Design and implementation of an autonomous EGR cooling system using deep neural network prediction to reduce NOx emissio
- PDF / 3,047,912 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 69 Downloads / 150 Views
(0123456789().,-volV)(0123456789(). ,- volV)
ORIGINAL ARTICLE
Design and implementation of an autonomous EGR cooling system using deep neural network prediction to reduce NOx emission and fuel consumption of diesel engine Halil ˙Ibrahim Akolas¸1 • Alirıza Kaleli2
•
Kadir Bakırcı3
Received: 26 November 2019 / Accepted: 4 June 2020 Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract This study includes the design of an autonomous exhaust gas recirculation (EGR) cooling system and implementation of the system on diesel engine by using deep neural network (DNN)-based controller. The NOx formation and BSFC were optimized as output variables considering engine speed, load, EGR ratio and the exhaust gas temperature entering the intake manifold as input variables using deep and traditional NN modelling method. The DNN modelling method has stronger prediction capability than traditional NN and can deal with the modelling problem in the complex ICE data. The developed NN models were compared with the relative error analysis to verify the effectiveness of the deep NN modelling method. The activation function, the number of neurons, learning rate and backpropagation optimization method were considered as algorithm performance parameters for the design of optimum network by using the searching method. R2, MSE, MAE and RAE are used to evaluate the output performance of the deep and traditional NN model. Then, the autonomous and independent EGR cooling system is designed apart from the main cooling system of the engine with electric pump and fan components, unlike conventional systems. PID controllers are designed for these components, and the parameters of the controller are determined by pattern search optimization method. In this way, the exhaust gas temperature entering the intake manifold and EGR ratio are set to the desired value according to the BSFC and NOx output parameters under different operating conditions of the engine. Experimental results showed that the designed DNN controller-based autonomous EGR cooling system outperforms compared to the conventional system in terms of reducing the NOx and BSFC, with a 5.751% and 2.997% reduction, respectively. Keywords Autonomous EGR cooling system NOx BSFC Deep neural networks Backpropagation Optimal tuned PID controller Abbreviations ANN Artificial neural network BMEP Brake mean effective pressure
& Alirıza Kaleli [email protected] Halil ˙Ibrahim Akolas¸ [email protected] Kadir Bakırcı [email protected] 1
Vocational School of Ahlat, Bitlis Eren University, 13400 Bitlis, Turkey
2
Department of Electrical-Electronics Engineering, Faculty of Engineering, Samsun University, 55080 Samsun, Turkey
3
Department of Mechanical Engineering, Faculty of Engineering, Ataturk University, 25240 Erzurum, Turkey
BSFC DNN EGR ICE MSE MAE NOx NN PID RAE R2 SGD
Brake-specific fuel consumption Deep neural network Exhaust gas recirculation Internal combustion engine Mean squared error Mean absolute error Oxides of nitrogen Neural network Pr
Data Loading...
