To enhance the reliability of unmanned hydro-mechanical transmission tractors, a fault diagnosis method for their power-shift system was developed. First, fault types were identified, and sample data was collected via a test bench. Next, a feature extraction method for data dimensionality reduction and a deep learning network called W_SCBAM were introduced for fault diagnosis. Both W_SCBAM and conventional algorithms were trained 20 times, and their performance was compared. Further testing of W_SCBAM was conducted in various application scenarios. The results indicate that the feature extraction method reduces the sample length from 46 to 3. The fault diagnosis accuracy of W_SCBAM for the radial-inlet clutch system has an expectation of 98.5% and a variance of 1.6%, respectively, outperforming other algorithms. W_SCBAM also excels in diagnosing faults in the axial-inlet clutch system, achieving 97.6% accuracy even with environmental noise. Unlike traditional methods, this study integrates the update of a dimensionality reduction matrix into network parameter training, achieving high-precision classification with minimal input data and lightweight network structure, ensuring reliable data transmission and real-time fault diagnosis of unmanned tractors.

The power-shift system employed in agricultural multi-range continuously variable transmissions (CVTs) features a complex structure and control logic, presenting significant challenges to the reliability of agricultural machinery. To enable timely detection of faults, constructing an intelligent fault diagnosis classifier to monitor the system’s health status is essential. Typically, fault samples utilized for classifier development originate from ideal bench tests, characterized by uniform patterns and limited diversity, thereby hindering the algorithm’s generalization capability. This study addresses this issue by proposing a generative adversarial network (GAN) model, integrated with a triple loss function and a novel generator architecture, to augment the fault dataset under laboratory conditions. The generator architecture comprises a variational autoencoder module and an oil pressure point attention mechanism, enabling the generation of diverse and fluctuating virtual samples. Building on this augmented dataset, a fault classifier based on one-dimensional ConvNeXt was developed. Experimental results indicate that the classifier achieves an accuracy of 99.73%. While classifier accuracy decreases with increasing noise levels, the GAN-generated dataset provides more comprehensive training, resulting in an accuracy approximately 3% higher than that achieved using the original dataset.