No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Interpretable Switching Deep Markov Model for Industrial Process Monitoring via Cloud-Edge Collaborative Framework
Abstract
In modern manufacturing, process monitoring is crucial in ensuring the stability and safety of production processes. However, the frequent changes in industrial conditions necessitate timely updates and retraining of the on-site deployment of data-driven process monitoring methods, which is a task unattainable with the limited computational resources of edge devices. To address this issue, this paper proposes an interpretable switching deep Markov model (ISDMM)-based cloud-edge collaborative framework for industrial process monitoring. ISDMM defines discrete switching variables that represent working conditions and corresponding multiple transition networks. The transition networks are concurrently trained and switched based on the current working condition, enabling ISDMM to capture different dynamics of the system under different conditions. Before deployment, model simplification is performed to construct a model base where each model is tailored to individual working conditions. Moreover, in the edge layer, a model update strategy is designed to determine whether the model can be invoked from the model library based on the current working condition, alleviating the burden of unnecessary retraining. Finally, the effectiveness of ISDMM and its cloud-edge collaborative framework is validated through a numerical example and a real-world industrial process.
ResearchGate has not been able to resolve any citations for this publication.
Missing data is a common phenomenon in the industrial field. The recovery of missing data is crucial to enhance the reliability of subsequent data-driven monitoring and control of industrial processes. Most existing methods are limited by the confined scope of feature extraction, which makes it impossible to rely on global information to impute missing data. In addition, they usually assume that industrial data is a uniform distribution across all working conditions, ignoring the differences in data evolution patterns across different conditions. To address these issues, this paper proposes an innovative scope-free global multi-condition-aware imputation framework based on diffusion transformer (SGMCAI-DiT). First, it extends the diffusion model by introducing conditional probability to capture the condition distribution of the entire data. Then, a noise prediction model is designed based on a novel double-weighted attention mechanism (DW-SA) to broaden the horizons of feature extraction. By discerning the inter-conditional interactions and the intra-conditional local information, the missing data imputation performance can be improved. Finally, the effectiveness and suitability of the proposed SGMCAI-DiT are verified on four real datasets sourced from industrial processes and two public non-industrial datasets. Extensive experimental results demonstrate that the proposed method outperforms several state-of-the-art methods in different missing data scenarios.
Internet of Things (IoT) is made up with growing number of facilities, which are digitalized to have sensing, networking and computing capabilities. Traditionally, the large volume of data generated by the IoT devices are processed in a centralized cloud computing model. However, it is no longer able to meet the computational demands of large-scale and geographically distributed IoT devices for executing tasks of high performance, low latency, and low energy consumption. Therefore, edge computing has emerged as a complement of cloud computing. To improve system performance, it is necessary to partition and offload some tasks generated by local devices to the remote cloud or edge nodes. However, most of the current research work focuses on designing efficient offloading strategies and service orchestration. Little attention has been paid to the problem of jointly optimizing task partitioning and offloading for different application types. In this paper, we make a comprehensive overview on the existing task partitioning and offloading frameworks, focusing on the input and core of decision engine of the framework for task partitioning and offloading. We also propose comprehensive taxonomy metrics for comparing task partitioning and offloading approaches in the IoT cloud-edge collaborative computing framework. Finally, we discuss the problems and challenges that may be encountered in the future.
Proper monitoring of quality-related variables in industrial processes is nowadays one of the main worldwide challenges with significant safety and efficiency implications. Variational Bayesian mixture of Canonical correlation analysis (VBMCCA)-based process monitoring method was proposed in this paper to predict and diagnose these hard-to-measure quality-related variables simultaneously. Use of Student's t-distribution, rather than Gaussian distribution, in the VBMCCA model makes the proposed process monitoring scheme insensitive to disturbances, measurement noises and model discrepancies. A sequential perturbation method together with derived parameter distribution of VBMCCA is employed to approach the uncertainty levels, which is able to provide confidence interval around the predicted values and give additional control line, rather than just a certain absolute control limit, for process monitoring. The proposed process monitoring framework has been validated in a Wastewater Treatment Plant (WWTP) simulated by Benchmark Simulation Model (BSM) with abrupt changes imposing on a sensor and a real WWTP with filamentous sludge bulking. The results show that the proposed methodology is capable of detecting sensor faults and process faults with satisfactory accuracy.
