Search

The performance of feature is essential to the degradation state identification for hydraulic pumps. The initial feature set extracted from the vibration signal of the hydraulic pump is often high-dimensional and contains redundant information, which undermines the effectiveness of the feature set. The novel three-stage feature fusion scheme proposed in this paper aims to enhance the performance of the original features extracted from the vibration signal. First, sparse local Fisher discriminant analysis (SLFDA) performs intra-set fusion within the two original feature sets, respectively. SLFDA has a good effect on samples with intra-class multimodality, and the feature set fused by it has obvious multivariate normal distribution characteristics, which is conducive to the next fusion. Second, our modified intra-class correlation analysis (MICA) is used to fuse two feature sets in the second stage. MICA is a CCA (Canonical correlation analysis) -based method. A new class matrix is used to modify the covariance matrix between two feature sets, which allows MICA to conveniently inherit the discriminating structure while fusing features. Finally, we propose a feature selection algorithm based on kernel local Fisher discriminant analysis (KLFDA) and kernel canonical correlation analysis (KCCA) to select the desired features. This algorithm based on Max-Relevance and Min-Redundancy (mRMR) framework solves the problem that CCA cannot properly evaluate the correlation between features and the class variable, as well as accurately evaluates the correlation among features. Based on the experimental data, the proposed method is compared with several popular methods, and the feature fusion methods used in some previous studies related to the fault diagnosis of rotating machinery are compared with it as well. The results show that the fusion effectiveness of our method is better than other methods, which obtains higher recognition accuracy.

Degradation state identification for hydraulic pumps is crucial to ensure system performance. As an important step, feature extraction has always been challenging. The non-stationary and non-Gaussian characteristics of the vibration signal are likely to weaken the performance of traditional features. In this paper, an efficient feature extraction algorithm named multi-scale ternary dynamic analysis (MTDA) is proposed. MTDA reconstructs the phase space based on the given signal and converts each embedding vector into a ternary pattern independently, which enhances its capacity of describing the details of non-stationary signals. State entropy (SE) and state transition entropy (STE) are calculated to estimate the dynamical changes and complexity of each signal sample. The excellent performance of SE and STE in detecting frequency changes, amplitude changes, and the development process of fault is verified with the use of four simulated signals. The proposed multi-scale analysis enables them to provide a more precise estimation of entropy. Furthermore, support vector machine (SVM) and nondominated sorting genetic algorithm II (NSGA-II) are introduced to conduct feature selection and state identification. NSGA-II and SVM can conduct the joint optimization of these two goals. The details of the method proposed in this paper are tested using simulated signals and experimental data, and some studies related to the fault diagnosis of rotating machinery are compared with our method. All the results show that our proposed method has better performance, which obtains higher recognition accuracy and lower feature set dimension.

Blood glucose (BG) prediction is an effective approach to avoid hyper- and hypoglycemia, and achieve intelligent glucose management for patients with type 1 or serious type 2 diabetes. Recent studies have tended to adopt deep learning networks to obtain improved prediction models and more accurate prediction results, which have often required significant quantities of historical continuous glucose-monitoring (CGM) data. However, for new patients with limited historical dataset, it becomes difficult to establish an acceptable deep learning network for glucose prediction. Consequently, the goal of this study was to design a novel prediction framework with instance-based and network-based deep transfer learning for cross-subject glucose prediction based on segmented CGM time series. Taking the effects of biodiversity into consideration, dynamic time warping (DTW) was applied to determine the proper source domain dataset that shared the greatest degree of similarity for new subjects. After that, a network-based deep transfer learning method was designed with cross-domain dataset to obtain a personalized model combined with improved generalization capability. In a case study, the clinical dataset demonstrated that, with additional segmented dataset from other subjects, the proposed deep transfer learning framework achieved more accurate glucose predictions for new subjects with type 2 diabetes.

As a kind of renewable energy with economic and environmental friendliness, wind energy is widely used, although the high uncertainty of wind speed has significant impact on the operation and planning of wind power systems. Thus, scenario generation of wind speed is the primary step to obtain optimal decisions. However, as to the wind power plants to be newly-built or expanded, adequate wind data may not be available, or even the wind data are missing or invalid, which may lead to the inaccuracy of data-driven scenario generation. In this paper, considering that multiple wind plants in neighboring areas may have similar wind patterns, a novel scenario generation framework is proposed to transfer the knowledge extracted from existing data-rich plants to help establish the uncertainty model of newly-built plant. Transfer component analysis (TCA) is used to minimize the distribution gap between multiple source plants and target plant, and the results of maximum mean discrepancy (MMD) can be obtained to construct the mixture distribution model of target wind speed. Experimental results show that the scenarios generated by transferring the knowledge of existing wind plants can better reflect the real characteristics of the target wind speed in the case of insufficient historical data.

