An Artificial Intelligent Prediction Model For Evaluating The Engine Performance And Emission Characteristics Using Waste Cooking Oil Biodiesel Blends

Abstract

Bio-fuels or bio-diesels are biodegradable fuels and are not toxic in nature. In this study, bio-fuels are produced by transesterification process from waste oils of cooking, animal fat and vegetable oils. During the process, these oils are reacted with an alcohol generally ethanol or methanol in the presence of sodium hydroxide as catalyst resulting in ester called as bio-diesel with a byproduct of glycerin. The manufactured pure bio-diesel is generally less flammable when compared to diesel and having burning point of 50 degrees celsius. Bio-diesel blends are formed by combination of bio-diesel and petroleum diesel in different proportions and the flash and gel points lie between those of pure fuels depending on the mixture. Artificial neural networks (ANNs) are soft computational models that mimic the behaviour of human neurons. ANNs are formed by the interconnections between the building blocks called neurons which are simple processing units that process the data and the performance of network depends on parameters involved and the architecture used. These are used for obtaining the correlation between the dependent and independent process parameters that are nonlinear in nature. ANNs find application in classification and prediction problems, provides results that are fast and are very close approximation to actual output. This study indicates bio-diesel is produced from waste cooking oil methyl ester and various blends are made with different proportions of bio-diesel and petroleum diesel. The experiments are conducted on diesel engine with bio-diesel blends and the performance parameters are brake power (BP), specific fuel consumption (SFC), brake thermal efficiency and the engine exhaust emissions. The ANNs are trained with the partial blend data used for experimentation as input to the model and the performance parameters as output using different activation functions. The model is tested with the remaining data to assess the percentage error between the actual and predicted values. The results reveal that blends of bio-diesel provide good performance and emission characteristics. It is concluded from the study that the ANN model can predict the exhaust emissions and engine performance well.