Signal measurement and diagnosis of wind turbine gearbox are very important for equipment maintenance. Generative adversarial networks (GAN) are particularly outstanding in data generation due to its game mechanism. An improved gear fault diagnosis method based on GAN for unbalanced data sets is proposed in this paper. Firstly, the fault data is encoded by binary vectorization based on kurtosis perceptron. Secondly, the individuals that fit the fitness are selected by the deterministic selection and the macro factor code string is crossed by multiple points. Then, gaussian mutation is focused on searching for local fault points. Finally, the nonlinear decision boundary is established by the logistic regression auxiliary classifier. The effectiveness of the proposed method was verified by three groups of comparison experiments. Compared with the existing methods, the proposed method has a better ability for fault feature generation, classification, and diagnosis accuracy under unbalanced data sets.

Fault coupling and fault override are common phenomena when faults occur in different parts of the planetary gearbox. Labeled compound fault samples are very rare or even unavailable in industrial scenarios. Thus, it is a challenging task to decouple the compound fault and detect the compound fault with unequal severity only utilizing single faults. In this paper, an untrained planetary gearbox compound fault diagnosis method based on Adaptive Learning Variational Mode Decomposition (ALVMD) and Dual Scale Squeeze-and-Excitation Convolutional Neural Network (DSSECNN) is proposed. An ALVMD algorithm is established to enhance weak fault characteristics and enrich the diversity of sample information. The DSSECNN intelligent fault diagnosis model is presented, and the lightweight SENet can enhance the sensitivity of the model to channel features. Moreover, the mapping relationship between the compound fault mode and single faults is revealed through the proposed probability formula, which can reduce the dependence of the training model on compound fault samples. Experiments have been performed on the gearbox fault experiment bench in the laboratory to verify the effectiveness and generalization performance of the method. Besides, wind turbine gearbox test results show that the ALVMD-DSSECNN can achieve the highest average accuracy of 98.83% compared with other related methods in variable working conditions, which has a certain guiding significance for practical engineering.

Signal processing is of vital importance to the incipient fault diagnosis and the safety running of wind turbines. To adaptively eliminate noise and retain the underlying fault characteristic signal, an adaptive exponential wavelet threshold denoising method based on chaotic dynamic weight particle swarm optimization with sigmoid-based acceleration coefficients (SBAC-CDWPSO) is proposed in this paper. Firstly, a high-order continuous differentiable adaptive exponential threshold function (AETF) based on stein unbiased risk estimation is put forward to improve the defects of the traditional threshold functions. Secondly, the sine map and the sigmoid-based acceleration coefficients are applied for the velocity updating mechanism in particle swarm optimization (PSO). Meanwhile, the dynamic weight, the acceleration coefficient and the best-so-far position are introduced to update the new position with the previous position and the velocity in PSO. Moreover, the gaussian mutation strategy is added, which can effectively maintain the diversity of the swarm and get rid of local optimization. Thirdly, the SBAC-CDWPSO is used to optimize the threshold iteration process in AETF, which can greatly improve the iteration speed of the optimal threshold, and enhance the noise reduction effect. Experimental results showed that the signal-to-noise ratio of our proposed method was higher and the root mean square error was smaller comparing with the pre-existing algorithms. Moreover, wind turbine generator bearing fault diagnosis classification results illustrated that the fault diagnosis rate of the proposed denoising algorithm was up to 96.67%, indicating that the proposed method has great potential in the incipient fault diagnosis of wind turbine bearings.

The prediction of performance degradation is of great significance for the health monitoring of rolling bearings. When predicting the performance degradation trend for the whole-life data of bearings, there are problems such as long prediction time, high cost and single evaluation criteria for prediction effect, which hinder the accuracy of degradation prediction. To this end, a self-checking long and short-term memory (Sc-LSTM) prediction model is proposed in this paper for predicting the performance degradation trend of bearings. First, Relation was used as a performance degradation indicator and improved using the mean and normalization methods. This approach can effectively alleviate the problems of long forecasting time and high cost. Secondly, the long and short-term memory (LSTM) sets up many different prediction schemes and proposes a test layer by segmental prediction and comparison to select the appropriate test index σ. According to the robust stability of the self-checking method and the high accuracy of LSTM nonlinear prediction, an Sc-LSTM performance degradation prediction model is established. The model introduces the prediction effect assessment of prediction pass rate η to make up for the problem of inadequate assessment of prediction effect by a single error. Finally, using the Sc-LSTM model, relation is used to perform segmental predictive analysis and experimental validation of the overall prediction. Experiments show that the model can effectively improve the prediction accuracy and is feasible in the prediction of bearing performance degradation.

The technology of fault diagnosis helps improve the reliability of wind turbines. Difficulties in feature extraction and low confidence in diagnostic results are widespread in the process of deep learning-based fault diagnosis of wind turbine bearings. Therefore, a probabilistic Bayesian parallel deep learning (BayesianPDL) framework is proposed and then achieves fault classification. A parallel deep learning (PDL) framework is proposed to solve the problem of difficult feature extraction of bearing faults. Next, the weights and biases in the PDL framework are converted from deterministic values to probability distributions. In this way, an uncertainty-aware method is explored to achieve reliable machine fault diagnosis. Taking the fault signal of the gearbox output shaft bearing of a wind turbine generator in a wind farm as an example, the diagnostic accuracy of the proposed method can reach 99.14%, and the confidence in diagnostic results is higher than other comparison methods. Experimental results show that the BayesianPDL framework has unique advantages in the fault diagnosis of wind turbine bearings.

Advancement in measurement techniques has dramatically contributed to the development of the modern manufacturing industry. As the primary fault causing unplanned downtime of mechanical equipment, gearbox compound faults are usually coupled by single faults with unequal severity and are difficult to obtain. In industrial scenarios, monitoring data for extreme operating conditions is not available in advance, and labeling samples is time-consuming and costly. However, current research on unseen compound fault diagnosis relies on numerous labeled samples to train the model, and few studies are conducted on transfer learning and unseen compound fault diagnosis under low label ratios. To address the issue, a gearbox compound fault diagnosis method based on deep adversarial graph convolution transfer learning network (DAGCTLN) under low label ratios is proposed. Specifically, a novel DAGCTLN model, including a feature extractor, two label classifiers, and a discriminator, is constructed to realize the diagnosis of faults in the transfer domain and unseen compound faults in the source and target domains. The feature extractor of a three-layer graph convolutional network is presented to achieve deep extraction of fault features under low label ratios. Then a domain space adversarial mechanism between the feature extractor and discriminator is used to achieve global alignment of the source and target domain features. Furthermore, two label classifiers are constructed, and the adversarial adaptation of the decision boundary is realized by maxi-min the classifier difference to achieve subdomain alignment of the same class features in all domains. Experimental results indicate that DAGCTLN can achieve the highest fault diagnosis accuracy in the transfer domain compared to state-of-the-art algorithms. The average diagnosis accuracy of compound faults in all domains can reach 98.41% even if the label ratio is only 0.1, which provides guiding significance for the safe operation and predictive maintenance of mechanical equipment.