Fuel, cilt.385, 2025 (SCI-Expanded)
This paper studies the prediction of vibration and noise levels in a four-stroke, four-cylinder diesel engine fueled with biodiesel and diesel, alongside hydrogen injection through the inlet manifold using three distinct artificial intelligence techniques: Radial Basis Function Neural Network, Adaptive Neuro-Fuzzy Inference System, and Least-Squares Boosting. The objective of this study is to forecast noise and vibration at varied engine speed, biodiesel ratio, and hydrogen flowrate. In the study, the optimal number of learners for Least-Squares Boosting was determined to be 88, while the best spread number for Radial Basis Function Neural Network was 100. In addition, Adaptive Neuro-Fuzzy Inference System is configured with three Gbell membership functions. The results indicate that Radial Basis Function Neural Network and Least-Squares Boosting outperform Adaptive Neuro-Fuzzy Inference System, with the best mean average percent errors and R2 values for the developed models being 0.9929, 0.9973, and 0.9233 for vibration acceleration and 0.9991, 0.9971, and 0.9901 for noise, respectively. Ultimately, it is concluded that the Radial Basis Function Neural Network and Least-Squares Boosting methods are effective choices for simulating and predicting the noise and vibration of biodiesel fueled, hydrogen aspirated diesel engine.