Temporal action localization in untrimmed videos is an important but difficult task. Difficulties are encountered in the application of existing methods when modeling the temporal structures of videos. In the present study, we develop a novel method, referred to as the Gemini Network, for effective modeling of temporal structures and achieving high-performance temporal action localization. The significant improvements afforded by the proposed method are due to three major factors. First, temporal dependencies are explicitly distinguished as long-term temporal dependencies and short-term temporal dependencies and are separately captured by two dedicated subnets. Second, a long-range temporal dependency capture module combined with a self-adaptive pooling module is proposed to capture long-term temporal dependency. Third, the proposed method uses auxiliary supervision, with the auxiliary classifier losses affording additional constraints for improving the modeling capability of the network. As a demonstration of its effectiveness, the Gemini Network is used to achieve a state-of-the-art temporal action localization performance on two challenging datasets, namely, THUMOS14 and ActivityNet.

An automatic recognition of discontinuities in borehole images is a desirable way to overcome the inefficiency and inconsistency inherent in the conventional method of manual annotation. The rough borehole walls and prominent noise failed the application of existing recognition methods to borehole images taken in the coal-measure strata. This paper presents a novel approach to automatically convert coal-measure strata borehole images into identified discontinuity maps. The developed procedure formed an integrated feature representation for the borehole image through combining the color information of image and the textural features generated from multi-channel filtering. Image regions containing discontinuities are then separated from other regions by implementing fuzzy c-means clustering on the acquired feature representation. The identification of discontinuities is finally accomplished by searching for four predefined patterns (named topographic model) on the intensity transection of image regions. The proposed method is proven to be superior in the respects of noise suppression, discontinuity positioning, and recognition completeness.

The accuracy of wind speed scenario generation is critical to the energy dispatching and planning of the wind-integrated power system. The accuracy of the modeling of wind speed can be improved with a comprehensive consideration of the dependency structure between wind speeds, and then the accuracy of wind speed scenario generation can be improved. Therefore, spatial and temporal dependence have been widely studied. However, linear correlation does not fully reflect the nonlinear nature of wind speed. Tail dependence, as a kind of non-linear dependency structure of wind speed in space and time, is studied in this paper. This paper proposes a wind speed scenario generation method based on the C-vine copula that considers spatiotemporal tail dependency structure. With the consideration of spatiotemporal dependency structure, the number of modeled random variables increase significantly. Therefore, a two-step wind speed scenario generation method is proposed to avoid the dimensional disaster. The wind speed data of two wind farms provided by the National Renewable Energy Laboratory are used in simulation analysis, the results demonstrate that reasonable consideration of the spatial and temporal tail dependence of wind speed can improve the accuracy of the spatial and temporal models, and further improve the accuracy of scenarios.

It is important for Bayesian network (BN) structure learning, a NP-problem, to improve the accuracy and hybrid algorithms are a kind of effective structure learning algorithms at present. Most hybrid algorithms adopt the strategy of one heuristic search and can be divided into two groups: one heuristic search based on initial BN skeleton and one heuristic search based on initial solutions. The former often fails to guarantee globality of the optimal structure and the latter fails to get the optimal solution because of large search space. In this paper, an efficient hybrid algorithm is proposed with the strategy of two-stage searches. For first-stage search, it firstly determines the local search space based on Maximal Information Coefficient by introducing penalty factors p1, p2, then searches the local space by Binary Particle Swarm Optimization. For second-stage search, an efficient ADR (the abbreviation of Add, Delete, Reverse) algorithm based on three basic operators is designed to extend the local space to the whole space. Experiment results show that the proposed algorithm can obtain better performance of BN structure learning.

The file attached to this record is the author's final peer reviewed version. The Publisher's final version can be found by following the DOI link. Population-based evolutionary algorithms have great potential to handle multiobjective optimisation problems. However, the performance of these algorithms depends largely on problem characteristics. There is a need to improve these algorithms for wide applicability. References, often specified by the decision maker’s preference in different forms, are very effective to boost the performance of algorithms. This paper proposes a novel framework for effective use of references to strengthen algorithms. This framework considers references as search targets which can be adjusted based on the information collected during the search. The proposed framework is combined with new strategies, such as reference adaptation and adaptive local mating, to solve different types of problems. The proposed algorithm is compared with state-of-the-arts on a wide range of problems with diverse characteristics. The comparison and extensive sensitivity analysis demonstrate that the proposed algorithm is competitive and robust across different types of problems studied in this paper.

This paper describes an example of a successful attempt to migrate an intensive hardware-software lab into a fully personalized scheme with team interactions. Each student in the course performed each lab individually, and was expected to collaborate with another selected or assigned class team member remotely. When completed, the team demonstrated the lab results to the Teaching Assistants and the Instructor through a secure Learning Management System installed at the university. The importance of this lab scheme is that it is universal in that the specific form and method of delivery of this scheme is suitable for other secure networking environments.

A steady-state 1-D code called BWIC is developed. The validation of the code is performed and comprehensive results for a 1-THz staggered double vane TWT are presented in this paper. Results from the BWIC code show good agreement with output power, gain, and drive curve of the CST.

