As well-known, deep learning models have achieved great success in the field of intelligent fault diagnosis. However, once the working condition changed, the diagnostic accuracy of the trained models would be greatly affected, which seriously limits the application of deep learning models in real industry. Therefore, signal-transformer (S-Transformer), an intelligent fault diagnosis model focusing on the problem of variable operating conditions, is proposed in this study. First, signal embedding is employed to complete the segmentation and up-dimensional representation of the 1-D signal, thus enriching the information in a high dimensional space. Then, those embedded subsignals are further processed via the multihead self-attention mechanism to explore the state features of the signal in different spaces for a deep representation. Finally, an attention visualization method is proposed for vibrating signals to increase the interpretability of the proposed model and overcome the drawbacks of black boxes. With an experimental validation, the proposed model outperforms than other six deep learning methods in terms of diagnostic accuracy under unknown operating conditions. According to the principle of the model, the crucial weights for signal are further explained in visualization. Furthermore, it can be foreseen that the S-Transformer can achieve a robust recognition effect with the ability of interpretable feature extraction in signal denoising.

Feature extraction from a time sequence signal without manual information is an important part for bearing intelligent diagnosis. With the merits of signal information and feature structure information excavation, Deep ConvNet is widely used in bearing fault diagnosis and analysis under complex working conditions. However, due to the complexity of the bearing operating environment in the actual operation process, the sensitive features show different scale distribution characteristics. Meanwhile, it is known that the convolution kernel of ConvNet is usually small, which mainly focuses on the small-scale details of state distribution characteristics while ignores the identification of the overall trend of characteristic distribution. Considering that the size of convolution kernel can sense information hidden in different scales, this paper designed a one-dimensional vision ConvNet (VCN), where the architecture is composed of multilayer small kernel network and single-layer large kernel network side by side. The multi-kernel structure improves the ability of network to detect fault characteristic frequency band. By analyzing the artificially generated data and experimental data, the setting method of large convolution kernel and stride is discussed. Compared with the traditional CNN, wide first-layer kernels (WDCNN) and multiscale kernel-based ResCNN (MK-ResCNN), this network improves the recognition accuracy with a better stable training process for rolling bearing fault classification.

The transient features caused by local fault are of vital importance for bearing fault diagnosis in intelligent industry. Due to the uncertainty of fault forms and non-stationarity of operating conditions, fault feature distribution influenced by physical dynamic response of actual defect is always complex and irregular. Motivated by this, a convolution sparse self-learning (CSSL) is proposed to accomplish an adaptive feature enhancement. Firstly, the desired morphological structures is promoted by a two-dimensional manifold sensing. From a randomly selected fragment, time-frequency manifold learning is then applied to mine the latent structures. Therewith, this kernel is used to operate a shift-invariant sparse analysis on raw time-frequency image. Combined this rebuilt image with raw phase, enhanced signal is synthesized. Finally, the desired transient morphology can be automatically mined. Practical defective bearing data are analyzed to illustrate its effectiveness. A comparison further illustrates that the proposed CSSL is superior in morphological transient features enhancement.

Deep learning, with its ability of feature mining and logical judgement, has been widely studied in industrial intelligent diagnosis, including bearing fault diagnosis. However, an explicable and representable expression of deep learning architecture for the variational working conditions has been rarely discussed while it is known that vibration features from bearings are seriously influenced by variational working conditions. In this article, a deep interpolation ConvNet (DICN) architecture with three special layers, consisting of multiple sub-ConvNet units, weight unit, and fusion unit, is presented with the basic deep ConvNet architecture. Different from the traditional network, the first sub-ConvNet extracts the fault features under different working conditions, while the corresponding condition weight unit is learned from a working condition identification task. With the principle of interpolation theory, fusion unit is employed to achieve a sound fault feature representation under unknown working condition, which is named as ConditionSenseNet (CSN). This CSN architecture provides a way to dynamically express the crucial features hidden in the samples with the influence of working conditions suppressed, especially the variational working factors will be interpolated in this nonlinear fitting model. Additionally, three experimental studies are tested to verify the effectiveness of the proposed DICN method for bearing intelligent diagnosis under variational working conditions. The results and comparisons with other seven deep learning models show the proposed method shows outstanding robustness and higher accuracy where the accuracy of DICN is higher than the one of convolution neural network by more than 9% even if the working condition is variational.