Comparative Evaluation of Strain, Displacement, and Acceleration Signals for Vertical Load Sensing in Agricultural Tractor Tires

Intelligent tire technology enables real-time monitoring of tire–ground interaction states and provides an effective approach for improving the intelligence of wheeled tractors. However, the relative effectiveness of different sensing signals for vertical load sensing in agricultural tractor tires remains insufficiently understood. To address this issue, a feature selection framework combining finite element analysis, random forest, and Pearson correlation analysis was developed. A finite element model of a 280/85 R24 tractor radial tire was established in ABAQUS and validated through static loading experiments, with the maximum error in radial deformation remaining below 7%. Based on the validated model, tire rolling simulations were conducted under different vertical loads, inflation pressures, and slip ratios, and strain, radial displacement, and radial acceleration signals were extracted during tire–ground interaction. Seventeen time- and frequency-domain features were constructed for each sensing signal, and informative and non-redundant features were identified using random forest importance and Pearson correlation analysis. The resulting feature subsets contained 10, 7, and 6 features for strain, displacement, and acceleration signals, respectively. Within the final selected feature subsets, the highest correlations with vertical load were obtained for the strain T-mean feature (r = 0.97), the acceleration FA-max feature (r = 0.96), and the displacement FA-max feature (r = 0.72). These results demonstrate that strain signals contain the strongest load-related information, followed by acceleration signals, whereas displacement signals show relatively limited effectiveness. The findings provide guidance for sensor selection and the development of tire-based vertical load sensing systems for agricultural vehicles.