Search

Working conditions of rolling bearings of wind turbine generators are complicated, and their vibration signals often show non-linear and non-stationary characteristics. In order to improve the efficiency of feature extraction of wind turbine rolling bearings and to strengthen the feature information, a new structural element and an adaptive algorithm based on the peak energy are proposed, which are combined with spectral correlation analysis to form a fault diagnosis algorithm for wind turbine rolling bearings. The proposed method firstly addresses the problem of impulsive signal omissions that are prone to occur in the process of fault feature extraction of traditional structural elements and proposes a “W” structural element to capture more characteristic information. Then, the proposed method selects the scale of multi-scale mathematical morphology, aiming at the problem of multi-scale mathematical morphology scale selection and structural element expansion law. An adaptive algorithm based on peak energy is proposed to carry out morphological scale selection and structural element expansion by improving the computing efficiency and enhancing the feature extraction effect. Finally, the proposed method performs spectral correlation analysis in the frequency domain for an unknown signal of the extracted feature and identifies the fault based on the correlation coefficient. The method is verified by numerical examples using experimental rig bearing data and actual wind field acquisition data and compared with traditional triangular and flat structural elements. The experimental results show that the new structural elements can more effectively extract the pulses in the signal and reduce noise interference, and the fault-diagnosis algorithm can accurately identify the fault category and improve the reliability of the results.

T-distributed stochastic neighbor embedding (t-SNE) is an effective visualization method. However, it is non-parametric and cannot be applied to steaming data or online scenarios. Although kernel t-SNE provides an explicit projection from a high-dimensional data space to a low-dimensional feature space, some outliers are not well projected. In this paper, bi-kernel t-SNE is proposed for out-of-sample data visualization. Gaussian kernel matrices of the input and feature spaces are used to approximate the explicit projection. Then principal component analysis is applied to reduce the dimensionality of the feature kernel matrix. Thus, the difference between inliers and outliers is revealed. And any new sample can be well mapped. The performance of the proposed method for out-of-sample projection is tested on several benchmark datasets by comparing it with other state-of-the-art algorithms.

A visualization approach for process fault detection using bikernel t-distributed stochastic neighbor embedding (bikernel t-SNE) is described in this paper. Bikernel t-SNE preserves the dimension-r...

The SCADA data of wind turbine contains a number of abnormal data, which seriously affects the prediction and evaluation of the health state of the turbine. Therefore, an abnormal data cleaning algorithm is proposed according to the measured wind speed - power and generator speed - power data. In this algorithm, the key SCADA parameters were selected as data cleaning objects based on the combination of the Copula function and the mutual information feature (CMI). After that, an algorithm combining the Copula function and the local outlier factor (Copula-LOF) was proposed to clean the abnormal data and filter out the abnormal data of outliers and stacking points. Finally, the actual operation data of wind turbines are used to verify the algorithm. The experimental results show that the algorithm can effectively identify all kinds of abnormal data and has a good cleaning effect on abnormal data.

To improve the wind turbine shutdown early warning ability, we present a generalized model for wind turbine (WT) prognosis and health management (PHM) based on the data collected from the SCADA system. First, a new condition monitoring method based on kernel entropy component analysis (KECA) was developed for nonlinear data. Then, an aggregate statistic T was designed to express the state change of the monitoring parameters. As the features were submerged because of the diversity and nonlinearity of SCADA data, an enhanced generalized regression neural network (GRNN) method—KECA-GRNN—for failure prediction was developed by adding KECA for feature extraction to improve the predictive performance. Finally, the results of the KECA-GRNN model were visualized by a bubble chart, which made the health assessment results of the WT more intuitive. Similarly, the fusion residual was defined to analyze the health trend of the WT, and the health status of the WT was represented by two visualization methods—bubble chart and fuzzy comprehensive evaluation. Furthermore, they were evaluated using SCADA data that were collected from a wind farm. Observations from the results of the model indicated the ability of the approach to trend and assess turbine degradation before known downtime occurrences.

In the study of brain-computer interfaces (BCIs) based on steady-state visual evoked potentials (SSVEPs), how to improve the classification accuracies of BCIs has always been the focus of researchers. Canonical correlation analysis (CCA) is widely used in BCI systems of SSVEPs because of its rapidity and scalability. However, the classical CCA algorithm always encounters the difficulty of low accuracy in a short time.For targetless stimuli, this paper proposes a fusion algorithm (CCA-CWT-SVM) that is combined with CCA, a continuous wavelet transform, and a support vector machine (SVM) to improve the low classification accuracies when a single feature extraction method is used.This fusion algorithm achieves high accuracies and information transfer rates (ITRs) in the SSVEP paradigm with few targets.Through the study of 400 groups of experimental data from 10 subjects, the results show that CCA-CWT-SVM has a classification accuracy of 91.76% within 2 s and an ITR of 48.92 bits/min, which are 10.88% and 13.18 bits/min higher than those of the standard CCA. Compared with a mainstream EEG decoding algorithm, filter bank canonical correlation analysis (FBCCA), the classification accuracy and ITR of the CCA-CWT-SVM algorithm also improved (4.45% and 5.69 bit/min, respectively). Using a dataset from Tsinghua University (THU), we also showed that the fusion algorithm is better than the classical algorithms. The CCA-CWT-SVM algorithm obtained an 89.1% accuracy and a 39.91 bit/min ITR in a time window of 2 s. The results were significantly improved compared with those of CCA and the FBCCA (CCA: 79.44% and 28.23 bits/min, FBCCA: 84.03% and 33.4 bits/min). Hence, this work provides an experimental basis for designing an SSVEP-based BCI system with a high task classification accuracy in some crucial biomedical applications.

