Feature Information Of Data Research Articles

Internet of Things intrusion detection: Research and practice of NSENet and LSTM fusion models

To address the problems of complex environment, limited device computational resources and limited memory resources in the existing IoT, SELSTM, an intrusion detection system composed of NSENet and LSTM fusion based on SENet, is investigated. The NSENet part of the SELSTM system is based on the squeeze-and-excitation network (SENet). The lightweight computational modules NonLocal, SKConv and inverted residuals are fused into SE blocks, and self-attention of Nonlocal is used to improve the local receptive field of feature extraction. The channel attention and spatial attention of each part of the data are strengthened by the use of SKConv To enhance the adaptive convolution ability of the model and ensure the completeness of the information, the properties of the inverted residual structure are used to ensure that the gradient of the model decreases steadily without gradient explosion or disappearance. For the problem of data imbalance, the dataset is randomly resampled using the weight resampling technique to improve the balance of the dataset to ensure that the final detection effect of the model is more effective and generalized, while the data flow is divided into two parts for processing, and the model parameters are optimized using the model gradient optimizer consisting of the optimizer Lion and the optimization function Lookahead. The model extracts the spatial and temporal features of the data through multidimensional extraction to ensure the completeness of the data feature information in multiple dimensions, thus obtaining better detection results. The results of the experiments comparing the SELSTM model with other models on the intrusion dataset show that the intrusion detection model has a higher detection precision and accuracy than the traditional deep learning intrusion detection model, which indicates that the SELSTM has better detection performance properties and better practicality and effectiveness on IoT devices.

MSGAFN: Multi-scale graph attention fusion network for machine fault diagnosis

Fault diagnosis is essential in various fields, such as industrial manufacturing and engineering maintenance. Graph neural networks that take graph data as input can explore the relationships between data, which has strong feature expression capabilities. Currently, traditional fault diagnostic methods based on graph neural networks (GNN) are challenging to capture local and global feature information of data effectively. Most GNN models fail to consider the inherent differences between adjacent nodes, and they perform poorly in processing vibration signals in real-world industrial scenarios that commonly have strong noise. To resolve these concerns, this paper proposes a method for fault diagnosis based on a multi-scale graph attention fusion network (MSGAFN). In MSGAFN, data samples are constructed as clan graphs with multiple information scales to effectively represent local and global information of the graph structure data. Additionally, MSGAFN designed a new multi-scale feature fusion layer (MSFFL) to automatically learn the weights of adjacent nodes to represent their importance to the central node and reflect the differences between different adjacent nodes. This method is fully validated on bearing and gear datasets. The experimental results demonstrate that the proposed method exhibits excellent performance under conditions of imbalanced datasets and strong noise, providing a promising approach for bearings and gears fault diagnosis in real-world industrial scenarios with strong noise.

Compensation of accuracy by increased data “thickness” for high timeliness in fault diagnosis of chemical process

BackgroundFault diagnosis rate (FDR) and fault diagnosis timeliness (FDT) are critical to a fault diagnosis model applied to industrial processes. Generally, high FDR requires process data of longer time, which means a low timeliness; while, the time-length reduction of data will result in a decreased FDR as learning features from less data is more difficult. MethodsTo address this problem, an idea is proposed that the time-length reduction of process data could be compensated by enhancing the data “thickness”, i.e., the shortage of feature information of data could be compensated by investigating more patterns from multiple different aspects of the data. Based on this, the multiple pattern representation-convolutional neural network (MPR-CNN) is proposed, in which multiple pattern representation algorithms are used first to pre-extract features from data. The resulting three-dimensional matrix contains richer features, which could be learned more efficiently by the CNN. Significant FindingsThe diagnosis results from a continuous and a batch chemical process show that MPR-CNN significantly improves the fault diagnosis performance, which achieved 96.1% of FDR in the Tennessee Eastman process and 98.1% of FDR in a semi-batch crystallization process, and outperforms other CNN structure-based models.

Prediction of lithium-ion battery SOC based on the fusion of MHA and ConvolGRU

If the charging state of the lithium-ion battery can be accurately predicted, overcharge and overdischarge of the battery can be avoided, and the service life of the battery can be improved. In order to improve the prediction accuracy of SOC, a prediction method combined with convolutional layer, multi-head attention mechanism and gated cycle unit is proposed to extract data feature information from different dimensions of space and time. Using the data set of the University of Maryland, we simulated the battery in real vehicle operating conditions at different temperatures (0 °C, 25 °C, 45 °C). The test results showed that the mean absolute error, root mean square error and maximum prediction error of the model were 0.53%, 0.67% and 0.4% respectively. The results show that the model can predict SOC accurately. At the same time, the comparison with other prediction models shows that the prediction accuracy of this model is the highest.

