A New Method of Tractor Engine State Identification Based on Vibration Characteristics

Abstract

Based on signal decomposition, a tractor engine state recognition method is proposed to explore the degree of information recognition of vibration signals at measurement points at different distances from the engine and the degree of correlation in different directions. The accuracy of engine operating state information recognition was obtained by analyzing the vibration signals of the tractor at different measurement points. The main contents are as follows: Based on variational mode decomposition (VMD), the modal component, which includes the state information, was obtained by measuring the vibration signal of the tractor at each measurement point under different driving conditions, and the exogenous excitation of the tractor under different road conditions was simulated by changing the tire pressure. Then, the state characteristics of the modal component were quantified based on permutation entropy (PE), and the correlation coefficient was used as the evaluation index to select the entropy of the optimal modal component as the feature vector. Finally, a support vector machine and random forest classification models were trained with 4800 feature vectors under 25 working conditions, and the remaining 900 feature vectors were used to verify the classification results. Compared with the results of empirical mode decomposition (EMD), the superiority of this method was proved. A comparative study with backpropagation demonstrated the superiority of the support vector machine and random forest identification method using a small sample size. The results indicate the following: (1) the accuracy of engine condition recognition, which was measured by longitudinal vibration signals, was better than that of vertical vibration signals at different measurement points; and (2) the closer the vibration transmission distance between the measurement point and the engine, the higher the recognition accuracy of the measured signals. This study provides a reference for the condition identification of agricultural machinery in complex working environments and lays a foundation for the fault diagnosis of agricultural machinery under working conditions.