A FUZZY MODEL OF THE DISPERSED COMPOSITION OF SOLID PARTICLES FOR THE ASSESSMENT OF ATMOSPHERIC AIR DUST

Abstract

Models based on the rules of fuzzy logic, or fuzzy models (Fuzzy Models), have gained significant development recently due to the practical tasks of creating and operating various kinds of expert systems in those fields where it is impossible or extremely difficult to obtain the optimal result by analytical methods. This applies, first of all, to quality assessment systems. The input parameters of such systems are numerical data (a vector of real numbers), and the output is some qualitative assessment from an expert, which can be expressed in a point scale. The authors suggested applying fuzzy modeling to estimate the level of atmospheric dust. A feature of the model is the consideration of the dispersed composition of dust. The model was developed in the FUZZY LOGIC TOOLBOX module of the MATLAB package. The main channel for obtaining input data for the model was data from meteorological stations that continuously monitor the environmental condition in the city of Kyiv. Improving the quality of the rating scale by more accurately taking into account the dispersion of solid particles in atmospheric emissions is an urgent research task. The use of three channels for measuring the concentration of solid particles with different dispersions and vague modeling of the general state of air dustiness is the recipe that will allow solving the outlined problem.

The work considers a fuzzy model for estimating the level of dustiness of atmospheric air based on taking into account the dispersed composition of solid particles. The model is built according to Mamdani's algorithm. Three inputs and one output are used, each having three membership functions. The input data are the values of the concentration of PM1, PM2.5 and PM10 solid particles. The source of input data is the results of measurements by environmental monitoring stations available from the network in real time (ECOBOT resource). This allows not only the developer of the program, but also its user (expert) to promptly carry out approbation and, if necessary, correction of the model. The database of fuzzy rules is compact and includes only 27 formulas of fuzzy derivation. The model was developed in the FUZZY LOGIC TOOLBOX module of the MATLAB package. The separate consideration of the concentration of particles with submicron size in the overall dustiness coefficient is an important feature of the model and gives it an advantage over the standard assessment, as it increases the expert's attention to the pollution factor that is extremely harmful to human health.