International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024
p-ISSN: 2395-0072
www.irjet.net
“COMPARISON OF DIFFERENT ARTIFICIAL NEURAL NETWORK MODELS FOR PREDICTING RESPIRABLE PARTICULATE MATTER (PM10) CONCENTRATION IN BENGALURU CITY” Chethan D M 1, Dr B Santhaveerana Goud 2 1 PG Student, Department of Civil Engineering, UVCE, Bengaluru University, Bengaluru, Karnataka.560056 Email: 2
Professor, Department of Civil Engineering, UVCE, Bengaluru University, Bengaluru, Karnataka.560056 ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Rapid increase in industrialization and
challenge, The term encompasses any physical, chemical, or biological agent that disrupts the natural composition of the atmosphere, thereby deteriorating its quality and posing health risks to inhabitants. Nowhere this issue is more pronounced than in metropolitan hubs like Bengaluru, where the Air Quality Index (AQI) fluctuates dramatically with each passing season, reflecting the alarming levels of air pollution prevalent in the region (Gurjar et al.2016). Bengaluru ranks 6th among the most polluted cities in India, with the Air quality Index 101 are found to be at alarming levels at some severely polluted areas.
urbanization is a threat to the public health because of adverse impact on the quality of air caused by the accumulation of unwanted particles. Studies conducted in Delhi have also documented the levels of different pollutants in the air have reached an alarming heights. From past few decades Bengaluru is also growing in an exponential way caused concern about the quality of air, it happens to be sixth most polluted city in India. Hence the present study focused on predicting PM10 concentrations at four different air quality monitoring stations of Bengaluru by the application of Artificial neural network models(ANN). An attempt also is being made to assess the efficiency of models in the predictions. Six years daily average PM10 data was used for the study, Four different ANN models namely Feed forward back propagation neural network(FFBP), ELMAN neural network, Recurrent neural network(RNN) and Nonlinear Autoregressive with Exogenous input(NARX) were used in predictions. The assessment of efficiency was based on the correlation coefficient(R) and Mean Squared Error (MSE). The results have shown that NARX model was found to be better than other models with a correlation coefficient of 0.88774 and Mean Squared Error of 0.008094. Hence for the city of Bengaluru NARX model may found to be more suitable for prediction of PM10 concentrations.
These places are dispersed throughout the city, represent focal points of heightened air pollution, where concentrations of harmful pollutants exceed permissible limits. The immediate impact of such pollutants on respiratory systems underscores the urgency of developing accurate prediction models to serve as early warning systems, safeguarding public health and well-being. While many prediction models attempt to correlate air pollutant concentrations with meteorological conditions. The complexity of these interactions necessitates advanced methodologies. Traditional deterministic models often fall short in predicting extreme pollutant concentrations and require extensive computational resources, thereby limiting their practical utility (Marjovi et.al 2016; Wang 2017). In contrast, statistical approaches, such as Multiple Linear Regression (MLR) and Auto Regressive Moving Average (ARMA) methods, struggle to capture non-linear patterns and may prove inadequate for extreme concentration scenarios (Li, X Peng et al. 2016). Recognizing these limitations, researchers have increasingly turned to artificial neural network (ANN) models, appreciating their efficiency and predictive accuracy. Artificial neural networks offer a versatile framework capable of effectively managing non linearities, data distortions, and missing values inherent in environmental datasets (Chaloulakou et al. 2003). Comparative studies between ANN models and conventional techniques consistently demonstrate the superior predictive capabilities of neural networks across various domains (Jiang 2004; Ghazi et al 2009. Gundogdu 2009).
Key Words: Particulate matter; Artificial neural network; feedfarward back propagation; Recurrent neural network; Nonlinear Autoregressive exogenous input.
1.INTRODUCTION In the midst of rapid increase in industrialization and urbanization, cities in developing countries are witnessing unprecedented population growth. Urban expansion, coupled with the surge in industrial activities, increased automobiles poses a grave threat to the natural environment, including vital resources such as air, water, and soil. The Continuous addition of harmful substances into the atmosphere, collectively known as atmosphere pollution, emerges as a pressing concern with far-reaching implications for human health, property, and ecological balance https://www.afro.who.int/health-topics/airpollution. Among the myriad forms of environmental pollution, air pollution stands out as a critical global
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 913