Seyed Saeed Keykhosravi; Farhad Nejadkoorki; Mahmood Amintoosi
Abstract
Background and Objective: Dust modeling can be considered as an appropriate tool for predicting future industrial dust and identifying pollutant emission control strategies. Perceptron (MLP) and radial base (RBF) neural networks were used as a means for predicting the outflow dust from the main cogeneration ...
Read More
Background and Objective: Dust modeling can be considered as an appropriate tool for predicting future industrial dust and identifying pollutant emission control strategies. Perceptron (MLP) and radial base (RBF) neural networks were used as a means for predicting the outflow dust from the main cogeneration of Sabzevar cement factory located in Khorasan Razavi Province. Method: the concentration of dust from the main cement chimney in the study area was measured through field measurements. Then, the parameters of the production line (temperature, speed of gas output, voltage, fuel, raw materials, and time of sampling) were used as input data to the nerve networks to predict the concentration of dust. The values obtained from the implementation of the models were compared with the results of field measurements as a superior model selection. Results: The analysis of figures and statistical parameters showed that the mean squared errors for the two MLP and RBF models were as much as 1.787 and 21.263, respectively, and the correlation coefficients were as much as 0.99693 and 0.95811, respectively, which indicates a lower error and greater correlation between the MLP and RBF model in predicting the concentration of dust. Conclusion: Because of the high ability of perceptron nervous networks to predict dust concentration, this model can be a convenient and fast solution to predict the amount of dust in the industry.
Alireza Ehsanzadeh; Farhad Nejadkoorki; Sattar Khodadoostan
Abstract
Background & objectives: Air particulate matters which have natural and human made origins have significant effects on the climate, the environment and human health. Several epidemiological studies have shown a direct relationship among the concentrations of suspended particles with different ...
Read More
Background & objectives: Air particulate matters which have natural and human made origins have significant effects on the climate, the environment and human health. Several epidemiological studies have shown a direct relationship among the concentrations of suspended particles with different adverse health effects. The general purpose of this research was to determine the most important parameters affecting on the concentration of PM10 in Tajrish station (Tehran) and develop an estimator model for PM10. Materials & methods: In this study, a model is constructed using principal component regression (PCR) for the relationship between the hourly concentration of particulate matter smaller than 10 microns with meteorological parameters (WD, WS, T, P, H) and air pollution parameters (CO, NO2, SO2, NOx, NMHC, THC) in Tajrish station (Tehran). The results of the performance evaluation of PCR model were measured in training and testing stages using RMSE, MAE, R and IA as statistical indicators. Results: The results of principal components import into multiple regression model showed that the most important variable affecting on the concentrations of PM10, are air temperature and the wind speed. Also, CO and SO2 emissions were known as synergic factors for PM10 concentration. The results showed that PCR model is able to estimate 41% of PM10 concentrations in the testing. Conclusion: The principal components regression analysis showed that meteorological parameters are one of most important factors affecting on the reduction of PM10 concentration in Tajrish station (Tehran).
Alireza Ehsanzadeh; Farhad Nejadkoorki; Ali Taleb
Abstract
Background and objective: Air pollution in Tehran, because of high concentration of pollutants, has caused various diseases and many problems concerning the public health and welfare of citizens and also damages to the environment and living organisms. Materials & Methods: Air Quality Index ...
Read More
Background and objective: Air pollution in Tehran, because of high concentration of pollutants, has caused various diseases and many problems concerning the public health and welfare of citizens and also damages to the environment and living organisms. Materials & Methods: Air Quality Index (AQI) is a key tool to monitor the air quality, to realize the effects of air pollution on health and to choose methods against air pollution. This study aimed at modeling and estimation AQI by CART algorithm and adaptive boosting algorithm (AdaBoost). Hourly data on concentration of air pollutants and meteorological parameters related to Gholhak stations in Tehran was used for modeling and estimation of AQI. Results: The results showed that CART model had better performance than AdaBoost model. To evaluate these models, root mean square error (RMSE), mean absolute error (MAE), mean squared error (MSE) and correlation coefficient (R) of the CART model for the test, were respectively, 0.75, 0.101, 0.563, and 0.99 when compared to the AdaBoost model (RMSE=7.1, MAE=5.11, MSE=50.52 and R=0.95) which implies the absolute superiority of the CART model than the AdaBoost model. Conclusion:The results of this study showed that regression decision tree model can be used as an efficient model for modelling and estimation of urban air quality index.