Gaussian state space models have been used for decades as generative models of sequential data. They admit an intuitive probabilistic interpretation, have a simple functional form, and enjoy widespread adoption. We introduce a unified algorithm to efficiently learn a broad class of linear and non-linear state space models, including variants where the emission and transition distributions are modeled by deep neural networks. Our learning algorithm simultaneously learns a compiled inference network and the generative model, leveraging a structured variational approximation parameterized by recurrent neural networks to mimic the posterior distribution. We apply the learning algorithm to both synthetic and real-world datasets, demonstrating its scalability and versatility. We find that using the structured approximation to the posterior results in models with significantly higher held-out likelihood.
High-dimensional and time-dependent data pose significant challenges to Statistical Process Monitoring (SPM). Most of the high-dimensional methodologies to cope with these challenges rely on some form of Principal Component Analysis (PCA) model, usually classified as non-adaptive and adaptive. Non-adaptive methods include the static PCA approach and Dynamic PCA for data with autocorrelation. Methods, such as Dynamic PCA with Decorrelated Residuals, extend Dynamic PCA to further reduce the effects of autocorrelation and cross-correlation on the monitoring statistics. Recursive PCA and Moving Window PCA, developed for non-stationary data, are adaptive. These fundamental methods will be systematically compared on high-dimensional, time-dependent processes (including the Tennessee Eastman benchmark process) to provide practitioners with guidelines for appropriate monitoring strategies and a sense of how they can be expected to perform. The selection of parameter values for the different methods is also discussed. Finally, the relevant challenges of modeling time-dependent data are discussed, and areas of possible further research are highlighted. This article is protected by copyright. All rights reserved.
Data-driven modeling of industrial processes is fundamental for predictive maintenance of key equipment and optimization of process operations, playing a crucial role in ensuring high quality production. However, the harsh environments of industrial settings often result in process data being heavily contaminated with noise, significantly impairing the accuracy of modeling. Unfortunately, existing denoising methods are predominantly developed for scenarios with well-defined models. This renders them ineffective for complex industrial processes. To address this problem, this paper proposes a deep Koopman Kalman filter (DKKF) for nonlinear model-free industrial process data denoising. First, DKKF defines a discrete dynamic system capable of conducting explicit Kalman filtering. This discrete system models the dynamics via the Koopman operator. Transition noise represents the approximation error of the Koopman operator, while observation noise is directly estimated from the observations. Then, the explicit implementation of the Kalman filter by the DKKF, both during training and inference phases, significantly enhances the accuracy and interpretability of the model. Moreover, a three-stage training strategy is defined for DKKF, which effectively addresses issues of slow training speed and gradient explosion. Finally, the effectiveness of the proposed method is validated on the Lorenz system and a real-world industrial process dataset. Notably, the denoised industrial process data is used for soft sensor modeling, resulting in more accurate prediction results compared to some existing representative methods.
With the rapid development of cloud computing, edge computing, and deep learning technologies, the implementation of soft sensor modeling within a cloud-edge collaboration architecture shows great potential and plays a pivotal role in achieving consistent and highly accuracy for industrial processes. However, traditional soft sensor models encounter significant challenges such as difficulties in the extraction of spatial correlation features, updating, and model mismatch, etc. For this reason, this paper proposes a soft sensor modeling method for adaptive quality prediction based on cloud-edge collaboration. The method first designs a basic model for the cloud and edge, which is variable reordering based convolutional neural network (VRCNN) for non-neighboring features learning. Then, it deploys deep training model and shallow fine-tuning prediction model in the cloud and edge, respectively. After that, the adaptive cloud-edge collaborative modeling is achieved through parameter collaboration, data collaboration and inference collaboration between cloud and edge models. The proposed method is validated through comparative experiments on two tasks of the industrial hydrocracking process.