Randomness complexity is a kind of features which is widely used to describe bearings’ degradation. However, different randomness complexities present different properties. It is necessary to figure out different randomness complexities’ properties. In this paper, we are going to make comparisons of seven commonly used randomness complexities namely approximate entropy, sample entropy, fuzzy entropy, Shannon entropy, permutation entropy, Lempel-Ziv complexity and C0 complexity by simulation signals with three different aspects and two run-to-failure bearing’s data. By comparisons, we have found that there are a kind of similarity between them and we have proposed a trend similarity index to expound this similarity. Based on the comparisons, we can infer that randomness complexities are a family feature of rolling bearings’ degradation. Among the seven discussed complexities, sample entropy has the best performance, and it can be a good representative of the complexity features. In this paper, the difference between complexity features and other features when monitoring bearings’ degradation have been discussed. The research will provide a reference for rolling bearings’ multi-features dimensionality reduction by attribute selection method.

The bearings’ degradation features are crucial to assess the performance degradation and predict the remaining useful life of rolling bearings. So far, numerous degradation features have been proposed. Many researchers have devoted to use dimensionality reduction methods to reduce the redundancy of those features. However, they have not considered the properties and similarity of those features. In this paper, we present a simple way to reduce dimensionality by classifying different features based on their trends. And the degradation features can be classified into two subdivisions, namely, uptrends and downtrends. In each subdivision, there exists visible trend similarity, and we have introduced two indexes to measure this similarity. By selecting the representative features of the subdivision, the multifeatures can be dimensionality reduced. Through the comparison, the root mean square and sample entropy are two good representatives of uptrend and downtrend features. This method gives an alternative way for dimensionality reduction of the rolling bearings’ degradation features.

Fault prognosis plays a key role in the framework of Condition-Based Maintenance (CBM). Limited by the inherent disadvantages, most traditional intelligent algorithms perform not very well in fault prognosis of hydraulic pumps. In order to improve the prediction accuracy, a novel methodology for fault prognosis of hydraulic pump based on the bispectrum entropy and the deep belief network is proposed in this paper. Firstly, the bispectrum features of vibration signals are analyzed, and a bispectrum entropy method based on energy distribution is proposed to extract the effective feature for prognostics. Then, the Deep Belief Network (DBN) model based on the Restrict Boltzmann Machine (RBM) is proposed as the prognostics model. For the purpose of accurately predicting the trends and the random fluctuations during the performance degradation of the hydraulic pump, the Quantum Particle Swarm Optimization (QPSO) is introduced to search for the optimal value of initial parameters of the network. Finally, analysis of the hydraulic pump degradation experiment demonstrates that the proposed algorithm has a satisfactory prognostics performance and is feasible to meet the requirements of CBM.

In the scenario-based stochastic programming problem, the solving complexity and computational burden increases as the number of scenarios increase, which involves necessary scenario reduction operations. For the scenario reduction problem with multiple random variables, we present a comprehensive optimal scenario reduction method based on a new optimization framework to eliminate redundant initial scenarios. In this framework, we propose the concept of correlation loss, to solve the problem of serious deviation of correlation in the reduced scenarios. The method uses a corrloss weight to balance the two objectives of the proposed scenario reduction framework. One is to minimize the correlation loss before and after scenario reduction, and the other is to maximize the similarity between the original scenario set and the reduced scenario set. In particular, it has strong universality. Numerical comparative experiments are carried out to evaluate the performance of the method. Application studies to microgrid economic operation optimization are also made to prove the effectiveness of the proposed method.

Piezoelectric ceramic cracking (PCC) is the main reason leading to failure of ultrasonic motors. To solve the problem that the fault information is too weak to reflect the cracking condition especially in the early degradation stage, a degradation state identification method based on morphological boundary span coverage statistics was proposed in this paper. Firstly, the average morphological cover area of standard data was adopted as the degradation feature for PCC. Then standard degradation state rectangles (SDSRs) were constructed based on the degradation feature. The height of SDSR was optimized to improve the classification accuracy via training data with the help of genetic algorithm. Lastly, the coverage statistics obtained by the relationship between test data’s morphological boundary span signal and the constructed SDSR can be taken as a fair indicator for the actual degradation state. The experimental results show that this method is feasible and effective, and could achieve a satisfying performance to identify the different degradation states of PCC.

To adapt to the change in the environment, the model needs to own the ability to update to ensure the performance of the decision. In this paper, a new updating strategy is proposed for the safe control Bayesian network (BN) model under the abnormity in the thickening process of gold hydrometallurgy. First of all, the abnormality in the thickening process of gold hydrometallurgy is analyzed deeply. The new safe control BN model is established for three main abnormities. Furthermore, the general framework of BN model updating strategy based on the new expert knowledge and the new dataset is proposed, which mainly includes the parameter updating learning and the structure updating learning. For the structure updating learning, the useful information in the established model is reserved, and only the partial structure which does not adapt to the change in the environment is changed by searching for the unconformable nodes as the target nodes. Finally, the proposed method is applied to update the safe control BN model in the thickening process of gold hydrometallurgy. The simulation results demonstrate that it is effective and owns the better performances to update the established model as the change of the dosage of flocculants.

