Search

Adverse weather conditions such as haze, rain, and snow often impair the quality of captured images, causing detection networks trained on normal images to generalize poorly in these scenarios. In this paper, we raise an intriguing question – if the combination of image restoration and object detection, can boost the performance of cutting‐edge detectors in adverse weather conditions. To answer it, we propose an effective yet unified detection paradigm that bridges these two subtasks together via dynamic enhancement learning to discern objects in adverse weather conditions, called TogetherNet. Different from existing efforts that intuitively apply image dehazing/deraining as a pre‐processing step, TogetherNet considers a multi‐task joint learning problem. Following the joint learning scheme, clean features produced by the restoration network can be shared to learn better object detection in the detection network, thus helping TogetherNet enhance the detection capacity in adverse weather conditions. Besides the joint learning architecture, we design a new Dynamic Transformer Feature Enhancement module to improve the feature extraction and representation capabilities of TogetherNet. Extensive experiments on both synthetic and real‐world datasets demonstrate that our TogetherNet outperforms the state‐of‐the‐art detection approaches by a large margin both quantitatively and qualitatively. Source code is available at https://github.com/yz-wang/TogetherNet. [ABSTRACT FROM AUTHOR]

Urban trees are often densely planted along the two sides of a street. When observing these trees from a fixed view, they are inevitably occluded with each other and the passing vehicles. The high density and occlusion of urban tree scenes significantly degrade the performance of object detectors. This article raises an intriguing learning-related question—if a module is developed to enable the network to adaptively cope with occluded and unoccluded regions while enhancing its feature extraction capabilities, can the performance of a cutting-edge detection model be improved? To answer it, a lightweight yet effective object detection network is proposed for discerning occluded and dense urban trees, called occluded and dense-urban tree detection network (OD-UTDNet). The main contribution is a newly designed dilated attention cross stage partial (DACSP) module. DACSP can expand the fields of view of OD-UTDNet for paying more attention to the unoccluded region, while enhancing the network’s feature extraction ability in the occluded region. This work further explores both the self-calibrated (SC) convolution module and GFocal loss, which enhance OD-UTDNet’s ability to resolve the challenging problem of high densities and occlusions. Finally, to facilitate the detection task of urban trees, a high-quality urban tree detection dataset is established, named Urban Tree Detection (UTD); to our knowledge, this is the first time. Extensive experiments show clear improvements of the proposed OD-UTDNet over 12 representative object detectors on UTD. The code and dataset are available at https://github.com/yz-wang/OD-UTDNet. [ABSTRACT FROM AUTHOR]

The visualization of an operating state of industrial processes allows operators to identify and diagnose faults intuitively and quickly. The identification and diagnosis of faults are important for ensuring industrial production safety. A method that combines variable-weighted Fisher discriminant analysis (VWFDA), t-distributed stochastic neighbor embedding (t-SNE), and multiple extreme learning machines (ELMs) is proposed for visual process monitoring. First, the VWFDA weighs variables on the basis of their contribution to the fault, thereby amplifying the fault information. The VWFDA is used to extract feature vectors from industrial data, and normal state and various fault states can be separated from each other in the space formed by these feature vectors. Second, t-SNE is used to visualize these feature vectors. Third, given that t-SNE lacks a transformation matrix during dimension reduction, one ELM is used for each class data of t-SNE to obtain the mapping relation from its input data to its mapping points. Finally, the VWFDA and multiple trained ELMs are combined for online process monitoring. The performance of the proposed approach is compared with that of FDA–t-SNE and other methods on the basis of the Tennessee Eastman process, thereby confirming that the proposed approach is advantageous for visual industrial process monitoring. • Variable-weighted FDA is proposed for feature extraction. • The model consisting of t-SNE and multiple ELMs is proposed for data visualization. • VWFDA–t-SNE is proposed for visual process monitoring. • Verifying the proposed approach by the TE process and an actual application. [ABSTRACT FROM AUTHOR]

Although the model based on an autoencoder (AE) exhibits strong feature extraction capability without data labeling, such model is less likely to consider the structural distribution of the original data and the extracted feature is uninterpretable. In this study, a new stacked sparse AE (SSAE) based on the preservation of local and global feature structures is proposed for fault detection. Two additional loss terms are included in the loss function of SSAE to retain the local and global structures of the original data. The preservation of the local feature considers the nearest neighbor of data in space, while that of the global feature considers the variance information of data. The final feature is not only a deep representation of data, but it also retains structural information as much as possible. The proposed model demonstrates remarkable detection performance in case studies of a numerical process and the Tennessee Eastman process. [ABSTRACT FROM AUTHOR]

