Large and expensive mechanical equipment such as wind turbines generally has limited fault datasets from real-world operations for digital model development. This often leads to poor accuracy in implementing a model based on the life prediction. To address this data shortage issue in developing deep learning models, a remaining useful life prediction approach is proposed in this paper, which combines digital twin technology with transfer learning theory and the embedded convolutional long short-term memory (CLSTM) extended model. First, the main bearing of a direct-drive wind turbine is mapped to the digital world using digital twin technology, allowing the fault datasets of main bearings to be generated and thereby ensuring the model is trained sufficiently with a balanced dataset. The CLSTM network then performs convolutional operations on input-to-state and state-to-state transitions, thereby integrating the time dependence and time-frequency characteristics of the data. Meanwhile, transfer learning is used to transfer the trained model to the wind field for real-world fault diagnostics and the life prediction of the main bearings. Finally, the approach is applied to predict the life of the main bearings, and is also compared with other methods of similar type. The results verified that the proposed approach can effectively overcome the low data density of large equipment, greatly improving the accuracy of life prediction.

Accurate diagnosis of incipient faults in wind turbine (WT) assets will provide sufficient lead time to apply an optimal maintenance for the expensive WT assets which often are located in a remote and harsh environment and their maintenance usually needs heavy equipment and highly skilled engineers. This paper presents an online bearing clearance monitoring approach to diagnose the change of bearing clearance, providing an early and interpretable indication of bearing health conditions. A novel dynamic load distribution method is developed to efficiently gain the general characteristics of vibration response of bearings without local defects but with small geometric errors. It shows that the ball pass frequency of outer race (BPFO) is the primary exciting source due to biased load distribution relating to bearing clearance. The geometric errors, including various orders of runouts on different bearing parts, can be the secondary excitation source. Both sources lead to compound modulation responses with very low amplitudes, being more than 20 dB lower than that of a small local defect on raceways and often buried by background noise. Then, Modulation Signal Bispectrum (MSB) is identified to purify the noisy signal and Gini index is introduced to represent the peakness of MSB results, thereby an interpretable indicator bounded between 0 and 1 is established to show bearing clearance status. Datasets from both a dedicated bearing test and a run-to-failure gearbox test are employed to verify the performance and reliability of the proposed approach. Results show that the proposed method is capable to indicate a change of about 20 µm in bearing clearance online, which provides a significantly long lead time compared to the diagnosis method that focuses only on local defects. Therefore, this method provides a big opportunity to implement more cost-effective maintenance works or carry out timely remedial actions to prolong the lifespan of bearings. Obviously, it is applicable to not only WT assets, but also most rotating machines.

Under the variable speed condition, the rolling bearing vibration signal shows non-smooth characteristics such as frequency modulation and amplitude modulation, which makes it difficult to effectively extract the fault characteristics by relying on the analysis method of traditional signals based on the smooth assumption, as well as the excessive speed change and the excitation response caused by the failure of the equipment components will lead to the change of the detected vibration signal. In this paper, a lightweight fault diagnosis framework called variable speed fault diagnosis framework (VSFD-Net) is proposed to address the above challenges. Firstly, a separable multi-scale convolutional block is proposed to achieve the extraction of multi-local sensory field features, thus reducing the learning parameters and computational effort during the training process. Secondly, the development of a broadcast self-concern block is introduced to achieve capturing the critical fine-grained global extent of the signal while avoiding complex operations such as matrix multiplication and multi-dimensional exponentials. Through experiments on three mechanical systems, it is demonstrated that the framework has significant advantages over existing neural network models, possessing lightweight and robustness.

Due to the symmetry of their structure and the way they rotate, rolling bearings are often used at high temperatures and high speeds. When the temperature changes, the material properties of the rolling bearing change, which in turn causes the dynamic model equation of the rolling bearing to change. Given the problem in that the conventional dynamic model equation does not consider temperature changes, but because temperature changes cause changes in material performance parameters resulting in differences between the dynamic simulation signals and the actual signals, this paper fully considers three factors, namely, the thermal expansion coefficient, the hardness value, and the friction coefficient. With the influence of temperature changes, a study on the mechanism of the symmetric mechanism-rolling bearing dynamic system considering the influence of temperature factors is proposed. First, combined with the material properties of the rolling bearing, the changes in thermal expansion coefficient, hardness value, and friction coefficient caused by temperature changes were analyzed. Second, the functional relationship expressions obtained above were substituted into the existing conventional dynamic model. Finally, the comprehensive functional relationship formula was substituted into the conventional dynamic model to obtain the rolling bearing dynamic model under temperature difference changes. By studying the mechanism of the symmetric mechanism-rolling bearing dynamic system that considers the influence of temperature factors, this method can undertake a more comprehensive consideration of the dynamic analysis research into rolling bearing fault diagnosis, thereby verifying the effectiveness of the method.

The turn domain resampling (TDR) method is proposed in the paper on the basis of the existing angle domain resampling for solving the problem of non-fixed fault frequency under variable working conditions. TDR can select the appropriate sampling order according to the influence of frequency conversion, which avoided the error caused by the spline interpolation method. It can provide accurate parameters for the subsequent calculation of the equivalent frequency order. Variable multi-scale morphological filtering (VMSMF) method is proposed for the purpose of further reducing the interference of noise in resampling signal to feature extraction. VMSMF adaptively selects structural elements according to the parameter change of impact signal to make its scale more targeted. It only needs to calculate once using the optimal structural unit for a particular impact, and the filtering accuracy and operating efficiency have been greatly improved. The main steps of this article are as follows. First, the TDR is used to resample the original signal as to get the resampling signal which is still submerged by the strong noise. In the second step, VMSMF is used to filter the resampling signal to obtain the signal with less noise interference. Finally, the fault characteristics of the filtering signal was extracted and compared with the possible fault frequency calculated by the sampling parameters provided by resampling, so as to determine the fault type of the planetary gearbox. By analyzing the simulation signal and the experimental signal respectively, this method can find out the corresponding fault characteristics effectively.

As the critical parts of wind turbines, rolling bearings are prone to faults due to the extreme operating conditions. To avoid the influence of the faults on wind turbine performance and asset damages, many methods have been developed to monitor the health of bearings by accurately analyzing their vibration signals. Stochastic resonance (SR)-based signal enhancement is one of effective methods to extract the characteristic frequencies of weak fault signals. This paper constructs a new SR model, which is established based on the joint properties of both Power Function Type Single-Well and Woods-Saxon (PWS), and used to make fault frequency easy to detect. However, the collected vibration signals usually contain strong noise interference, which leads to poor effect when using the SR analysis method alone. Therefore, this paper combines the Fourier Decomposition Method (FDM) and SR to improve the detection accuracy of bearing fault signals feature. Here, the FDM is an alternative method of empirical mode decomposition (EMD), which is widely used in nonlinear signal analysis to eliminate the interference of low-frequency coupled signals. In this paper, a new stochastic resonance model (PWS) is constructed and combined with FDM to enhance the vibration signals of the input and output shaft of the wind turbine gearbox bearing, make the bearing fault signals can be easily detected. The results show that the combination of the two methods can detect the frequency of a bearing failure, thereby reminding maintenance personnel to urgently develop a maintenance plan.