The real-time recognition of operating conditions is always critical to ensuring the efficient and stable operation of industrial flotation processes. Although the widespread use of smart devices enables the availability of multimodal data in flotation processes, recognizing operating conditions using cross-modal data information is still challenging due to the modality gap. To address this issue, this article first proposes an innovative bimodal manifold autoencoder model to predict interested quality variables from the perspective of visual modality and auditory modality. Specifically, the well-designed intra- and intermodal manifold regularization constraints are introduced to fully learn the intrinsic manifold features within each modality and the interdependencies across modalities, thereby enhancing the cross-modal data representation ability of the developed prediction models. Then, based on the prediction values of quality variables, an adaptable multimodal fuzzy decision inference module is designed to recognize the operating conditions while counteracting the influence of fluctuations in feedstock properties. Finally, extensive experiments are conducted on two industrial flotation process datasets at distinct periods to validate the superiority of the proposed methods in terms of quality prediction and operation condition recognition tasks.
Nowadays, with the rapidly growing size of random orders and machine scales, cloud manufacturing (CMfg) is suffering from complex difficulties in scheduling enormous production services in real time. Especially when coupled with preventive maintenance (PM) scheduling for massive distributed machines, most current methods fail to promise agility and effectiveness simultaneously under this service-oriented paradigm. Therefore, this article aims to propose a collaborative scheduling algorithm to support real-time production and opportunistic maintenance (OM) in the highly dynamic CMfg environment. First, a novel collaborative scheduling model that combines real-time production and PM decision-making is formulated. Then, to address significant complexity in production scheduling, a dynamic suborder-oriented scheduling (DSS) strategy is designed to utilize frequent triggered events for allocating optimal manufacturing services in real time. Then, for improving computational capability in PM decision-making, a production-driven PM policy with two stage adjustment (TSA) is developed to adopt suborder changeovers as PM opportunities to formulate group PM decisions for massive machines. Finally, a collaborative DSS-TSA algorithm is designed to cyclically coordinate the real-time production scheduling with OM decision-making to avoid repetitively solving large-scale joint problems. Our collaborative scheduling algorithm is verified in a real CMfg scenario for gas turbine blades. The results prove the significant improvement in production timeliness, machine availability, and system profitability.
Frequent switching of operating conditions in industrial processes tends to make the data distribution time varying and variable correlations nonuniform, which brings considerable challenges for explicit representation and monitoring of nonstationary processes. This study addresses the abovementioned problem based on the following recognitions: 1) despite changes in operating conditions, there exist time-invariant process mechanisms, which are nonlinearly coupled with nonstationarity; 2) such a coupling relationship is not uniform but may show diverse modes changing with time, defined as multimodal coupling here; 3) diverse coupling relations can be derived from the superposition of nonstationarity induced by definite driving forces, i.e., changeable operating condition settings, reflecting their intrinsic association under complex changes. Thereupon, a cascaded deep information separation (CDIS) architecture with a compatibility learning algorithm is proposed to extract multimodal decoupled representations. We design an information refinement module (IRM) to capture the basic coupling source (BCS) under the influence of driving forces, where a shortcut connection is incorporated into the nonlinear autoencoding structure. Multiple IRMs can be flexibly cascaded to achieve the superposition of BCSs, thus, portraying diverse multimodal couplings. Furthermore, by balancing the stationarity requirement and reconstruction constraint, the designed compatibility learning algorithm induces the cascaded IRMs to capture and filter out nonstationarity, thus, obtaining stationary refined data. In this way, stationarity and nonstationarity with multimodal couplings can be fully separated. The validity of CDIS is illustrated through a simulated example and a real condensing system experimental rig.
