TY  - CONF
EP  - 148
VL  - 4
A1  - Pathmanathan, E.
A1  - Ibrahim, R.
A1  - Asirvadam, V.S.
UR  - https://www.scopus.com/inward/record.uri?eid=2-s2.0-79957600700&doi=10.1109%2fICCRD.2011.5763872&partnerID=40&md5=2058c31d8dc5034a8174e82158b0dda6
SN  - 9781612848372
Y1  - 2011///
ID  - scholars2074
TI  - Development of predictive emission models for various applications using ANN
SP  - 144
KW  - Back-propagation neural networks; Data division; Data preprocessing; Data sets; Emission model; Emission monitoring system; Emissions model; Feedforward backpropagation neural networks; Gasoline fuels; Generalized Regression Neural Network (GRNN); Generalized regression neural networks; Large combustion plants; Pollutant emission; Predictive monitoring; Radial basis; RBF kernels; Unburned hydrocarbons
KW  -  Backpropagation; Fuel tanks; Hydrocarbons; Monitoring; Neural networks; Nitrogen compounds; Pollution; Research
KW  -  Air pollution control equipment
N1  - cited By 0; Conference of 2011 3rd International Conference on Computer Research and Development, ICCRD 2011 ; Conference Date: 11 March 2011 Through 15 March 2011; Conference Code:84959
N2  - This paper aims to develop intelligent Predictive Monitoring Emission Systems (PEMS) for three distinct case studies involving traffic, gasoline fuel tanks and large combustion plants (LCP). The underlying theme of pollutant emissions exists in all three case studies whereby the gases that are monitored are NO2, unburned hydrocarbons, and SO2. These pollutants can cause grievous harm to health, environment and infrastructure hence they are vital to be monitored. Emissions models are required because this will allow countermeasures to be taken in order to control the attributes that contribute to the emission of these pollutants. The datasets are collected online via database libraries, and consequently data preprocessing and data division are done. Back-propagation neural networks (BPNN) are first used to model the emission, and then to compare, generalized regression neural networks (GRNN) are used. From the results it is shown that GRNN models outperform BPNN algorithms for complex and nonlinear datasets, because of the underlying radial basis kernel transfer function. The RBF kernel has fewer numerical difficulties; one of it is that the kernel output is contained between 0 and 1; hence the solution provided by GRNN is stable, certain and localized. © 2011 IEEE.
AV  - none
CY  - Shanghai
ER  -