The full vector spectrum is an effective and efficient tool for homologous multi-sensor data fusion in rotating machinery. However, this methodology just takes Fourier Transform to obtain the harmonic trajectory information hidden in multi-sensor data and it has some drawbacks of processing nonlinear, multi-frequency and noise-containing data. To address this critical issue, this paper provides a novel approach called EWT-VCR based on Empirical Wavelet Transform (EWT) and Variance Contribution Rate (VCR) to improve the adaptability and accuracy of the fusion method. EWT is introduced as a signal preprocessing technique to decompose complex signals into variable frequency bands. And VCR is proposed to denoise, fuse EWT components at different frequency bands, and enhance useful harmonic components. The full vector spectrum technology is utilized to carry out the full vector information fusion of the improved multi-sensor signals for further spectrum analysis. The proposed methodology is applied to multi-channel vibration signal fusion for hydraulic pumps to detect specific frequencies related to pump’s degradation process and a novel degradation feature named Full Vector Factor Entropy (FVFE) is extracted to describe hydraulic pump’s degradation process during its life cycle. The effectiveness of the proposed methods is validated through two experimental cases.

When the data of target domain are scarce, the established model will not be accurate enough to analyze the target problem. For the abnormal condition identification modeling problem of electro-fused magnesia smelting process, this paper proposes the new Bayesian network (BN) parameters transfer learning method based on the expert knowledge from target domain to increase the accuracy of abnormal condition identification. First of all, the electro-fused magnesia smelting process is introduced and the existing research results on the abnormal condition identification are analyzed. The problem to solve in this paper is described. Furthermore, the constraints from expert knowledge for the target model are shown in two forms. The new BN parameters transfer learning method is proposed. Finally, the proposed method is verified by the Asia network, and it is applied to establish the abnormal exhausting condition identification model for the electro-fused magnesia smelting process. The simulation results demonstrate the effectiveness of proposed method which owns the better performances.

Feature extraction is a key step of Prognostics and Health Management (PHM). To improve the feature performance, a method based upon the modified composite spectrum and relative entropy is proposed. The DCS algorithm is firstly presented by the modification of earlier composite spectrum for making fusion of multi-channel vibration signals. Considering Shannon entropy and Tsallis entropy, the DCS power entropy and singular entropy are initially extracted. According to max relation entropy criterion and gradual fusion strategy, the relative entropy algorithm is built to fuse the initial features into a new one, which is considered to be the degradation feature. Result of the application in hydraulic pump degradation experiment demonstrates that the proposed algorithm is feasible and the fused feature is effective to measure the performance degradation of pump.

For the complex batch process with characteristics of unequal batch data length, a novel data-driven batch process monitoring method is proposed based on mixed data features analysis and multi-way kernel entropy component analysis &#x0028 MDFA-MKECA &#x0029 in this paper. Combining the mechanistic knowledge, different mixed data features of each batch including statistical and thermodynamics entropy features, are extracted to finish data pre-processing. After that, MKECA is applied to reduce data dimensionality and finally establish a monitoring model. The proposed method is applied to a reheating furnace industry process, and the experimental results demonstrate that the MDFA-MKECA method can reduce the calculated amount and effectively provide on-line monitoring of the batch process.

The purpose of wearable technology is to use multimedia, sensors, and wireless communication to integrate specific technology into user clothes or accessories. With the help of various sensors, the physiological monitoring system can collect, process, and transmit physiological signals without causing damage. Wearable technology has been widely used in patient monitoring and people’s health management because of its low-load, mobile, and easy-to-use characteristics, and it supports long-term continuous work and can carry out wireless transmissions. In this paper, we established a Wi-Fi-based physiological monitoring system that can accurately measure heart rate, body surface temperature, and motion data and can quickly detect and alert the user about abnormal heart rates.

According to the one-dimensional characteristics of the vibration signal, this paper proposes an elevator operation fault monitoring method based on one-dimensional convolutional neural network (1-DCNN). It can solve the problems of traditional elevator fault monitoring methods that require complex feature extraction processes and a large amount of diagnostic experience. Because the elevator fault monitoring field has less fault information, it is different from the large sample situation in the field of face recognition. Aiming at the problem of small samples, this paper first preprocesses elevator vibration signals through singular value decomposition (SVD) and wavelet transform, then uses wavelet transform to extract wavelet energy features of the original vibration signals, and then use PCA to reduce the feature data to the dimension with a cumulative contribution rate of greater than 85%. When reducing the dimensionality, the original characteristics of the features are preserved as much as possible. When designing the 1-CNN, the K-fold cross-validation method is added to obtain as many abnormalities from the sample set as possible. The information is finally trained using the 1-CNN and classified by softmax regression. In order to verify the performance of the algorithm, the original vibration signal was used as the input of the 1-CNN, and the wavelet energy feature without PCA dimensionality reduction was used as the input of the 1-CNN. The experimental results showed that the 1-DCNN model with PCA dimension-reduced feature data as input can effectively extract and identify the features of normal and abnormal states and has high fault identification accuracy, and good results have been obtained.