For traditional pattern recognition fault diagnosis methods, a large number of labeled data is a necessary condition for training effective models, but it is almost difficult to collect enough labeled data in industrial application scenarios. Therefore, an intelligent fault diagnosis algorithm based on wavelet transform and twin convolution depth neural network is proposed and applied to the fault identification of water chillers. Firstly, the fault data is transformed by wavelet transform to extract the change information of amplitude and frequency, and then it is transformed into dynamic time-frequency map as the depth network input; secondly, the pooling layer and convolution layer of convolution neural network are used to construct a multi input twin network to map the contrast image data to a low dimensional space, and the similarity comparison is carried out based on Euclidean distance, so as to realize the water chiller detection Fault type intelligent self-learning. Through the simulation of chiller data, it is verified that this method not only has good robustness to the same type of fault with different severity levels, but also can realize the autonomous learning of unknown fault and enhance the intelligence of diagnosis process under incomplete data modeling.

With the features of control flexibility, multiple functions and compact size, solid state transformer (SST) is considered to be a promising candidate for the energy conversion in applications such as charge station, data center, distribution grid, traction systems and so on. Generally, SST is used to transform medium-voltage AC to low-voltage AC, through the intermediate DC link. Two modular SST topologies based on the back-to-back H-bridge cells with the high-frequency bus as the extra intermediate link are proposed in this paper. Compared with existing modular SST topologies, the two topologies have smaller module counts, thus lower lost and higher efficiency. The working principle of one topology is analyzed as an example, and the corresponding control strategies and design criterion are provided. The feasibility of the proposed topology is verified using the simulation results of a 10kV/380V SST with the capacity of 1MVA.

BCI (Brain Computer Interface, BCI) system means that between the human brain and a computer or other electronic device established communication system. Motor imagination EEG signals are derived from the active consciousness of the human brain, spontaneously in the brain's movement imagination process of different parts of the limbs, without external stimulation.At present, as a mainstream brain-computer interface input signal, it has been increasingly studied. Traditional motor imagination EEG signal classification research is often based on the C3 and C4 dual-channel EEG signal features for extraction and differentiation, and the classification and recognition accuracy is low. In this paper, the wavelet transform coefficients of F3, F4, C3, C4, FZ, CZ, FC1, FC2, FC5 and FC6 multi-channel motor imagination EEG signals are introduced as features, and the extracted features are classified by Support Vector Machines (SVM). Proposed multi-channel Characteristics Analysis Method based on wavelet transform (Multi-channel Characteristics Analysis Method, MCCAM) with dual channel Analysis Method and Common Spatial Pattern (Common Spatial Pattern, CSP) Method are carried on the comparative study, found that multi-channel Analysis Method to imagine the classification accuracy of EEG signals is significantly higher than other two methods, so as to verify the effectiveness of the proposed multi-channel Method.

Performance degradation of wind turbines (WT) is inevitable in long-term operation. Therefore, reliable and intuitive visualization of their health state are essential for condition-based maintenance. It is, however, difficult to extract effective information from the SCADA data due to its massive scale, high dimensionality and nonlinearity. Dimensionality reduction (DR) techniques provide an efficient way for structure visualization of complex datasets. In this paper, we use the diffusion map (DM) algorithm for DR and propose a health state visualization methodology, diffusion map-features geodesic distance (DM-FGD). The DM-FGD fulfills online operation by calculating multivariate statistics of the SCADA data and then reducing its dimensionality. The extracted features present the typical manifold and clustering characteristics. Accordingly, we construct a health indicator, confidence value (CV), and analyze the embedding features by clustering. The CV describes the performance degradation process of the WT and the clustering results illustrate its health status. We further propose a centroid sliding technique to extract the monotonicity characteristic of the CV. Case studies confirm the efficiency of the DM-FGD in real-time health state visualization, and justify its effectiveness and reliability by comparing it with other deviation-based methods. We also verify its extendibility through comparison studies with other DR algorithms.

The modeling accuracy of an AGV (Automated Guided Vehicle) mathematical model is a crucial factor affecting the control accuracy of the AGV. Traditional AGVs are usually controlled by PID control algorithms when the system structures are unknown. The parameter configuration of the controllers mainly rely on manual trial and error-method. Therefore, the task volume is large, the performance of the controller is difficult to be guaranteed. In order to solve this problem, this paper uses ARX model and parameter identification method to model the controlled object according to the actual engineering requirements of the magnetic navigation AGV. Based on the identified model, a hybrid particle swarm optimization (HPSO, Hybrid Particle Swarm Optimization) is used to optimize the parameters of the PID controller of the magnetic navigation AGV. Using the integrated error criterion (ITAE) as the error performance index, the control effects of the optimal parameters found by the trial and error method, standard PSO, and HPSO on the magnetic navigation AGV were compared. The results show that the PID controller designed with the HPSO method has better control performance indicators and can improve the adaptability and robustness of the control system.

