Risk of gaseous release assessment based on artificial intelligence methods

Abstract Based only on current pollutant measured concentrations and atmospheric parameters the paper presents a novel procedure able to predict pollutant emission concentrations and to estimate the risk of pollution. Instead of deterministic or probabilistic methods, cumbersome regression analysis or physical models, a minimax decision procedure based on support vector machine in a minimax approach implemented in MATLAB object oriented language, was utilised. This procedure can perform highly complex mappings on nonlinearly related data, inferring subtle relationships between inputs and outputs. Numerical experiments were reported to gaseous emissions of pollutant sulphur dioxide from a thermo power station smokestack.