Assessment and Prediction of Whole Body Vibration Exposure in Agricultural Tractors During Tillage Operations Using Supervised Machine Learning

Ride comfort during field operation is required for the operator’s health and safety. The extended exposure to whole body vibration (WBV) can adversely impact operator health, causing fatigue, spinal injuries, and long-term musculoskeletal disorders. This study focuses on analyzing and predicting tractor ride comfort using supervised machine learning techniques. Ride comfort was evaluated using metrics such as Overall Vibration Value (OVV), A(8), and Seat Effective Amplitude Transmissibility (SEAT, %). Experimental data were collected from three different tractor models operating under different tillage operations. Variables such as engine power, engine speed, working depth, implement type, and the operator’s body mass were considered to understand their influence on vibration levels. The results revealed that the OVV and A(8) values exceeded the exposure values of ISO 2631-1:1997. These higher levels indicate a risk to operator health during prolonged tractor use. Supervised machine learning models, including Least Absolute Shrinkage and Selection Operator (LASSO) Regression, Support Vector Regression (SVR), Extreme Gradient Boosting (XGBoost) Regression, and Artificial Neural Networks (ANN), were developed to predict OVV values using the identified input features. The effectiveness of each model was assessed using evaluation metrics like Mean Absolute Error (MAE) and the coefficient of determination (R²). The results showed progressive improvement in predictive accuracy, with R² scores of 0.62 for LASSO, 0.74 for SVM, 0.82 for XGBoost, and 0.92 for ANN. The ANN model outperformed with the higher R² value and lower Mean Square Error (MSE) value of 0.16.