For batch processes that are extensively applied in modern industry and characterized by nonlinearity and dynamics, quality prediction is significant to obtain high-quality products and maintain production safety. However, some quality variables and key performance indicators are difficult to measure online. In addition, the mechanism-based model for batch processes is usually tough to acquire due to the strong nonlinearity and dynamics, which makes quality prediction a challenge. With the accumulation of historical process data, data-driven methods for quality prediction gain increasing attention, among which convolutional neural network (CNN) is quite successful for its automatic feature extraction of nonlinear features from raw data. Considering that most CNN-based methods mainly take the variety of extracted features into account and ignore the redundancy between them, this paper introduces the minimal-redundancy-maximal-relevance algorithm to select features obtained by original CNN and further improves it with a feature selection layer to form the proposed method referred as mRMR-CNN. Then, a quality prediction model is established based on mRMR-CNN and the effectiveness of it is verified on the penicillin fermentation process, where the proposed method shows remarkable performance. [ABSTRACT FROM AUTHOR]

Visual process monitoring is the application of a visualization method to map the real-time operating information of an industrial process to a 2-D map, followed by process monitoring. However, owing to the complexity of industrial production processes and the complex correlations among industrial process variables, the structure and distribution of high-dimensional industrial data are very complicated. Therefore, a general visualization method cannot effectively separate the different fault data in a 2-D map for process monitoring. Accordingly, in this article, a deep double supervised embedding neural network (DDSE) is proposed for visualizing high-dimensional industrial data. The DDSE consists of two supervised deep neural networks: a deep class centres uniform distribution neural network (DCCUD), and a deep supervised t-stochastic neighbor embedding neural network (DSSNE). The DCCUD maps the high-dimensional industrial data to a new feature space in which the class centres obey a uniform distribution, promoting a good and separable situation for subsequent visualization procedures. The DSSNE then maps these high-dimensional features into a 2-D space. The training of the DDSE can be conducted through pre-training and fine-tuning. A proposed visual process monitoring approach combines the DDSE with the local outlier factor and k-nearest neighbor approaches. The proposed approach is tested on a Tennessee Eastman process, and the results illustrate that the proposed approach outperforms traditional methods in terms of visualization and visual process monitoring. [ABSTRACT FROM AUTHOR]

Process monitoring based on deep learning has attracted considerable attention. Generally, several hidden layers exist in the deep-learning model, and only the output information of the last hidden layer neurons extracted by deep learning is applied. Considering that each hidden layer is a kind of information representation of the original data, the information of different hidden layers may contain positive elements for process monitoring. In this article, we found that when a fault occurs, there are some neurons in each hidden layer that the information they output are different, compared with the normal condition. These neurons are called unstable neurons. Obviously, the information they output are beneficial for process monitoring. Motivated by theoretical analysis and experimental studies on unstable neurons, a novel method (UN-DBN) based on the unstable neurons in hidden layers is proposed to integrate the useful information for process monitoring, the Euclidean metric, the moving average filter, and the kernel density estimation technique are employed to provide an intuitionistic expression of the working state. The comparable result applied on a mathematic simulation process and the TE process with other advanced monitoring methods confirms the superiority and feasibility of the proposed method UN-DBN in this article. [ABSTRACT FROM AUTHOR]

Nonlinear process fault detection remains a challenge, with representation learning being a key step. In this article, a deep neural network (DNN)-based discriminative representation learning approach is proposed to achieve efficient fault detection for nonlinear plant-wide processes. An early-stage fault rarely affects several independent variables concurrently; hence, mutual information-based block division and randomized fault construction are performed to generate faulty validation data. By using the training data from the normal operation training data and the constructed validation data, a DNN with stacked autoencoders and a softmax classifier is trained to generate discriminative representations that maximize the capability of discriminating normal and abnormal statuses. Finally, on the basis of the learned deep discriminative representations, support vector data description is employed to discriminate the normal and abnormal process statuses. The proposed monitoring approach is tested on a numerical example and an industrial tail-gas treatment process, through which the efficiency is verified. Note to Practitioners—A modern process is generally characterized by a large scale and complex nonlinear correlation, and monitoring of such nonlinear plant-wide processes is imperative. Nowadays, a large amount of process data is generally available, and deep neural network-based monitoring is promising in dealing with such data on nonlinear processes. This article proposes a deep discriminative representation learning method for efficient nonlinear plant-wide process monitoring. The key idea is to first decompose a large-scale process into multiple units according to a variable relationship, and then generate faulty validation data based on randomized fault construction. Then, deep discriminative representations are learned by optimizing a stacked autoencoder-based deep neural network (DNN). This article provides guidelines for designing an efficient monitoring algorithm for plant-wide nonlinear processes in the industrial big data environment. [ABSTRACT FROM AUTHOR]

