Optimization of performance and emission characteristics of a diesel engine in dual-fuel mode with LPG using adaptive-neuro fuzzy inference system model

Abstract

In this work, a dual-fuel engine was tested. The engine was converted to run on a dual fuel system, using diesel as the ignition source and liquefied petroleum gas (LPG) at varying flow rates depending on the load. The engine was modified so that LPG could be used as fuel and diesel could be used as ignition source. The engine parameters were optimized with the help of Adaptive Neuro Fuzzy Inference Systems (ANFIS). Diesel engine performance and emission characteristics when running on LPG as a secondary fuel have been modelled using ANFIS. ANFIS combines the self-learning of ANN with the reasoning of Fuzzy Inference System (FIS). The five layers of ANFIS model contains fuzzification, product, normalized, defuzzification, and output. ANFIS predicts GRG using injection pressure, LPG flow rate, and braking power. Grey relational analysis was used to converts multi-objective to single-objective optimization with Grey Relational Coefficient, grey relational grade (GRG) and Rank. ANFIS prediction modelling uses gradient descent and least squares to train parameters. Functional signals are used till the defuzzification layer. To reduce inaccuracy, outcome parameters are regulated using least squares. The backward pass uses gradient descent to improve parameters. The results were shown using Surface plots representing the effect of input parameters on the GRG value. According to the results of the performance evaluation, the ANFIS projected data was consistent with the experimental data, with an overall correlation coefficient of 0.99415. The optimum operating conditions are found to be 194.32 bar injection pressure, 1 LPM LPG flowrate and 1.13 BP with 0.835084 Optimum GRG.