Aiming at nonlinearity and multimodal batch trajectories in semiconductor manufacturing processes, principal component analysis and k nearest neighbor (kNN)-related methods were previously presented. However, these methods require data unfolding and are not capable of extracting crucial features, which affects the performance of fault detection. In this paper, an automated fault detection method using convolutional auto encoder (CAE) and k nearest neighbor rule is proposed. Firstly, data collected in one batch is considered as a two-dimensional gray-scale image, and is input to CAE for feature unsupervised learning, with no need of data preprocessing and data labels. Secondly, kNN rule is incorporated into CAE to construct the monitoring index and perform fault detecting. Finally, the effectiveness of the proposed method is verified with a benchmark semiconductor manufacturing process.

Due to the complex dynamic behavior in fermentation process, real-time online fault monitoring is very difficult. In this paper, an ensemble learning method, based on a statistical model and a mechanism model, is presented to monitor the fault. First, the linear and nonlinear information which have great effect on fault monitoring are extract by principal component analysis (PCA) and kernel entropy component analysis (KECA). Second, the judgment conditions of faulty information were determined by the dynamic parameter change information of the mechanism model. Third, Bayesian inference is used to transform the monitoring statistics into fault probabilities to integrate the monitoring statistics. Finally, based on the data in a real fermentation process, simulation experiments are carried out. The results show that the monitor model using the ensemble learning has better monitor accuracy than some other methods.

Multiple phases with transitions from phase to phase are important characteristics of many batch processes. The linear characteristics of batch processes are usually taken into consideration in the traditional algorithms while the nonlinearity is neglected. However, to monitor batch processes more accurately and efficiently, such process features are needed to be considered carefully. In this paper, a new similarity index based on KECA (kernel entropy component analysis) is defined for batch processes with nonlinear characteristics. A new phase division and monitoring method based on the proposed similarity index is brought forward simultaneously. First, nonlinear characteristics can be extracted in feature space via performing KECA on each preprocessed time-slice data matrix. Then phase division is achieved with the similarity change of the extracted feature information. By establishing a series of KECA models for transitions and steady phases, it reflects the diversity of transitional characteristics objectively and can preferably solve the stage-transition monitoring problem in multistage batch processes. Finally, in order to overcome the problem that the traditional contribution plot cannot be applied to the kernel mapping space, a nonlinear contribution plot diagnosis algorithm is proposed. Both results of simulation study and industrial application clearly demonstrate the effectiveness and feasibility of the proposed method.

To quantitatively assess the safety against seismic collapse of eccentrically braced steel frame (EBSF) system, 24 typical EBSFs with K-shape and V-shape braces with seismic precautionary intensities 8 and 9 were designed complying with China seismic design code and relative codes to constitute archetype space of this structure system. In the archetype space, the collapse probability of the structural system under maximum considered earthquakes (MCE) was researched. The results show that the structures possess necessary safety against seismic collapse when they respectively encounter the maximum considered earthquakes corresponding to their seismic precautionary levels, and their collapse probabilities increase with increasing seismic precautionary intensities. Moreover, the EBSFs with V-shape braces have smaller collapse probability, thus greater capacity against seismic collapse than those with K-shape braces.

As the main concerns of abnormal event management in process engineering, fault detection and diagnosis have attracted more and more attention recently. A new monitoring method based on kernel entropy component analysis(KECA) is proposed for nonlinear chemical process. Then, an angle-based statistic is designed to express the distinct angular structure that KECA reveals, which is able to measure the similarity between probability density functions. Likewise, each KECA classifier is dedicated to a specific fault, which provides an expendable framework for incorporating new faults identified in the process. As to the fault features are submerged because of multi-scale property of process data, an enhanced KECA method for fault detection and diagnosis is developed, by adding multi-scale principal component analysis(MSPCA) for features extraction to improve the classification effect of KECA. The effectiveness of the proposed approach is demonstrated by applying to Tennessee Eastman process. The MSPCA based method essentially captures the fault-symptom correlation, whereas KECA can be an effective method for process fault diagnosis.

The conventional data-driven soft sensor methods such as multiway partial least squares have been encountering nonlinear problems in predictions of batch processes, and kernel methods have been used to deal with these problems. In this work, a new data-driven soft sensor method is proposed by developing a Reduced Dual Kernel multiway partial least squares algorithm. First, the number of kernel vectors is reduced by the feature vector selection method. Then, by projecting both input data and the output data into two reduced kernel spaces, dual kernel matrices are established. These two matrices can be used to build PLS models. Finally, the predicted data in the kernel space can be reversely projected onto its original space during online prediction. Comparisons were made among the proposed method and some pervious algorithms through a numerical example and an Escherichia coli fermentation batch process.