AN APPLICATION OF ARTIFICIAL NEURAL NETWORK TO EXHAUST EMISSION MODELLING FROM DIESEL ENGINE

Krzysztof Brzozowski, Jacek Nowakowski

Akademia Techniczno-Humanistyczna
Willowa 2
43-309 Bielsko-Biała
tel./fax: +48 (33) 8279289
e-mail: kbrzozowski@ath.bielsko.pl
e-mail: jnowakow@ath.bielsko.pl

Abstract

The paper presents an application of artificial neural network to modelling emission from Diesel engine. In our study six independent engine control variables have been identified and used. In order to found the relations between engine control parameters and emission the radial artificial neural network (ANN) has been applied.
The ANN has been trained on the base of measurement results whose cover a wide range of steady state engine operating conditions. The accuracy of ANN output has been compared with calculation based on power function proposed in earlier works. In the paper has been shown that ANN gives better emission approximation than power function. A good correspondence of ANN prediction values has been achieved also in comparison to results of additional experiment data.