Evaluation of Machine Learning Models in Predicting Diesel Engine Behaviour and Emissions with Biodiesel Blends

Abstract

The fuel qualities and compatibility of biodiesel with diesel fuel derived from petroleum have been drawing more and more attention to it. As a result, testing the efficiency and pollution levels of diesel engines powered by biodiesel blends and conventional diesel fuel derived from petroleum is essential. The objective of this research is to use a variety of machine learning (ML) methods to analyse the efficiency and pollution levels of diesel engines that run on biodiesel mixes. An example of one such method is supporting vector regression, which falls under artificial neural networks (ANNs), while another is adaptive neuro-fuzzy inference systems (ANFIS). Utilizing biodiesel blends B0, B5 and B10 in a Toyota 2GD-FTV engine is the focus of the case study. The model takes the engine's process parameters as inputs and produces NOX emissions and predicted torque as outputs. We construct and evaluate numerous prediction models that rely on ML. Using RMSE (Root Mean Squared Error), MAPE and R (Mean Absolute Error and Correlation Coefficient), we evaluate and contrast the models' efficacy. Based on the outcomes, Support Vector Regression (SVR) seems like a good fit for the model that will be employed to forecast behaviour and emissions. In addition, study elucidates the impact of several engine process parameters on emissions and performance.