In existing research on rotating machinery diagnosis using graph neural networks, most methods are based on vibration analysis under contact sensor monitoring. However, the gathered vibration signal is sensitive to the increased mass of the sensor, and after prolonged close contact, the vibration sensor may cause structural damage to the object. A multi-sensor fusion based on modal analysis and graph attention network (MFMAGAT) for bearing fault diagnosis is suggested as a solution to these issues. First, we proposed a phase-based full-field non-contact measurement method based on direction-controlled pyramids to extract phase information at different scales and directions and use the phase information to characterize the vibration of the structure. Then, based on the vibration and acoustic signals of the characterized structure, we constructed a kernel function that can fuse the heterogeneous information, and its singular value decomposition can jointly characterize the response of the structure at same frequency bands from the video data and the acoustic data. Finally, considering that the decomposed features are independent and orthogonal to each other, the graph attention network is introduced at this time to find the weights among the components to jointly characterize the structural damage and improve the damage recognition accuracy. Comparison results show that the proposed method is effective and performs better than other conventional graph neural network diagnosis methods.
This article studies the performance monitoring problem for the potassium chloride flotation process, which is a critical component of potassium fertilizer processing. To address its froth image segmentation problem, this article proposes a multiscale feature extraction and fusion network (MsFEFNet) to overcome the multiscale and weak edge characteristics of potassium chloride flotation froth images. MsFEFNet performs simultaneous feature extraction at multiple image scales and automatically learns spatial information of interest at each scale to achieve efficient multiscale information fusion. In addition, the potassium chloride flotation process is a multistage dynamic process with massive unlabeled data. To overcome its dynamic time-varying and working condition spatial similarity characteristics, a semi-supervised froth-grade prediction model based on a temporal-spatial neighborhood learning network combined with Mean Teacher (MT-TSNLNet) is proposed. MT-TSNLNet designs a new objective function for learning the temporal-spatial neighborhood structure of data. The introduction of Mean Teacher can further utilize unlabeled data to promote the proposed prediction model to better track the concentrate grade. To verify the effectiveness of the proposed MsFEFNet and MT-TSNLNet, froth image segmentation and grade prediction experiments are performed on a real-world potassium chloride flotation process dataset.
The use of traditional methods in anomaly detection of multi-class ship trajectories showed some limitations in terms of robustness and learning ability of trajectory features. In view of this, an anomaly detection model for ship trajectory data based on Gaussian Mixture Variational Autoencoder (GMVAE) is proposed in this study using an unsupervised classification method. The proposed model modifies Variational Autoencoder (VAE) by changing the inferential distribution of prior distribution and approximate posterior to the Gaussian mixture model. A high-dimensional hidden space is constructed to learn the features of multi-class trajectory data, and the Dynamic Time Warping (DTW) method is applied to measure the error between the reconstructed trajectory and the original trajectory in order to judge whether the ship trajectory is abnormal. The Automatic Identification System (AIS) data from the US coastal areas are used to verify the proposed model, and the results are compared with other commonly used models in a manually labeled dataset. The research results indicate that the detection rate of the proposed model is 91.26%, and the false alarm rate is 0.68%, which performs the best. Using the Gaussian mixture model to describe the distribution of hidden space can improve the learning ability of multi-class trajectories of VAE, thus increasing the robustness of the model. This research can provide technical support for ship trajectory data analysis and risk management of maritime transportation.
Latent variable-based process monitoring (PM) models have been generously developed by shallow learning approaches, such as multivariate statistical analysis and kernel techniques. Owing to their explicit projection objectives, the extracted latent variables are usually meaningful and easily interpretable in mathematical terms. Recently, deep learning (DL) has been introduced to PM and has exhibited excellent performance because of its powerful presentation capability. However, its complex nonlinearity prevents it from being interpreted as human-friendly. It is a mystery how to design a proper network structure to achieve satisfactory PM performance for DL-based latent variable models (LVMs). In this article, a variational autoencoder-based interpretable LVM (VAE-ILVM) is developed for PM. Based on Taylor expansions, two propositions are proposed to guide the design of appropriate activation functions for VAE-ILVM, allowing nondisappearing fault impact terms contained in the generated monitoring metrics (MMs). During threshold learning, the sequence of counting that test statistics exceed the threshold is considered a martingale, a representative of weakly dependent stochastic processes. A de la Peña inequality is then adopted to learn a suitable threshold. Finally, two chemical examples verify the effectiveness of the proposed method. The use of de la Peña inequality significantly reduces the minimum required sample size for modeling.
