The prediction of diesel engine NOx emissions using artificial neural network / Mohd. Mahadzir Mohammud and Khairil Faizi Mustafa
This paper describes an experimental and computer simulation studies used to develop a suitable algorithm to predict and control the oxides of nitrogen (NO ) emitted from the Yanmar L60AE-D single cylinder direct injection diesel engine, fitted in a Cusson's Engine Test Bed Model P8160. NOx con...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universiti Teknologi MARA, Pulau Pinang
2003
|
| Subjects: | |
| Online Access: | http://ir.uitm.edu.my/id/eprint/11377/ http://ir.uitm.edu.my/id/eprint/11377/ http://ir.uitm.edu.my/id/eprint/11377/2/AJ_MOHD.%20MAHADZIR%20MOHAMMUD%20ESTEEM%2003.pdf |
| Summary: | This paper describes an experimental and computer simulation studies used to develop a suitable algorithm to predict and control the oxides of nitrogen (NO ) emitted from the Yanmar L60AE-D single cylinder direct injection diesel engine, fitted in a Cusson's Engine Test Bed Model P8160. NOx contained in the exhaust gases of diesel engines have been identified as elements responsible for polluting our atmosphere. In order to reduce or to control diesel engine polluting emissions, the formation mechanism of NOx can be predicted. A neural network model is developed to obtain the NOx emission concentration under various operating condition. The neural network, well suited for
non-linear phenomena modelization, is able to deal with high uncertainly input level and able to operate outside of their range of training experience. A feedforward neural network structure has been selected with a backpropagation training procedure. Four operating parameters (engine speed,
engine load, exhaust temperature and air fuel ratio) have been used as an input data in the modelling process. The modelling algorithm implemented, takes a large set of measurements to learn how to predict the NOx emission from four operating parameters. The predicted values obtained using neural network model are compared with the experimental values. The studies show that the predicted results are in good agreement with experimental values, within less 9 % relative error. |
|---|