Transformer model is being gradually studied and applied in bearing fault diagnosis tasks, which can overcome the feature extraction defects caused by long-term dependencies in convolution neural network (CNN) and recurrent neural network (RNN). To optimize the structure of existing transformer-like methods and improve the diagnostic accuracy, we proposed a novel method based on the multiscale time-frequency sparse transformer (MTFST) in this paper. First, a novel tokenizer based on shot-time Fourier transform (STFT) is designed, which processes the 1D format raw signals into 2D format discrete time-frequency sequences in the embedding space. Second, a sparse self-attention mechanism is designed to eliminate the feature mapping defect in naive self-attention mechanism. Then, the novel encoder-decoder structure is presented, the multiple encoders are employed to extract the hidden feature of different time-frequency sequences obtained by STFT with different window widths, and the decoder is used to remap the deep information and connect to the classifier for discriminating fault types. The proposed method is tested in the XJTU-SY bearing dataset and self-made experiment rig dataset, and the following work is conducted. The influences of hyperparameters on diagnosis accuracy and number of parameters are analysed in detail. The weights of the attention mechanism (AM) are visualized and analysed to study the interpretability, which explains the partly working pattern of the network. In the comparison test with other existing CNN, RNN, and transformer models, the diagnosis accuracy of different methods is statistically analysed, feature vectors are presented via the t-distributed stochastic neighbor embedding (t-SNE) method, and the proposed MTFST obtains the best accuracy and feature distribution form. The results demonstrate the effectiveness and superiority of the proposed method in bearing fault diagnosis.

Establishing effective feature descriptors is a crucial step in 3D point cloud registration task. Existing manual‐based methods are noise‐susceptible and time‐consuming when running on low‐cost edge computing devices. To this end, the authors proposed a Local Reference Frame (LRF) based approach that can quickly and robustly register point clouds by using a novel lightweight local‐spherical grid weighted descriptor (LSGWD). Firstly, the LRF of the proposed algorithm is established by the covariance matrix eigenvector of KeyPoint's spherical support set and the centroid vector's projection on its orthogonal plane. Then the spherical support is grided to 32 bins, and the 4D geometric features of each subset are constructed by the centroid moment and the cosine value of the angles between the centroid vector and axes of LRF. Secondly, to restrain the insufficient discriminative information presented in the purely geometric features, the Gaussian projection and gradient mapping are proposed to calculate the smooth density and the correlation of structural characteristics, which are obtained as the distribution information of each bin to weigh the feature representation. Finally, the 32 × 4‐dimensional KeyPoint descriptor is obtained and used in the 3D point cloud registration framework. Experiments are carried out on three test datasets and real scene data. Compared to previous baselines, our descriptor achieves the state‐of‐the‐art performance in terms of efficiency and accuracy owing to its compact structure and noise robustness. The proposed method enhances the recognition and registration performance of 3D point cloud matching in low‐cost edge computing applications.