Fault tree analysis is often used in elevator fault diagnosis because of its simplicity and reliability. However, the traditional fault tree method has the problems of low efficiency due to ignoring location change of bottom events during troubleshooting. This paper proposes a rapid diagnosis method based on multiattribute decision making to solve the problem. The fault tree of the elevator system is constructed based on expert knowledge and multisource data, and the location-related matrix is constructed according to the complex vertical structure of the elevator. Then, the attributes of bottom events such as the failure probability, search cost, location time cost, and location-related attributes are comprehensively analyzed in this paper. Finally, the TOPSIS method for dynamic attributes is used based on the work above to achieve the optimal troubleshooting sequence of elevator vibration fault. The results show that the proposed method is more efficient when compared to the optimal troubleshooting sequence obtained by the traditional method.

We propose an absolute phase retrieval method based on three phase-shifting amount codes (3-PSA-codes) to measure the colored object with one additional pattern. 3-PSA-codes adopt the coding concept of 3-digit-codes, in which the code elements of three consecutive periods are treated as a unique code word for one period. However, to measure the colored object more effectively in the proposed method, each code element is embedded into the PSA domain and retrieved from the phase difference. Fringe patterns for the wrapped phase are artfully employed in the code element retrieval. Hence, for the first time, to the best of our knowledge, the code element related to the phase can be determined by one additional pattern. It breaks the constraint that temporal methods require multiple additional patterns to overcome the adverse effect of the surface color of objects on absolute phase retrieval. Experimental results demonstrate that the proposed 3-PSA-codes have strong robustness in the measurement of the colored object.

As continuous glucose monitoring (CGM) systems provide critical feedback information of blood glucose concentration to the artificial pancreas for patients with type 1 diabetes (T1D), faults in CGM may seriously affect the computation of insulin infusion rates which can lead to fatal consequences accompany with hypoglycemia or hyperglycemia. In the present work, the k-medoids clustering algorithm is modified by calculating cluster number with a Bayesian Information Criterion (BIC)-based cost function and the SAC (SSE-ASW Criterion) evaluation coefficient which considers both SSE (Sum of Square due to Error) and ASW (Average Silhouette Width) criteria. Then, the modified k-medoids clustering algorithm is proposed to detect sensor failures online with CGM measurements. Different from the qualitative model-based methods and quantitative model-based methods, sufficient clean data are the only requirement of the proposed method. During online monitoring, the new glycemic variability is then tracked against predefined confidence limits during training period to indicate abnormality. The feasibility of the proposed method is successfully assessed using CGM data collected from the UVa/Padova metabolic simulator.

Continuous glucose monitoring (CGM) collects a host of time-series sensor data and is committed to fully automated systems for glucose control as an essential part. Outlier data in CGM measurements caused by faults may seriously affect the computation of insulin infusion rates and endanger the safety of patients. In this paper, a semi-supervised outlier detection method is proposed for anomaly detection of glucose concentration measurements based on a density-based clustering algorithm, named independent central point OPTICS (ICP-OPTICS). An optimization function is designed to improve the clustering performance and solve the problem of outlier detection with the distance measurement and information entropy of weighted time series. The proposed method in the application can be configured automatically with no prior knowledge except only some clean samples. The UVa/Padova simulator and real dataset are used to evaluate the performance of the method. Compared with the present work, the statistical results show that the method is much better in effectiveness and superiority. Furthermore, the proposed method also provides a possible reference significance for outlier detection in other practice applications.

Modern complex industrial processes are prone to errors because of interactions between humans, the external environment and the equipment. When the abnormity degree of a system increases, the system can generate failures or even accidents, which result in serious economic loss or even personal casualties. Therefore, it is necessary to take effective measures to remove the abnormity as soon as possible. This problem can be described as the least-time optimization problem. This paper analyses an abnormity by summarizing and comparing related concepts in the researched results. Based on these concepts, a control strategy for the abnormity in a complex industrial process is proposed by analysing the experience of operators on site. Taking the abnormity in the thickening process of gold hydrometallurgy as an example, this paper explores how the abnormity control problem can be transformed into the least-time optimization problem. Technical and mechanical constraints are described. Simulation results indicate that the proposed strategy can assist the operators to regulate the control variables and recover the abnormity as soon as possible. This produces better performance than the existing regulations on site.

To improve the performance of the abnormal condition identification, the multi-source information of the abnormal conditions in the electro-fused magnesia smelting process is analyzed in this paper. An intelligent abnormal condition identification method is proposed based on Bayesian network (BN). By analyzing three main abnormal conditions and the experience of the operators, the characteristics related with the abnormal conditions are extracted. The BNs are established to identify the abnormal conditions by fusing the multi-source information. The simulation results show that the proposed method can realize abnormal condition identification, distinguish the degree of the abnormal condition, and obtain better performance.

