An ANN_Simulation Hybrid Approach To Forecast Polycyclic Aromatic Hydrocarbons Concentrations in Indoor Industrial Work Environment

oil, and coal. Sources of PAHs include emissions from industrial activities such as primary aluminum and coke production, petrochemical industries, rubber tire and cement manufacturing, bitumen and asphalt industries. Some studies outline a significant correlation between mortality by lung cancer in humans and exposure to PAHs. Monitoring activities to assess the human exposure to airborne PAHs in workplaces are expensive and time consuming since they require many samplers and analytical methods derived from analytical chemistry. The aim of this study is to develop a tool that, through the prediction of the PAHs concentrations in work environment, suggests a sampling strategy able to improving the reliability of measurements and reduce costs of environmental monitoring. Different workplaces, using a multi zone modeling simulation software, have been analyzed; for each case the concentration of the pollutants emitted in the indoor environment have been detected. The output data of the simulator have been adopted to train an Artificial Neural Network (ANN). The trained ANN is able to provide, through computational logic, a reliable forecast about the concentrations of different species of pollutants statistically distributed in the environment based both on characteristics of the workplace (as room dimensions, surfaces of aeration, etc.) that on type of contaminant source and on intensity of emission. This allows to determine the minimum number and the correct location of the samplers to perform the environmental monitoring. The jointly use of the simulation software and the ANN allowed to develop a tool characterized by advantages of both the technologies. In fact, the simulation software allows to test a wide variety of cases and, for each of them, provides reliable measures of the pollutants concentration at different points. The ANN, instead, is trained by means of aggregates values of the characteristic variables obtained from simulation, thus providing the number and the position of the samplers to be adopted as well as the typology of pollutants (belonging PAHs groups) to be measured. The main advantages of the toll are that it requires a limited number of input data, provide outputs in a very limited computational time, and, above all, it allows to reduce cost of sampler activity by reducing the number of sampling points.