Quantum walks-based classification model with resistance for cloud computing attacks

Cloud computing is considerably investigable and adoptable in both industry and academia, and Software Defined Networking (SDN) has been applied in cloud computing. Although SDN mitigates some security issues in cloud computing, new security issues related to its own architecture are also introduced. In this paper, we propose a quantum walks-based classification model which is available for intrusion detection in cloud computing. The proposed model concentrates feature information of data via Principal Component Analysis, and then aggregates the concentrated data in the way of quantum walks by a training-free clustering algorithm. The clustering algorithm constructs coin transformation and conditional shift transformation based on transition probabilities to move similar data toward each other. To enhance the usability of the proposed model in cloud computing security, we propose a new cloud architecture which adds security layer in SDN to ponder the protection of cloud computing fundamentally, and simplify transition probabilities equations of clustering algorithm without affecting clustering accuracy, decreasing the time complexity from O(nk2) to O(nk). The experimental results on popular datasets (Accuracy: 99.4% on InSDN, 95.8% on NSL-KDD, 98% on UNSW-NB15 and 96.4% on CSE-CIC-IDS2018) revealed that the proposed model is effective dealing with attacks on SDN-based cloud computing, and is able to maintain stable and excellent attack identification ability under different traffic intensities.

Parallel Network-Based Sliding Mode Tracking Control for Robotic Manipulators with Uncertain Dynamics

Robot dynamics model uncertainty and unpredictable external perturbations are important factors that influence control accuracy and stability. To accurately compensate for the dynamics model in sliding mode control (SMC), a new parallel network (PCR) is proposed in this paper. The network parallelizes the radial basis function and convolutional neural network, which gives it the advantage of making full use of one-dimensional data fitting results and two-dimensional data feature information, realizing the deep learning of multidimensional data and improving the model’s compensation accuracy and anti-interference ability. Meanwhile, based on the integration of adaptive control techniques and gradient descent, a new weight update algorithm is designed to realize the online learning of PCR networks under loss-free functions. Then, a new sliding mode controller (PCR-SMC) is established. The model-free intelligent control of the robot is accomplished without knowledge of the predetermined upper bounds. Additionally, the stability analysis of the control system is proved by the Lyapunov theorem. Lastly, robot tracking control simulations are performed on two trajectories. The results demonstrate the high-precision tracking performance of this controller in comparison with the RBF-SMC controller.

Detection and Identification of Potato-Typical Diseases Based on Multidimensional Fusion Atrous-CNN and Hyperspectral Data

As one of the world’s most crucial crops, the potato is an essential source of nutrition for human activities. However, several diseases pose a severe threat to the yield and quality of potatoes. Timely and accurate detection and identification of potato diseases are of great importance. Hyperspectral imaging has emerged as an essential tool that provides rich spectral and spatial distribution information and has been widely used in potato disease detection and identification. Nevertheless, the accuracy of prediction is often low when processing hyperspectral data using a one-dimensional convolutional neural network (1D-CNN). Additionally, conventional three-dimensional convolutional neural networks (3D-CNN) often require high hardware consumption while processing hyperspectral data. In this paper, we propose an Atrous-CNN network structure that fuses multiple dimensions to address these problems. The proposed structure combines the spectral information extracted by 1D-CNN, the spatial information extracted by 2D-CNN, and the spatial spectrum information extracted by 3D-CNN. To enhance the perceptual field of the convolution kernel and reduce the loss of hyperspectral data, null convolution is utilized in 1D-CNN and 2D-CNN to extract data features. We tested the proposed structure on three real-world potato diseases and achieved recognition accuracy of up to 0.9987. The algorithm presented in this paper effectively extracts hyperspectral data feature information using three different dimensional CNNs, leading to higher recognition accuracy and reduced hardware consumption. Therefore, it is feasible to use the 1D-CNN network and hyperspectral image technology for potato plant disease identification.

Abnormal Operation Detection of Access Equipment in Dispatching Exchange Network Based on Data Mining

To meet the requirements of a power dispatching service, a dispatching switching network has been widely used instead of circuit switching technology. However, with the growth of the number of access devices, the problem of network security is becoming more and more serious. At present, the commonly used access equipment abnormal operation detection method mainly relies on a two-stage clustering method to identify the abnormal operation state, but the processing effect is not good in the face of a large number of data, resulting in a low F1 value of the final detection result. Therefore, aiming at the access equipment of the dispatching exchange network, an abnormal operation detection method based on data mining is proposed. Scapy, a network tool, is used to capture the traffic data of access devices, and the collected data are normalized and fused. Data feature information is extracted through time sliding window and incremental calculation mode. With the clustering algorithm and related rules in data mining as the core, the abnormal operation detection method of access equipment is designed to obtain accurate abnormal detection results. Finally, based on the recursive graph structure, the detection results are visualized. The experimental results show that the average F1 value of the detection results of the proposed method is 0. 98, which indicates that the detection results of the proposed method have higher accuracy, and based on this, the network security isolation is carried out, which effectively improves the security of dispatching switching network.