Recently, latent models have continued to find advantages in statistical process monitoring, especially with multifarious extensions coping with the nonlinearity and dynamics confronted by the pattern extraction. However, when later performing the famous 𝑇 2 statistic on the patterns, the issue of non-Gaussianity hasrarely been focused alongside, hence vitiating the control of Type-I and -II error in practice. In this paper, amodel, in the name of recurrent Kalman variational auto-encoder (RKVAE), is theoretically derived for handling these properties simultaneously, and proposed for its use on monitoring general complex processes. The mainidea is to build a monitoring panel on the basis of further reasoning the dynamics underlying the nonlinearly extracted Gaussian features. The model hierarchically consists of three parts: the vanilla VAE for encoding latent Gaussian features, a linear Gaussian state-space model (LGSSM) for temporally reasoning these features,and an RNN for recurrently updating the online LGSSM by fusing several candidate parameters. For the latterK in the name, it is the Kalman filter (KF) that facilitates the feature noise filtering and LGSSM identification.The model has, structurally, a good capacity to be extensible from both the depth and width when modelling complex processes. Finally, based on the explicit formula of LGSSM, a detection panel is designed, and two case studies are given to demonstrate its effectiveness in monitoring and relevant analysis.
The effective fault diagnosis is an important guarantee for the safe and stable operation of mechanical system. Nowadays, the most of fault diagnosis methods are for known classes. However, unknown fault usually occurs in the process of mechanical operation, and may be wrongly divided into defined classes, which increases the risk of mechanical shutdown and the difficulty of maintenance. Therefore, in order to achieve high-accuracy diagnosis with unknown fault, a method based on transfer learning and deep transfer clustering is proposed, named Transfer Clustering Calculation Center Points (TCCP). TCCP completes the task of fault diagnosis with unknown class by calculating the feature space distribution between sample to be tested and the known samples, and further improves the accuracy by matching the target distribution. In two open-source bearing vibration datasets and an acoustic emission dataset of the self-made bearing, five bearing diagnostic models are used to compare the unknown failure diagnosis. The average accuracy with unknown failure by TCCP is 97.08%, 89.78%, and 95.24%, respectively. The accuracy is 24.45%, 23.16% and 10.08% higher than the average accuracy without TCCP. TCCP can get better diagnostic results under both vibration signals and audio signals. TCCP has predominant universality and generalization capabilities, and provides an effective approach for the identification of unknown failure.
Detecting out-of-distribution (OOD) data is a task that is receiving an increasing amount of research attention in the domain of deep learning for computer vision. However, the performance of detection methods is generally evaluated on the task in isolation, rather than also considering potential downstream tasks in tandem. In this work, we examine selective classification in the presence of OOD data (SCOD). That is to say, the motivation for detecting OOD samples is to reject them so their impact on the quality of predictions is reduced. We show under this task specification, that existing post-hoc methods perform quite differently compared to when evaluated only on OOD detection. This is because it is no longer an issue to conflate in-distribution (ID) data with OOD data if the ID data is going to be misclassified. However, the conflation within ID data of correct and incorrect predictions becomes undesirable. We also propose a novel method for SCOD, Softmax Information Retaining Combination (SIRC), that augments softmax-based confidence scores with feature-agnostic information such that their ability to identify OOD samples is improved without sacrificing separation between correct and incorrect ID predictions. Experiments on a wide variety of ImageNet-scale datasets and convolutional neural network architectures show that SIRC is able to consistently match or outperform the baseline for SCOD, whilst existing OOD detection methods fail to do so. Code is available at https://github.com/Guoxoug/SIRC.
