292th International Conference on Science, Technology, Engineering and Management(ICSTEM), Paris, Fransa, 25 - 26 Ağustos 2018, ss.52
Nowadays, artificial intelligence (AI) techniques are commonly used instead of time consuming and costly experimental
labor works. Especially, artificial neural networks (ANN) which are inspired from biological neural networks are prominent
among the other types of AI. In our study, ANN is used to estimate the characteristics of a four-stroke single cylinder
compression ignition engine such as performance and emission parameters. The engine was fueled with biodiesel-diesel
mixtures. The input parameters of ANN are blend ratio and engine speed while the output parameters to be estimated are
torque, power and CO, CO2, NOx values. The main architecture of ANN was consisted of three layers, namely; input,
hidden and output. Levenberg-Marquardt back-propagation training algorithm was used for weight and bias updates. MAPE
and R2 were used as performance evaluation criteria. Finally, it can be concluded that, ANN has considerable prediction
capability of the values with an acceptable accuracy and it can be suggested for the further estimations of engine parameters
instead of time consuming and costly experiments.