Deep neural network (DNN) extracts hierarchical representations from process data and is promising for nonlinear process monitoring. Obtaining meaningful representations and generating efficient fault detection residual are the main challenges in DNN-based monitoring. This study proposes a regularized deep correlated representation (RDCR) method that incorporates deep belief networks (DBNs) and canonical correlation analysis (CCA) for nonlinear process monitoring. Hierarchical representations are initially extracted using DBN to process input and output variables. Second, hierarchical representations from process input and output are modeled through CCA to characterize the relationship between them. Efficient fault detection residuals are then generated, and monitoring statistics are established. CCA-based monitoring relies on the most correlated representations; thus, a multiobjective evolutionary optimization-based regularization is performed to select the most correlated representations and eliminate the influence of unrelated representations. The advantages of the RDCR monitoring are verified through experimental studies on a numerical example and the Tennessee Eastman process. [ABSTRACT FROM AUTHOR]

Conventional methods for distributed monitoring commonly assume that complete process measurements are available. However, the problem of missing data is often encountered in the monitoring of large-scale multiunit processes. This paper proposes an approach based on a neighborhood variational Bayesian principal component analysis (NVBPCA) and canonical correlation analysis (CCA) for the efficient distributed monitoring of multiunit processes in the presence of missing data. Missing observations for a local unit are reconstructed through NVBPCA by considering information from both local and neighboring units. A CCA-based local monitor, which identifies the status of the local unit and the type of a detected fault using information from both the local and neighboring units, is then developed. The NVBPCA–CCA approach has a better performance since its missing data handling and local monitor construction consider information from both the local and neighboring units. The efficiency of the proposed monitoring method is demonstrated through its application in a numerical example and an industrial tail gas treatment process. [ABSTRACT FROM AUTHOR]

In complex and volatile unknown flight environments, the limited environmental information obtained by sensors in the face of sudden dynamic and static obstacles makes it extremely challenging for unmanned aerial vehicles (UAVs) to obtain a safe and efficient path to avoid obstacles and reach a designated target point. Therefore, a real-time dynamic path planning method based on an improved interfered fluid dynamical system (IFDS) and artificial neural network (ANN) is proposed to enhance path quality and computational efficiency. Firstly, to address the issue of insufficient sample quality and quantity, IFDS is employed as the fundamental method for path planning to simulate and generate an adequate amount of sample data for the ANN training. Then, an enhanced sand cat swarm optimization algorithm (ESCSO) with an adaptive social neighborhood search mechanism and Lévy flight strategy is proposed to improve the sample quality. Secondly, the information between the UAV and the target points and obstacles is extracted from the sample data as the input for the network, the parameters of the IFDS are used as the feature extraction at the output of the network, and the ESCSO is applied to optimize the weights and biases of the ANN, enabling offline training of the neural network. Finally, the trained neural network is utilized to dynamically output IFDS parameters based on the real-time environmental information obtained from the sensors, enabling the generation of real-time obstacle avoidance paths. Experimental results in a series of complex simulated environments demonstrate that the proposed method outperforms other algorithms in terms of path quality and meets real-time requirements. It provides excellent obstacle avoidance characteristics for the UAV. [ABSTRACT FROM AUTHOR]

Abstract Recently, based on the powerful capability of feature extraction, deep learning technique has been applied to the field of process monitoring, and usually, the researches utilize all the abstract features to establish the detection model and detect or classify the fault. However, whether all the extracted features are valid and beneficial for process monitoring have never been researched and discussed. If there are some features that are adverse for process monitoring, the detection performance of the model would be reduced once they are considered in the model, and utilized the features that are advantageous for process monitoring could ameliorate the performance of detection model. Motivated by this, a feasibility analysis on each feature captured by deep belief network for process monitoring is executed and the conception of active features (AFs) which have active expression for the occurrence of the fault is proposed. Based on AFs, utilized Euclidean metric to calculate the dissimilarity between the test sample and the training sample, and moving average technique is employed to reduce the effect of the burst noise in measurement variables on the result. Finally, the comparison of fault detection rate with other advanced methods on a numerical process and TE process demonstrate the feasibility and superiority of the proposed method, AF-DBN in this study. Highlights • Utilized the powerful feature extraction capability of deep learning to perform industrial process monitoring. • Perform the analysis of all the features extracted by DBN and propose the conception of "activity degree" and "active features". • Based on active features to perform the process monitoring and eliminate the adverse effects of inactive features on fault detection. • Avoid the adverse influence on monitoring results of the burst noise in measurement variables. • Different monitoring performance influenced by parameters of DBN are demonstrated in detail. [ABSTRACT FROM AUTHOR]

