Comparative evaluation of machine learning models for predicting noise and vibration of a biodiesel-CNG fuelled diesel engine

Improving engine operation through the implementation of intelligent modelling is crucial for reducing vibration and noise. For this reason, In the present study, advanced machine learning models including Radial Basis Function Neural Network (RBFNN), General Regression Neural Network (GRNN), Support Vector Machine (SVM), and ensemble models with Least Squares Boosting (LSboost) are employed to predict noise and vibration of a diesel engine. The engine is fuelled with low-sulphur diesel, sunflower biodiesel-diesel blends at 20 % and 40 % by volume and compressed natural gas (CNG) added through the intake manifold at various flow rates. Noise and vibration data were gathered at intervals of 300 rpm between 1200 rpm and 2400 rpm. Results show that the among proposed models, for noise predictions, GRNN yield the best results among all models with R2 accuracy of 0.9983 and Theil U2 of 0.073. Meanwhile, in the lights of vibration results, RBFNN outperforms other models with R2 accuracy of 0.9968 and Theil U2 of 0.214.