits suitable for the first monitoring of metal traces. A linear relationship between peak area and concentration has been observed for all tested metals. Therefore, a simple linear regression can be used by the analyzer to determine the actual metal concentration from the PSA peak area. The presence of a second metal or of a real matrix, often, interferes with stripping measurements in different ways: compounds with oxidation potentials close to that of the analytes of interest may interfere if the instrument does not have sufficient resolution to resolve the overlapping PSA peaks; species which form intermetallic compounds with the analytes may result in erroneously low analyte concentration readings, since the oxidation potential of the intermetallic compound is rarely near that of the original analyte. In order to overcome these problems, the described analyzer has been integrated with a Neural Network algorithm, able to separate the different contributions, after a proper training.