<![CDATA[The most complex aspect of predictive maintenance (PdM) for heavy vehicles is accurately forecasting fuel consumption as it is both critical and challenging to achieve optimal efficiency while minimizing expenses. Overfitting and failure to capture the existing data's linear relationships seem to remain the most persistent issues with traditional methods. In order to achieve this, the following techniques were analyzed to choose the best fuel consumption forecaster: Support Vector Regression (SVR), Decision Tree Regression (DTR), Random Forest Regression (RFT), and XGBoost. The models were implemented and their performance measured using Mean Squared Error (MSE). The analysis revealed that SVR surpassed the others with a linear kernel (C=10) achieving the lowest MSE rates of 0.26, while DTR, RFR, and XGBoost earned significantly higher 3.375, 2.857, and 3.857 (MSEs). The other models lagged behind SVR because SVR was more effective in capturing linear relations and managing overfitting, a dominating issue with decision-tree based models. This points out another important aspect of predictive maintenance (PdM) : the appropriate machine learning technique plays a very important role in accurately predicting fuel consumption of heavy trucks, which improves precision and fuel efficiency.]]>
Copyrights © 2025
