Analytical Formulation for Diesel Engine Fueled with Fusel Oil/Diesel Blends

AKÇAY, Mehmet ; ÖZER, Salih ; SATILMIŞ, Gökhan

Abstract

The experiments related to reduction of gases from the exhaust emissions of internal combustion engines, usually conducted in laboratory conditions, are quite laborious and costly. For these purposes, modelling engine experiments with algorithms have emerged as a way forward. In this paper, the operation of diesel engine is modelled through experimental dataset, which has input variables such as engine load, fuel type and output variables such as carbon monoxide (CO), carbon dioxide (CO2), oxides of nitrogen (NOx), hydrocarbon (HC), smoke, Brake Specific Energy Consumption (BSEC) and maximum in-cylinder pressure (Cpmax). Artificial intelligence based Symbolic Regression (SR) algorithms have been used to derive analytical equations of each output variable. The derived equations and experimental results are plotted on the same graph to show the accuracy of the obtained equations. The coefficient of determination (R2) is between 0.98 and 0.99 in all equations. In addition, Mean Error Percentage (MEP) value is less than 10 in all equations. The performance of SR algorithms is compared with Artificial Neural Network (ANN), Support Vector Machines (SVM), instance-based and K nearest based classifier (IBk), ensemble method-based bagging algorithm, and decision tree-based REPTree algorithms. SR algorithms exhibit the best performance for all output variables. IBk algorithm exhibits the second-best performance for the BSEC, CO, CO2, HC and NOx output variable. SVM algorithm exhibits the second-best performance for the Cpmax output variable and Bagging algorithms exhibits the second-best performance for the smoke output variable. The operation of diesel engine can be predicted using these equations and algorithms for further research.


Keyword(s)

Artificial intelligence, Diesel engine, Engine performance, Symbolic regression


Full Text: PDF (downloaded 41 times)

Refbacks

  • There are currently no refbacks.
This abstract viewed 128 times