Multivariate time series prediction has attracted growing interest in many research fields. Recently, deep learning has been applied to multivariate time series prediction and has achieved encouraging results. However, in real-world scenarios, the insufficient data of multivariate time series at the beginning of the observation causes the deep learning model unable to exert its expected performance. Furthermore, there is the distribution discrepancy between different multivariate time series caused by many factors, making it unfeasible to reuse existing data or models directly. Therefore, a novel Domain Adaptive Deep Recurrent Network (DADRN) is proposed for multivariate time series prediction with insufficient data, which transferring the knowledge of the target-related time series (source domain) to the target time series (target domain) by minimizing distribution mismatch in the feature sharing space. The DADRN automatically learns the temporal dependence of predictive time series and the dynamic dependencies between multiple time variables through the deep recurrent neural network. Besides, a special transfer learning method, domain adaptation, is embedded in the constructed deep recurrent network to reduce the distribution discrepancy between different domains. The proposed domain independence strategy and domain weighted loss further enhance the DADRN’s transfer learning capability by improving the distribution estimation of the target domain and balancing the network’s learning on two domains. The reasonable combination of deep recurrent network and domain adaptation endows DADRN with favorable transfer learning capability, and its effectiveness is demonstrated by the experimental results on two real-world datasets.

This paper develops a safe control scheme under the abnormity based on Bayesian network (BN) for the thickening process of gold hydrometallurgy. By analyzing the causes and corresponding solutions of the abnormity, the operator experience of removing the abnormity is transformed to BN which is used to make safe decisions automatically. The proposed BN-based approach can provide real-time decision support when the abnormity happens. The BN combines the expert knowledge with quantitative data analysis for the abnormity which can make decisions to remove the abnormity. The BN is established off-line and used to infer on-line. After receiving abnormal phenomena as evidences, the posterior probabilities of the decision variables with different grades can be obtained by BN reasoning, which provide real-time safety analysis. The posterior probabilities can be updated using the abnormal latest information. If the abnormity is not removed, once again the BN analysis is invoked based on the newly established situation and the appropriate set of decisions are obtained accordingly. The application results show that the proposed approach can make the safe operation control decisions to remove the abnormity in the thickening process effectively.

Temporal action localization in untrimmed videos is an important but difficult task. Difficulties are encountered in the application of existing methods when modeling temporal structures of videos. In the present study, we developed a novel method, referred to as Gemini Network, for effective modeling of temporal structures and achieving high-performance temporal action localization. The significant improvements afforded by the proposed method are attributable to three major factors. First, the developed network utilizes two sub-nets for effective modeling of temporal structures. Second, three parallel feature extraction pipelines are used to prevent interference between the extractions of different stage features. Third, the proposed method utilizes auxiliary supervision, with the auxiliary classifier losses affording additional constraints for improving the modeling capability of the network. As a demonstration of its effectiveness, the Gemini Network was used to achieve state-of-the-art temporal action localization performance on two challenging datasets, namely, THUMOS14 and ActivityNet.

The extraction of rolling bearings&rsquo, degradation features has been developed for decades. However, the degradation features always present different trends of different run-to-failure data. To find a consistent indicator of different data will be helpful to establish a general model and explore the nature of bearings&rsquo, degradation. In this study, we have found there is a trend of similarity between the energy and complexity features. By using the cointegration test, we found the two kinds of features exhibit a certain degree of cointegration relationship. Fused by the cointegration method, we have obtained a novel health indicator which can depict different run-to-failure data in a unified way. The difference between the energy features and complexity features can be explained by the novel health indicator. The indicator has &ldquo, two-stage&rdquo, characters. The first stage is the zero-line stage and the second stage is the quickly raise stage, which presents like an exponential function. It is easy to think about using an exponential degradation model to model this indicator. Next, we have compared the indicator to root mean square (RMS) by using the exponential degradation model. It shows that the indicator is more suitable for the exponential degradation model. In this paper, we used eleven run-to-failure data to verify the generality and &ldquo, characters of the proposed indicator. The result shows that the novel indicator is general and effective and that it will promote the development of bearings&rsquo, prognostics.

This paper presents a novel vibration signal fusion algorithm using improved empirical wavelet transform and variance contribution rate to fuse three-channel vibration signals for weak fault detection of hydraulic pumps. Firstly, empirical wavelet transform (EWT) is utilized to decompose the three-channel signals into several AM-FM components. Then in accordance with the statistical characteristics of these component data, variance contribution rate is defined to measure the weight of component data points. A series of fusion coefficients are computed and assigned to every component point. Finally, these component points are fused into one single signal and Hilbert transform is conducted to demodulate the fault characteristic frequency for weak fault detection. Moreover, to address the issue of improper EWT spectrum segmentation, we introduce Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to improve EWT in the full space and the frequencies corresponding to outlier points are taken as the boundaries of spectrum segmentation. Therefore, the number of boundaries is more reasonable and the AM-FM components are more consistent with inherent components existing in the vibration signals of pumps. Results of simulation and experiment analysis demonstrate the good performance of the exhibited fusion algorithm in weak fault detection of hydraulic pumps.

