To address the problems of traditional fault diagnosis methods, such as the insensitivity of fault feature extraction, strong dependence on expert experience, insufficient generalization, and low fault recognition rate in gearbox running state detection, a gearbox fault diagnosis method based on the Markov transition field (MTF) and a Swin Transformer is proposed. First, the original one-dimensional vibration signal dataset is preprocessed. then the one-dimensional vibration signal is encoded into a two-dimensional feature map by MTF, which preserves the correlation between data and time. Second, the encoded 2D feature image dataset was input into the Swin Transformer model with a moving window and hierarchical design structure for training to realize the recognition of different faults in the gearbox. Finally, the power transmission fault diagnosis test-bed (DDS) dataset was used for experimental verification, and the fault diagnosis accuracy rate was 99.69%. Experimental results show that the proposed method has higher computational efficiency, better generalization performance, and higher fault identification accuracy than the intelligent diagnosis model based on a time-frequency diagram, CWT diagram, GAF diagram, convolutional neural network (CNN), and Vision Transformer (Vi T), which can provide a reference for fault diagnosis of gearboxes in practical industries.

When analyzing vibration or acoustic signals in machinery, noise interference within the characteristic signals can significantly distort the results. This issue is particularly pronounced in complex environments, where mechanical signals are often overwhelmed by noise, making it extremely difficult or even impossible to determine the operational status of mechanical equipment by the analysis of characteristic signals. Existing methods for analyzing weak signals in the presence of strong Gaussian noise have limitations in their effectiveness. This paper proposes an innovative approach that utilizes a Variable-Step Gaussian-Sinusoidal Filter (VSGF) combined with rotational coordinate transformation to extract weak signals from strong noise backgrounds. The proposed method improves noise reduction capabilities and frequency selectivity, showing significant improvements over traditional Gaussian filters. Experimental validation demonstrates that the signal detection accuracy of the proposed method is 10–15% higher than that of conventional Gaussian filters. This paper presents a detailed mathematical analysis, experimental validation, and comparisons with other methods to demonstrate the effectiveness of the proposed approach.

Detecting weak signals embedded in strong noise remains a critical challenge across numerous engineering fields, including mechanical fault diagnosis, unde