HFCM-LSTM: A novel hybrid framework for state-of-health estimation of lithium-ion battery

Accurate estimating the state of health (SOH) of lithium-ion battery plays a significant role in the safe operation of electric vehicles. With the development of deep learning, neural network-based methods have attracted much attention from researchers. While most of the existing SOH estimation methods are built by a single network, failing to sufficiently extract data information, and thus leading to the limited accuracy and generality (i.e., such a single network makes it difficult to estimate the SOH of battery, with different types and working conditions). Towards this issue, a novel hybrid network, called HFCM (Hierarchical Feature Coupled Module)-LSTM (Long–short-term memory), is designed to fully extract the original data information, making it more accurate to estimate the SOH of battery, with different types and working conditions. Specifically, the proposed HFCM-LSTM mainly consists of two components, HCFM and LSTM. The HCFM is proposed to comprehensively extract the original data feature information from the original samples. On the other hand, following the HFCM, a LSTM module is employed to model time series information. Based on this HFCM-LSTM network, the data obtained directly from the battery are fed into the model as input, enabling an end-to-end SOH estimation of the battery. A series of experiments are conducted on both NASA dataset and Oxford dataset, the experimental results demonstrate that the proposed SOH estimation algorithm outperforms several existing state-of-the-art methods, in terms of accuracy and generality.

Ethereum-Based User Authentication Model Using EOG Data for a Service Provider

Unlike general passwords, user authentication technology using biometric data cannot be lost or forgotten, and it has the advantage of being impossible to forge or falsify by attackers. Since this biometric data contains sensitive information, a safe storage method is needed. However, if biometric data is managed on a central server, it is limited by being vulnerable to integrity infringement attacks by system attackers and persistent infringement attacks on the authentication service. To solve this problem, a model using blockchain-based information security technology is proposed in this paper. The aim is to safely manage the user biometric data by dispersing the storage of the biometric data feature information for user authentication in smart contracts and IPFS using Ethereum. We have verified that it takes an average of 1,472.733 ms and 217.829 ms, to register and authenticate a user in the proposed model and we expect that it will provide a reliable authentication service to the users compared to the existing authentication method in which the biometric data is managed by a single server.

Introducing the ETH/PRIO Civil Conflict Ceasefire Dataset

This article introduces the Civil Conflict CeaseFire (CF) dataset. The CF data covers all ceasefires in civil conflict between 1989 and 2020, including multilateral, bilateral and unilateral arrangements, ranging from verbal arrangements to detailed written agreements. In total, the CF data includes 2202 ceasefires across 66 countries and 109 civil conflicts. The data feature information on the actors involved in the ceasefire, and the class, purpose, coverage, and end date of the ceasefire. The CF data provide an empirical basis to assess the conditions that give rise to ceasefires, how ceasefires affect the dynamics of conflict, and the role of a ceasefire in the peace process. This article presents the rationale underlying the data collection, the coding rules and procedures, and how this data can be used for analysis.

Dynamic-static collaborative strategy for industrial data modeling based on hierarchical deep networks

With the rapid development of industrial processes, the complex nonlinear dynamic features of process data have created great challenges for deep learning models. However, existing deep learning models, such as stacked autoencoder (SAE), mainly focus on capturing static data feature information while ignoring the extraction of dynamic data evolution patterns. To combat this issue, this paper proposes a novel deep learning model based on dynamic SAE to collaboratively learn deep static and dynamic features of process data. First, sliding window technology is utilized to obtain temporal nearest neighbor samples within each time window. Then, multiple SAE modules are utilized hierarchically to extract static features in each window. Then, the proposed dynamic feature extraction module is exploited to extract the local dynamic information from the data within each window. Finally, the learned dynamic and static features are collaboratively fused to build a soft-sensor model for quality prediction tasks. To validate the superiority of the proposed model, it is applied to simulation experiments of a tobacco drying process and a hydrocracking process. The experimental results show that the proposed model performs better than other methods.