The cracking of piezoelectric ceramics is the main reason of failure of an ultrasonic motor. Since the fault information is too weak to reflect the condition of piezoelectric ceramics especially in the early degradation stage, a fault feature extraction method based on multiscale morphological spectrum and permutation entropy is proposed. Firstly, a signal retaining the morphological feature under different scales is reconstructed with multiscale morphological spectrum components. Then, the permutation entropy of the reconstructed signal is taken as the fault feature of piezoelectric ceramics. Furthermore, a sensitivity factor is defined to optimize the embedded dimension and delay time of permutation entropy according to double sample Z value analysis. Finally, a matrix composed of the probability distributions, obtained from permutation entropy calculation, is applied for the degradation state identification by means of probability distribution divergence. The analysis of actual test data demonstrates that this method is feasible and effective.

Temporal Action Localization in untrimmed videos is an important yet difficult task. It’s important for accurate temporal action localization to model temporal structure. In this work, we propose a novel model, Gemini CNN, which can effectively model temporal structure via several separated CNNs. Our first contribution is the idea of modular design of temporal structure modeling. Temporal structure is divided into long-range temporal structure and short-range temporal structure. Two types of temporal structures are respectively modeled in two subnets. The other contribution is the idea of auxiliary supervision. We add classifiers at not only the end of our framework, but also intermediate layers. The loss of auxiliary classifiers increases the gradient signal that gets propagated back, and boosts our framework’s performance. We achieve state-of-the-art performance for temporal action localization on two challenging temporal action localization datasets, THUMOS14 and ActivityNet.

Objective.An electrocardiogram (ECG) is one of the most common means to diagnose arrhythmia according to different waveforms clinically. Although there are advanced classification methods such as deep learning, the single view feature cannot meet the demand of classification accuracy for new individuals. To this end, a classification model based on multiview fusion was proposed.Approach.First, handcrafted view features were extracted from heartbeats and then deep view features were obtained from the deep learning model. The features of two different perspectives were fused in the fully connected layer, and the random forest classifier was used instead of the Softmax classifier for classification. Notably, Bayesian optimization was utilized in the hyper-parameter tuning of the classifier. The proposed method employed the MIT-BIH database to classify five classes: normal heartbeat (N), left bundle branch block heartbeat (LB), right bundle branch block heartbeat (RB), atrial premature contraction (APC) and premature ventricular contraction (PVC).Main results.The experimental results achieved a higher average accuracy of 98.93%, average precision of 96.92%, average sensitivity of 96.46%, and average specificity of 99.33% in five types of heartbeat classification for inter-patient.Significance.The proposed framework improves the performance of ECG detection for new individuals. And it provides an feasible algorithmic model for single-lead wearable devices with multiview fusion.

Performance degradation assessment (PDA) of hydraulic pumps is of great importance to preserve operational reliability and ensure safety of hydraulic systems. PDA of hydraulic pumps relies heavily on degradation monitoring data such as vibration data, acoustic emission data and oil data. However, besides the inherent degradation of pumps, the time-varying load conditions (e.g. output pressure) have a significant influence on the behavior of degradation data. This makes PDA more difficult under varying conditions. To address this issue, this paper proposes a conditional factor variational auto-encoder (CFVAE) model whereby variational theory is firstly applied to degradation data decoupling and degradation characteristics extraction. In the training phase, degradation data decoupling is realized by punishing the total correlation term of the model with a hyper-parameter γ and the degradation characteristics can be extracted from latent code units of the model. However, the punishment can result in large reconstruction losses of the model which is not conducive to degradation assessment. Then the degradation label information is integrated into the decoder of the model to decrease reconstruction loss caused by data decoupling. Moreover, a new metric based on inter-class distance and intra-class divergence is introduced to optimize the hyper-parameter and select the best latent code units for degradation characteristics extraction and description. Using these techniques, the CFVAE models corresponding to all degradation states can be well trained. In the testing phase, degradation data from unknown degradation states is fed into all of the trained CFVAE models respectively and minimum distance criterion is introduced to predict the actual degradation state. Taking return oil flow as degradation data, the proposed CFVAE model along with several advanced methodologies is applied to degradation assessment for hydraulic pumps under varying conditions. The assessment results of ten experimental cases verify that the proposed CFVAE model can adapt better to change of conditions and has higher assessment accuracy than other methodologies. The proposed model also shows good robustness in multiple cases, making it likely to apply this model to other mechanical components.