Nowadays, the intelligent fault diagnosis problem of modern industrial systems has received increasing attention. However, with the increasing scale of industrial systems, the same category of faulty data often contains multiple working conditions, which leads to the multi-granularity of faulty data and the increasing complexity of data distribution. The possible existence of fine-granularity inter-class similarities in multi-granularity data can interfere with coarse-granularity recognition tasks, which brings challenges to data-driven fault diagnosis tasks. To solve the above issue, a fault diagnosis method based on multi-granularity dictionary learning is proposed in this paper. Specifically, the proposed method integrates prior knowledge into the dictionary learning framework by introducing the concept of “centroid atom”, which embeds the multi-granularity structure in the dictionary and improves the discrimination of dictionary atoms. In order to validate the proposed approach, experiments are conducted on a numerical simulation case and a zinc smelting roasting process. Compared with some state-of-the-art methods, the proposed approach performs better in the classification accuracy which shows that it can make full use of the knowledge provided by the multi-granularity structure while it is difficult to be modeled by the other methods. Therefore, it can be applied to fault diagnosis tasks more precisely and efficiently.
Note to Practitioners
—Motivated by the fact that industrial faults often occur in different modes, and different faults may occur in the same mode, the fault data often has multi-granularity characteristics, this paper proposes a multi-granularity dictionary learning method for complex industrial fault diagnosis. The proposed method can learn the multi-granularity structure in the data, and the obtained model contains both coarse-grained and fine-grained information, which effectively improves the fault diagnosis accuracy. Intensive experimental results show that the proposed method performs better than some state-of-the-art methods, it can make full use of the knowledge provided by the multi-granularity structure, while it is difficult to be modeled by others. Above all, it is verified as suitable for fault diagnosis of real industrial systems.
We introduce deep switching auto-regressive factorization (DSARF), a deep generative model for spatio-temporal data with the capability to unravel recurring patterns in the data and perform robust short- and long-term predictions. Similar to other factor analysis methods, DSARF approximates high dimensional data by a product between time dependent weights and spatially dependent factors. These weights and factors are in turn represented in terms of lower dimensional latent variables that are inferred using stochastic variational inference. DSARF is different from the state-of-the-art techniques in that it parameterizes the weights in terms of a deep switching vector auto-regressive likelihood governed with a Markovian prior, which is able to capture the non-linear inter-dependencies among weights to characterize multimodal temporal dynamics. This results in a flexible hierarchical deep generative factor analysis model that can be extended to (i) provide a collection of potentially interpretable states abstracted from the process dynamics, and (ii) perform short- and long-term vector time series prediction in a complex multi-relational setting. Our extensive experiments, which include simulated data and real data from a wide range of applications such as climate change, weather forecasting, traffic, infectious disease spread and nonlinear physical systems attest the superior performance of DSARF in terms of long- and short-term prediction error, when compared with the state-of-the-art methods.
In cloud manufacturing systems, fault diagnosis is essential for ensuring stable manufacturing processes. The most crucial performance indicators of fault diagnosis models are generalization and accuracy. An urgent problem is the lack and imbalance of fault data. To address this issue, in this article, most of existing approaches demand the label of faults as
a priori
knowledge and require extensive target fault data. These approaches may also ignore the heterogeneity of various equipment. We propose a cloud-edge collaborative method for adaptive fault diagnosis with label sampling space enlarging, named label-split multiple-inputs convolutional neural network, in cloud manufacturing. First, a multiattribute cooperative representation-based fault label sampling space enlarging approach is proposed to extend the variety of diagnosable faults. Besides, a multi-input multi-output data augmentation method with label-coupling weighted sampling is developed. In addition, a cloud-edge collaborative adaptation approach for fault diagnosis for scene-specific equipment in cloud manufacturing system is proposed. Experiments demonstrate the effectiveness and accuracy of our method.
The development of cloud manufacturing enables data-driven process monitoring methods to reflect the real industrial process states accurately and timely. However, traditional process monitoring methods cannot update learned models once they are deployed to edge devices, which leads to model mismatch when confronted time-varying data. In addition, limited resources on the edge prevent it from deploying complex models. Therefore, this article proposes a novel cloud-edge collaborative process monitoring method. First, historical data of industrial processes are collected to establish a dictionary learning model and train the dictionary and classifier in the cloud. Then, the model is simplified and deployed to the edge. The edge layer monitors the process states, including fault detection and working condition recognition, and determines whether a model mismatch has occurred based on an error-triggered strategy. Both numerical simulation and industrial roasting process results verify the superiority of the proposed method.
