Artificial neural network prediction based performance and exhaust emission study of variable compression ratio engine with Undi ethyl ester diesel blends: A fuzzy based optimization

To achieve the essential requirement of the day and provision of forthcoming energy storage with lowest concession in performances and exhaust emissions, suitable source should use for energy supply. Hence, in this work the effect of varying compression ratio on performance-emissions profiles using Undi ethyl ester is investigated and presented. The experimental study was conducted on single cylinder, direct injection variable compression ratio Diesel engine to analyses the performance-emission using various blends (10%, 20%, 30%, 40% and 50% biodiesel content on volume basis) of Undi ethyl ester by changing compression ratio (CR) from 16 to 20. The outcomes exhibit that brake thermal efficiency (BTE) and brake specific energy consumption (BSEC) increases with rise in compression ratio (CR) and amount of Undi ethyl ester in blend. Furthermore, for all Undi ethyl ester blends, rise in compression ratio (CR) reduces carbon monoxide (CO), unburned hydrocarbon emissions with the cost of higher NOx emissions. In perception of experimental results, artificial intelligence-based artificial neural network model has established for exact prediction and locating optimal diesel engine operating conditions using Undi ethyl ester blends. A single hidden layer with feed-forward backpropagation, Levenberg- Marquardt training algorithm (TRAINLM) along with logsig transfer function were used for differing neurons from two to twenty-five. For selecting optimum network topology fuzzy-logic optimization technique adopted. A (2-6-4) topology was found to the optimum model with an overall correlation coefficient (R) of 0.99704, mean square error (MSE) of 0.00044 and mean absolute percentage error 1.34%. The study explores the competency of fuzzy logic for deciding the optimal network topology of artificial neural network model under Undi ethyl ester blends.