Vibration-Based Intelligent Fault Diagnosis of Agricultural Tractor Transmission Systems: Methods, Signal Processing Techniques, and Machine Learning Approaches

Agricultural tractor transmission systems operate under variable field conditions in which speed changes, load fluctuation, gear engagement, clutch behavior, PTO demand and soil-implement interaction can alter vibration responses. This structured review examines vibration-based intelligent fault diagnosis of tractor transmission systems with emphasis on fault mechanisms, data acquisition, signal processing, machine learning, robustness, data scarcity and practical deployment. Evidence from tractor-specific gearbox studies, transmission design papers and transferable rotating-machinery research was synthesized to evaluate whether reported methods support field-ready maintenance decisions. Classical approaches, including RMS analysis, frequency-domain interpretation, wavelet decomposition, statistical features, random forest and support vector machines, are compared with recent deep and hybrid approaches such as CNN-BiLSTM, Time-GAN-Transformer, Vision Transformer-BiGRU, dictionary learning and transfer learning. The review shows that high classification accuracy under controlled bench-test conditions does not necessarily indicate reliable field diagnosis. Key limitations remain limited field-representative datasets, weak cross-speed and cross-load validation, scarce real fault samples, insufficient mechanical interpretability and poor integration with maintenance planning. Future systems should be duty-cycle-aware, mechanically interpretable, robust across operating domains and directly connected to predictive maintenance decisions.

کلیدواژه ها:

Agricultural tractor transmission ، Vibration-based fault diagnosis ، Gearbox fault detection ، Signal processing ، Machine learning ، Deep learning ، Domain adaptation ، Predictive maintenance