Multivariate statistical process monitoring involves dimension reduction and latent feature extraction in large-scale processes and typically incorporates all measured variables. However, involving variables without beneficial information may degrade monitoring performance. This study analyzes the effect of variable selection on principal component analysis (PCA) monitoring performance. Then, it proposes a fault-relevant variable selection and Bayesian inference-based distributed method for efficient fault detection and isolation. First, the optimal subset of variables is identified for each fault using an optimization algorithm. Second, a sub-PCA model is established in each subset. Finally, the monitoring results of all of the subsets are combined through Bayesian inference. The proposed method reduces redundancy and complexity, explores numerous local behaviors, and provides accurate description of faults, thus improving monitoring performance significantly. Case studies on a numerical example, the Tennessee Eastman benchmark process, and an industrial-scale plant demonstrate the efficiency. [ABSTRACT FROM AUTHOR]

Abstract: Considering that the global and local structures of process data would probably be changed in some abnormal states, a multi-manifold projection (MMP) algorithm for process monitoring and fault diagnosis is proposed under the graph embedded learning framework. To exploit the underlying geometrical structure that contains both global and local information of sampled data, the global graph and local graph are designed to characterize the global and local structures, respectively. A unified optimization framework, i.e. global graph maximum and local graph minimum, is then constructed to extract meaningful low-dimensional representations for high-dimensional process data. In the proposed MMP, the neighborhood embedding is used in both global and local graphs and the extracted features are faithful representations of the original data. The feasibility and validity of the MMP-based process monitoring scheme are investigated through two case studies: a simple simulation process and the Tennessee Eastman process. The experimental results demonstrate that the whole performance of MMP is better than those of some traditional preserving global or local or global and local feature methods. [Copyright &y& Elsevier]

• A stacked reinforced discriminant autoencoder is proposed for feature extraction. • The proposed stacked autoencoder and MRMR are combined for feature selection. • A stacked supervised t-SNE is proposed for data visualization. • A new visualization-based process monitoring method is introduced. Visual process monitoring is to monitor industrial processes by projecting the high-dimensional process data into the two-dimensional space, which provides powerful insight for industrial processes, and accelerates fault diagnosis. The challenge of visual process monitoring lies in how to project the complex process data into the two-dimensional plane and separate different classes as much as possible. In this paper, a new visual process monitoring method is proposed. First, a stacked reinforced discriminant auto-encoder (SRDAE) which consists of multiple reinforced discriminant auto-encoders (RDAEs) is proposed to extract discriminant features. In SRDAE, the useful features in the original data and the hidden output of the previous RDAE are combined together as the input of the latter RDAE, and the error of class label is added into the loss function of RDAE. Therefore, SRDAE can prevent the loss of useful information in the original data in the high layers and make the extracted features have the powerful ability to separate different classes. Furthermore, in order to extract the more informative discriminant features, minimal redundancy maximal relevance (MRMR) technology is utilized to select important neurons from all layers of the SRDAE as the final feature representation of the original data. Finally, a stacked supervised t-distributed stochastic neighbor embedding network is proposed to visualize the discriminant features for process monitoring. The effectiveness of the proposed method is validated on the Tennessee Eastman process, the experiments show that the proposed method can effectively separate different classes to achieve intuitive and efficient process monitoring. [ABSTRACT FROM AUTHOR]

Current fault detection methods based on deep neural networks only consider process information and ignore quality indicators. In order to obtain features representing both process variables and quality indicators efficiently, this paper designs teacher and supervise dual stacked auto-encoder (TSSAE) for quality-relevant fault detection in industrial process which separates the feature extraction and model construction. To separate the feature extraction and model construction, a mixing stacked auto-encoder which consists of a nonlinear encoder and a linear decoder is designed to extract features of process variables and quality indicators. Another encoder is supervised by the extracted features and further predict the process variables and quality indicators only from process variables. Then quality-relevant, quality-irrelevant and residual subspaces are constructed in a linear way and fault detection is implemented in these subspaces based on Euclidean distance and kernel density estimation. Finally, the effectiveness of TSSAE is evaluated by a numerical example and the Tennessee-Eastman process. • Quality variables are considered in TSSAE. • Feature extraction and model construction are implemented separately. • Features are compact without much redundancy based on LWPD. [ABSTRACT FROM AUTHOR]