With the continuous addition of an abundant of heterogeneous devices, the limitation of task delay has become an obstacle to the development of the Industrial Internet of Things(IIoT). Task offloading based on edge computing can provide low-latency computing services for these tasks. However, in the actual IIoT scenario, contrast to cloud computing, edge computing has limited resources and computing capabilities. Resource-constrained edge resources can’t meet the offloading requirements of massive industrial devices. In this paper, we propose an optimal joint offloading scheme based on resource occupancy prediction for the problem of computing offloading with limited edge resources. The scheme is divided into two parts including edge resource occupancy prediction and task offloading. Simultaneously, considering multitask and the limitations of edge resources, gate recurrent unit is used to predict the occupancy of edge resources. Formulating an optimal strategy of task offloading by using a reinforcement learning algorithm according to the network state and predicted results. The simulation results show that the scheme can effectively reduce the average delay of tasks, while minimizing the task offloading failure rate.
In this paper, a cloud-edge collaboration based control framework is proposed for the voltage regulation and economic operation in incremental distribution networks (IDN). The voltage regulation and economic operation, usually considered in separated aspects, can be integrated in a hierarchical control method by coordinating the active power and reactive power of distributed generators (DGs) and distributed storages (DSs) in an ‘'active’' mode. Promising the voltage security of the IDN, the upper-level multi-objective optimization is formulated to maximize the consumption of the DGs, moreover, the lower-level model predictive control (MPC) aims to regulate the dynamics of the DGs and DSs based on the established state space model. Time delay in the downstream channel is considered due to the open environment of the proposed control framework, which can be eliminated by using the predictive compensation mechanism derived from the MPC with considering model uncertainty. Finally, simulation results demonstrate the validity and robustness of the proposed method.
Deep learning has recently emerged as a promising method for nonlinear process monitoring. However, ensuring that the features from process variables have representative information of the high-dimensional process data remains a challenge. In this study, we propose an adversarial autoencoder(AAE) based process monitoring system. AAE which combines the advantages of a variational autoencoder and a generative adversarial network enable the generation of features that follow the designed prior distribution. By employing the AAE model, features that have informative manifolds of the original data are obtained. These features are used for constructing monitoring statistics and improve the stability and reliability of fault detection. Extracted features help calculate the degree of abnormalities in process variables more robustly and indicate the type of fault information they imply. Finally, our proposed method is testified using the Tennessee Eastman benchmark process in terms of fault detection rate, false alarm rate, and fault detection delays.
Fault diagnosis is essential to ensure the operation security and economic efficiency of the chemical system. Many fault diagnosis methods have been designed for the chemical process, but most of them ignore the temporal correlation in the sequential observation signals of the chemical process. A novel deep learning method based on Stacked Long Short-Term Memory (LSTM) neural network is proposed, which can effectively model sequential data and detect the abnormal values. The proposed method is also able to fully exploit the long-term dependencies information in raw data and adaptively extract the representative features. The dataset of Tennessee Eastman (TE) process is utilized to verify the practicability and superiority of the proposed method. Extensive experimental results show that the fault detection and diagnosis model we proposed has an excellent performance when compared with several state-of-the-art baseline methods.
Recently, deep neural network (DNN) models work incredibly well, and edge computing has achieved great success in real-world scenarios, such as fault diagnosis for large-scale rotational machinery. However, DNN training takes a long time due to its complex calculation, which makes it difficult to optimize and retrain models. To address such an issue, this work proposes a novel fault diagnosis model by combining binarized DNNs (BDNNs) with improved random forests (RFs). First, a BDNN-based feature extraction method with binary weights and activations in a training process is designed to reduce the model runtime without losing the accuracy of feature extraction. Its generated features are used to train an RF-based fault classifier to relieve the information loss caused by binarization. Second, considering the possible classification accuracy reduction resulting from those very similar binarized features of two instances with different classes, we replace a Gini index with ReliefF as the attribute evaluation measure in training RFs to further enhance the separability of fault features extracted by BDNN and accordingly improve the fault identification accuracy. Third, an edge computing-based fault diagnosis mode is proposed to increase diagnostic efficiency, where our diagnosis model is deployed distributedly on a number of edge nodes close to the end rotational machines in distinct locations. Extensive experiments are conducted to validate the proposed method on the data sets from rolling element bearings, and the results demonstrate that, in almost all cases, its diagnostic accuracy is competitive to the state-of-the-art DNNs and even higher due to a form of regularization in some cases. Benefited from the relatively lower computing and storage requirements of BDNNs, it is easy to be deployed on edge nodes to realize real-time fault diagnosis concurrently.