An online phase measuring profilometry with new layered modulation method for pixel matching is presented. In this method and in contrast with previous modulation matching methods, the captured images are enhanced by Retinex theory for better modulation distribution, and all different layer modulation masks are fully used to determine the displacement of a rectilinear moving object. High, medium and low modulation masks are obtained by performing binary segmentation with iterative Otsu method. The final shifting pixels are calculated based on centroid concept, and after that the aligned fringe patterns can be extracted from each frame. After performing Stoilov algorithm and a series of subsequent operations, the object profile on a translation stage is reconstructed. All procedures are carried out automatically, without setting specific parameters in advance. Numerical simulations are detailed and experimental results verify the validity and feasibility of the proposed approach.

When the data information of target domain is very limited, it is difficult to establish the accurate model to analyze the target problem. For the safety control modeling problem, this paper develops a new Bayesian network (BN) transfer learning strategy for the thickening process of gold hydrometallurgy. First of all, the safety control modeling problem in this process is analyzed deeply. When the data information of abnormality is insufficient, the safety control modeling problem is transformed into the BN transfer learning problem. Furthermore, the new BN transfer learning strategy is proposed, which includes the structure and parameters transfer learning methods. For the structure transfer learning, by integrating the common structural information of multiple sources and the useful information of target, the final structure of target is determined. For the parameters transfer learning, by distinguishing the similarity of multiple sources, the parameters of target are obtained by the fusion algorithm. Finally, the proposed method is verified by the Asia network and it is applied to establish the safety control model for the thickening process of gold hydrometallurgy. The simulation results demonstrate that the proposed method is effective and owns the better performances than the traditional modeling method.

Bayesian network structure learning from data has been proved to be a NP-hard (Non-deterministic Polynomial-hard) problem. An effective method of improving the accuracy of Bayesian network structure is using experts&rsquo, knowledge instead of only using data. Some experts&rsquo, knowledge (named here explicit knowledge) can make the causal relationship between nodes in Bayesian Networks (BN) structure clear, while the others (named here vague knowledge) cannot. In the previous algorithms for BN structure learning, only the explicit knowledge was used, but the vague knowledge, which was ignored, is also valuable and often exists in the real world. Therefore we propose a new method of using more comprehensive experts&rsquo, knowledge based on hybrid structure learning algorithm, a kind of two-stage algorithm. Two types of experts&rsquo, knowledge are defined and incorporated into the hybrid algorithm. We formulate rules to generate better initial network structure and improve the scoring function. Furthermore, we take expert level difference and opinion conflict into account. Experimental results show that our proposed method can improve the structure learning performance.

In this paper, a novel dynamic programming technique is presented for optimal operation of a typical renewable microgrid including battery energy storage. The main idea is to use the scenarios analysis technique to proceed the uncertainties related to the available output power of wind and photovoltaic units and dynamic programming technique to obtain the optimal control strategy for a renewable microgrid system in a finite time period. First, to properly model the system, a mathematical model including power losses of the renewable microgrid is established, where the uncertainties due to the fluctuating generation from renewable energy sources are considered. Next, considering the dynamic power constraints of the battery, a new performance index function is established, where the Lagrange multipliers and interior point method will be presented for the equality and inequality operation constraints. Then, a feedback control scheme based on the dynamic programming is proposed to solve the model and obtain the optimal solution. Finally, simulation and comparison results are given to illustrate the performance of the presented method.

This paper presents a control framework to identify the abnormal condition and make safe control scheme by fusing the related multi-source information based on the analysis of the abnormal exhausting condition in the electro-fused magnesia smelting process. For the abnormal condition identification, the case-based reasoning is used to infer the degree of the abnormal condition. When the similarity of the matched case doesn't exceed the set threshold, Bayesian network is used to infer the degree. For the safe control scheme, taking advantage of the idea of the remaining lifetime predicting, and the adjustment of the control variables is obtained. The application results show that the proposed method is effective and it owns better performance than the manual and the existing research results.

The cracking of piezoelectric ceramic components is one of the main failure pattern of ultrasonic motors. The degradation characteristics can be extracted effectively by monitoring the voltage signal of piezoelectric sensor. High dimensional feature vector contains a lot of redundant information due to dimension disaster. Locality preserving projection (LPP) can effectively reduce dimension on the degradation feature, which can fuse the high dimension feature. The parameters of support vector machine (SVM) have an important influence on the generalization ability of the model. Fruit fly optimization algorithm (FOA) has the advantages of few parameters, fast calculation speed, strong ability of global optimization and easy to implement. FOA is utilized to optimize SVM in this paper accordingly, and the optimized SVM by FOA (FOASvM) is applied in degradation state recognition of ultrasonic motor. Finally, the effectiveness of the proposed method is verified through the comparative analysis.

A linear programming phase unwrapping method in dual-wavelength digital holography is proposed and verified experimentally. The proposed method uses the square of height difference as a convergence standard and theoretically gives the boundary condition in a searching process. A simulation was performed by unwrapping step structures at different levels of Gaussian noise. As a result, our method is capable of recovering the discontinuities accurately. It is robust and straightforward. In the experiment, a microelectromechanical systems sample and a cylindrical lens were measured separately. The testing results were in good agreement with true values. Moreover, the proposed method is applicable not only in digital holography but also in other dual-wavelength interferometric techniques.