ABSTRACT
The multi-sensor system based on IoT allowing a continuous and real time-evaluation of aerosol in the atmosphere. The project is based on the use of different commercially available semiconductor sensors that are integrated into a self-contained and portable device, so that it can operate on site. Different types of sensors are selected and their performances are tested on a bench simulating polluted atmospheres and developed for the occasion. In order to overcome the problems of non-repeatability and drift of the baseline and sensitivity, standardization pretreatment procedures are developed. Campaigns in measuring stations of one week and different types of sites are then conducted. There is a joint collection of sensor signals and the IoT system that allows management of all data. Methods based on neural networks are then applied in order to obtain jointly, from the sensor signals, a classification among three types of pollution as well as a global air quality indicator. These methods use an approach based on fuzzy logic in order to avoid edge effect problems.