This paper is primarily concerned with finite-time and predefined-time convergence design for a class of general Zeroing Neural Network (ZNN) by constructing different activation functions (AFs). Based on the limit comparison test for improper integrals, some useful theorical criteria are proposed to determine whether a nonlinear-activated ZNN model has finite-time convergence (FTC), predefined-time convergence (PTC), or not. This novel method can avoid the valuation-loss of the zoom method and the unsolvable barrier of the direct integration method that are widely used in the previous ZNN design. According to these convergence criteria, some instructive corollaries are derived to design valuable AFs to make ZNN models with FTC or PTC more easily. By taking a matrix-inversion ZNN model, some commonly-used AFs are used to verify the usability of the criterion. In addition, some new AFs are constructed to further design some better ZNN models with superior FTC or PTC. Finally, convergence types of the ZNN model based on different AFs are visualized in numerical experiments.
Prior work on deep learning models-based mobile applications in a cloud-edge computing environment focuses on performing lightweight data pre-processing tasks on edge servers for cloud-hosted cognitive servers. Those approaches have two major drawbacks. First, it is uneasy for the mobile applications to assure satisfactory user experience in terms of network communication delay, because the intermediary edge servers are used only to pre-process data (e.g., images and videos) and the cloud servers must be used to complete the required data processing tasks. Second, those approaches assume the pre-trained deep learning models deployed on cloud servers do not change, and do not attempt to automatically upgrade in a context-aware manner via the collected data. In this paper, we propose a cloud-edge collaboration framework that facilitates delivering mobile applications with long duration, fast response, and high accuracy properties. In the proposed framework, the edge server can provide cognitive service with long duration and fast response properties via a shallow convolutional neural network model, called EdgeCNN. Our experimental results show that EdgeCNN can reduce the average response time of cognitive services and improve accuracy.
Many natural systems, such as neurons firing in the brain or basketball teams traversing a court, give rise to time series data with complex, nonlinear dynamics. We can gain insight into these systems by decomposing the data into segments that are each explained by simpler dynamic units. Building on switching linear dynamical systems (SLDS), we present a new model class that not only discovers these dynamical units, but also explains how their switching behavior depends on observations or continuous latent states. These "recurrent" switching linear dynamical systems provide further insight by discovering the conditions under which each unit is deployed, something that traditional SLDS models fail to do. We leverage recent algorithmic advances in approximate inference to make Bayesian inference in these models easy, fast, and scalable.
This paper is concerned with developing a distributed k-means algorithm and a distributed fuzzy c-means algorithm for wireless sensor networks (WSNs) where each node is equipped with sensors. The underlying topology of the WSN is supposed to be strongly connected. The consensus algorithm in multiagent consensus theory is utilized to exchange the measurement information of the sensors in WSN. To obtain a faster convergence speed as well as a higher possibility of having the global optimum, a distributed k-means++ algorithm is first proposed to find the initial centroids before executing the distributed k-means algorithm and the distributed fuzzy c-means algorithm. The proposed distributed k-means algorithm is capable of partitioning the data observed by the nodes into measure-dependent groups which have small in-group and large out-group distances, while the proposed distributed fuzzy c-means algorithm is capable of partitioning the data observed by the nodes into different measure-dependent groups with degrees of membership values ranging from 0 to 1. Simulation results show that the proposed distributed algorithms can achieve almost the same results as that given by the centralized clustering algorithms.
Fuzzy c-means algorithm—A review
- R Suganya
- R Shanthi