Evaluation of Emission Pattern of Compression Ignition Engines Fuelled With Blends of Orange Peel Oil Based Biodiesel Using Artificial Neural Network Model

Abstract

It is well established that Orange peeled oil biodiesel (OPOB) is a suitable fuel in Compression Ignition (CI) engines because of its compatible physicochemical properties with diesel. Literature is however sparse on its emission characteristics in CI engines majorly because few investigations have been done due to seemingly difficult and time-consuming experiments. On this strength, this work first carried out experimental investigations on carbon monoxide (CO), unburned hydrocarbon (UHC), oxides of nitrogen (NOX) and smoke of orange peel oil based biodiesel in single cylinder, four stroke CI engine; and afterwards applied the power of artificial neural networks (ANNs) prediction model to predict to the full-scale CO, UHC, NOX, and SMOKE values of the CI engine. Brake load, orange oil-diesel mixture percentages and engine speed were the inputs of the ANN while Levenberg Marquardt (trainlm) and scaled conjugate gradient (trainscg) were the training algorithms to enable better comparative estimation. Performance indicators (correlation coefficient (R), mean absolute percentage error (MAPE) and root mean squared error (RMSE)) were used to compare the predicted and experimental values of the neural network. The ANN trained with Levenberg Marquardt algorithm predicted the emissions with highest accuracy (R very close to +1) and low error rates. The study concludes that ANN is a reliable tool to investigate the emissions profile of a CI engine.