PREDICTION OF PERFORMANCE AND EMISSION PARAMETERS OF AN SI ENGINE BY USING ARTIFICIAL NEURAL NETWORKS


Atik K., KAHRAMAN N. , Ceper B. A.

ISI BILIMI VE TEKNIGI DERGISI-JOURNAL OF THERMAL SCIENCE AND TECHNOLOGY, cilt.33, ss.57-64, 2013 (SCI İndekslerine Giren Dergi) identifier

  • Cilt numarası: 33 Konu: 2
  • Basım Tarihi: 2013
  • Dergi Adı: ISI BILIMI VE TEKNIGI DERGISI-JOURNAL OF THERMAL SCIENCE AND TECHNOLOGY
  • Sayfa Sayıları: ss.57-64

Özet

This study deals with artificial neural network (ANN) modeling of a spark ignition engine to predict the engine performances and exhaust emissions of the engine. The proposed ANN model was solved by a developed computer program which was written in the Visual Basic programming language. For training and testing of the proposed ANN, a four-cylinder, four-stroke test engine were used to be fuelled by methane hydrogen blended with various percentages of hydrogen (0, 10, 20, and 30%), at different excess air ratios (0.9, 1, 1.1, 1.2, 1.3 and 1.4) and operated at different engine speeds (1500, 2000, 2500 and 3000 rpm). An ANN model based on standard back-propagation algorithm for the engine was developed using some of the experimental data for training. The used ANN has three layer, three cells in the input layer (Speed, H-2 and Excess air ratio) and 8 cells in the output lager (HC, CO, CO2, and O-2 emissions, torque, specific fuel consumption, power and exhaust temperature). The performance of the ANN was validated by comparing the prediction dataset with the experimental results. In the hidden layer, 28, 29, 30, 31 and 32 cells were tested with artificial neural network structures. Results showed that the ANN provided the best accuracy in modeling of the emission indices with correlation coefficient equal to 0.9880, 0.9728, 0.9930 and 0.9623 for CO, CO2, O2 and HC and 0.8650, 0.9840, 0.9252 and 0.9605 for torque, brake power, specific fuel consumption and exhaust temperature, respectively. The overall results show that the networks can be used as an alternative way for predicting the performance and emission parameters of SI engine. The best result was obtained in the ANN with 28 hidden cells (R2 = 0.9860).