Fault diagnosis of rolling bearings is an essential process for improving the reliability and safety of the rotating machinery. It is always a major challenge to ensure fault diagnosis accuracy in particular under severe working conditions. In this article, a deep adversarial domain adaptation (DADA) model is proposed for rolling bearing fault diagnosis. This model constructs an adversarial adaptation network to solve the commonly encountered problem in numerous real applications: the source domain and the target domain are inconsistent in their distribution. First, a deep stack autoencoder (DSAE) is combined with representative feature learning for dimensionality reduction, and such a combination provides an unsupervised learning method to effectively acquire fault features. Meanwhile, domain adaptation and recognition classification are implemented using a Softmax classifier to augment classification accuracy. Second, the effects of the number of hidden layers in the stack autoencoder network, the number of neurons in each hidden layer, and the hyperparameters of the proposed fault diagnosis algorithm are analyzed. Third, comprehensive analysis is performed on real data to validate the performance of the proposed method; the experimental results demonstrate that the new method outperforms the existing machine learning and deep learning methods, in terms of classification accuracy and generalization ability.

With the use of sensorless control strategy, mechanical position sensors can be removed from the gearbox, so as to decrease the maintenance costs and enhance the system robustness. In this paper, a switching PI control based model reference adaptive system (MRAS) observer using Fuzzy-Logic-Controller (FLC) is introduced for sensorless control of permanent magnet synchronous motor (PMSM) drives. The main work and innovation of this paper include: 1) A disturbance observer based voltage compensation method is proposed to solve the problems of voltage distortion in voltage source inverter (VSI) fed PMSM systems. 2) A double closed-loop FLC is designed for speed and current control. The response performance of speed regulation and the accuracy of the command voltage are improved, and the difficulty of manual adjustment of control parameters is avoided. 3) A sensorless control strategy of a switching PI control based MRAS observer is proposed for stator resistance tracking. By setting different thresholds, the proposed MRAS observer can choose the appropriate PI adaptive mechanism based on the amplitude of the current error variable, which can improve the accuracy of resistance estimation and improve the dynamic performance of sensorless control. 4) The experimental results are given to verify the availability of the proposed scheme.

Fault diagnosis plays an indispensable role in prognostics and health management of rotating machines. In recent years, intelligent fault diagnosis methods based on domain adaptation technology have attracted the attention of researchers. However, a more extensive application scenario of fault diagnosis − partial domain adaptation (PDA) − has not been well-resolved. In this article, for the first time, a novel stacked auto-encoder based partial adversarial domain adaptation (SPADA) model is proposed to solve the fault diagnosis problem in PDA situations. Two deep stack auto-encoders are first designed to extract representative features from the training data (source domain) and test data (target domain), respectively. Then, a weighted classifier based on Softmax is used to weight the features from the source and target domains. Meanwhile, another domain discriminator and label predictor using the Softmax classifier are adopted to simultaneously implement domain adaptation and fault diagnosis. Comprehensive analysis is performed on real data to test the performance of the SPADA model and detailed comparisons are provided; the extensive experimental results show that the diagnosis performance of SPADA outperforms the existing deep learning and domain adaptation methods in dealing with the PDA problem.

As an effective tool for solving dynamic complex value problems, especially dynamic complex value matrix inversion (DCVMI), the complex value zeroing neural network (CVZNN) has been widely applied in various practical industrial fields. However, previous complex value zeroing neural network (PCVZNN) models only achieve unsatisfactory exponential or finite time convergence, which usually does not conform to actual practical situation. Different from PCVZNN models, a complex-valued variant parameter robust ZNN model (CVVPRZNN) with both fixed-time convergence and robustness for solving DCVMI is proposed in this paper. The fixed-time convergence and noise suppression ability of the CVVPRZNN model for solving DCVMI are strictly deduced. Then, three comparative numerical experiments for DCVMI solving, dynamic circuit currents computing and manipulator trajectory tracking of the proposed CVVPRZNN model with other existing PCVZNN models are provided for demonstrating its better solution performances and higher practical application value. It should be noted that the experimental results indicate that regardless of whether it is disturbed by noise or not, the CVVPRZNN model only requires about 0.01 s to complete the corresponding solving task. However, in the same experimental environment, other models either cannot complete the solving task at all, or require about 8 s which is almost 800 times the solving time of the CVVPRZNN model to complete the corresponding dynamic problem solving.

Remaining useful life (RUL) prediction for condition-based maintenance decision making plays a key role in prognostics and health management (PHM). Accurately predicting RUL of the rotating components of complex machines becomes a challenging task for PHM. For many existing methods, the current prediction error of RUL prediction may be accumulated into the future predictions, and thus can lead to a prediction error superposition problem. In this article, the formation mechanism of prediction error superposition is analyzed, and for the first time a deep adversarial long short-term memory (LSTM) prognostic framework is proposed to overcome the major issue related to prediction error superposition. In the proposed framework, a generative adversarial network (GAN) architecture combining the LSTM network and autoencoder (AE) is investigated for bearing RUL monitoring. In the proposed deep adversarial learning prediction framework, due to the potential involvement of long-term and complex tasks, the LSTM network (generator) is used to predict the degradation process of rolling bearings based on available historical data, and a simple but useful AE (discriminator) is used to determine and refine the accuracy of the prediction. Therefore, the AE plays the adversarial role of the LSTM network, and the prediction accuracy of the LSTM network can be significantly improved. For illustration purpose, two practical case studies, which use a series of bearing degradation data and the IEEE PHM 2012 PRONOSTIA datasets, respectively, are presented to show the prediction performance of the proposed method. Experimental results show that the proposed method works very well for vibration monitoring and performs better in comparison with the reference machine learning and deep learning approaches. Impact Statement—The damage of rolling bearing usually leads to a significant consequence to industrial production process. However, the existing remaining useful life (RUL) prediction methods for rolling bearing have a prediction error superposition problem that can affect the multistep prediction performance. The new adversarial learning prognostics model proposed in this article can overcome the problem. The proposed method uses LSTM network as a generator to predict RUL life for rolling bearing, and uses AE as a discriminator to estimate the prediction accuracy. The method can significantly improve the multistep prediction accuracy of RUL for rolling bearing, and provides reliable and scientific strategy in PHM of mechatronics equipment.

Rotating machinery is important to industrial production. Any failure of rotating machinery, especially the failure of rolling bearings, can lead to equipment shutdown and even more serious incidents. Therefore, accurate residual life prediction plays a crucial role in guaranteeing machine operation safety and reliability and reducing maintenance cost. In order to increase the forecasting precision of the remaining useful life (RUL) of the rolling bearing, an advanced approach combining elastic net with long short-time memory network (LSTM) is proposed, and the new approach is referred to as E-LSTM. The E-LSTM algorithm consists of an elastic mesh and LSTM, taking temporal-spatial correlation into consideration to forecast the RUL through the LSTM. To solve the over-fitting problem of the LSTM neural network during the training process, the elastic net based regularization term is introduced to the LSTM structure. In this way, the change of the output can be well characterized to express the bearing degradation mode. Experimental results from the real-world data demonstrate that the proposed E-LSTM method can obtain higher stability and relevant values that are useful for the RUL forecasting of bearing. Furthermore, these results also indicate that E-LSTM can achieve better performance.