VMD-SEAE-TL-Based Data-Driven soft sensor modeling for a complex industrial batch processes

On complex batch industrial processes, soft sensor modeling plays a key role in process control and monitoring. Considering the nonlinearity, time-varying, and repetitive nature of the batch process, this paper proposes a soft sensor modeling method (VMD-SEAE-TL) based on variational mode decomposition (VMD), stacked enhanced autoencoder (SEAE) and transfer learning (TL) algorithms for online detection of key variables in batch industrial production processes. Firstly, the raw industrial process data are decomposed and reconstructed using VMD to achieve denoising and reduce the non-smooth characteristics of the data series. Secondly, based on the reconstructed process data, the SEAE network is used to deeply extract data feature information and achieve high accuracy regression prediction. Here, during the SEAE training process, each layer of the enhanced autoencoder (EAE) network reconstructs both the network input and the original input. The purpose of this operation is to extract the deep feature information of the process data and to ensure that there is no cumulative loss of the original input information. Further, it is considered that the working conditions in the batch industrial production process are time-varying, and this often makes it difficult for the model trained on the source domain data to accurately predict the trend of the process variables in the target domain. For this problem, the maximum mean deviation (MMD)-based transfer learning algorithm is introduced, which is used to solve the domain adaption problem with changing working conditions. Finally, based on two actual industrial process cases, the effectiveness and practicality of the proposed soft sensor are verified by various experimental results.

Spatiotemporal non-negative projected convolutional network with bidirectional NMF and 3DCNN for remaining useful life estimation of bearings

Remaining useful life (RUL) estimation for bearings is crucial in guaranteeing the reliability of rotating machinery. With the rapid development of information science, deep-learning-based RUL estimation has become more appealing as it can automatically establish the mapping relationship between the monitored data and the degradation states through feature learning. Vibration analysis via time–frequency representation (TFR) has shown great advantages for the detection of bearing damage in deep-learning-based prognostics. However, the following two problems remain: 1) insufficient or ineffective utilization of the data feature information, and 2) the requirement for huge computational resources, which still present challenges for the accuracy and efficiency of TFR-based prognostics. A novel RUL estimation approach called spatiotemporal non-negative projected convolutional network (SNPCN) is hence proposed. The approach can fully learn the spatiotemporal degradation features of bearing TFRs with high computational efficiency. In detail, the continuous wavelet transform (CWT) was applied as a TFR analysis method to reveal the nonstationary properties of the bearing degradation signals. Then, a newly proposed bidirectional non-negative matrix factorization (BiNMF) method was used to obtain the low-rank eigenmatrices of the TFRs and greatly compress the calculations in TFR-based prognostics. A three-dimensional convolutional neural network (3DCNN) was next constructed to learn the spatiotemporal degradation features in adjacent BiNMF eigenmatrices and construct the mapping relationship between the bearing RUL and current monitored data. Experiments on the PRONOSTIA platform demonstrate the feasibility and superiority of the proposed SNPCN-based bearing RUL estimation approach.

Dimension Reduction Algorithm Based on Adaptive Maximum Linear Neighborhood Selection in Edge Computing

With the rapid development of the Internet of Things (IoT), large quantities of data have been generated. Due to the limitation of the network bandwidth, the time and energy consumption of data transmission are increased. Data feature information can be extracted in real-time by the deployment of a data processing center. In this article, a novel dimension reduction approach is proposed in edge computing. First, a four-layer data processing framework is designed for data acquisition. A task assignment algorithm (TAA) is used for the condition when the edge node stops working due to an accident. Second, a threshold strategy is proposed to filter the data and reduce the dimension. Finally, the dimension reduction algorithm based on adaptive maximum linear neighborhood selection (AMLNS) is proposed. The harmonic geodesic distance is introduced to avoid the deformation of the manifold structure in AMLNS algorithm. Particularly, multiple weights are used to construct linear structure, which has a better embedding effect than single weight. The maximum linear neighborhood error weight is used to calculate the data coordinates. Experimental results show that the TAA improves the task completion rate about 15% and 36% over the random assignment method in mobile layer and edge layer, respectively. Compared with the local linear embedding (LLE), the points distribution of AMLNS is more uniform and regular, the execution time of AMLNS is reduced by about 17%. Furthermore, the embedding errors are less than those of LLE.

Network Intrusion Detection Model Based on Artificial Intelligence

The Internet occupies a more and more important position in people’s life. The society has become more convenient because of the progress of network technology, and the network security problem has attracted more and more attention. Therefore, the security technology based on network is more and more important. Intrusion detection technology is the main research direction of dynamic security tools. Aiming at the problem of low accuracy of network intrusion detection, this paper proposes a network intrusion detection model based on AI, introduces artificial intelligence neural network and algorithm, extracts data feature information through repeated training of the algorithm, and constructs network intrusion monitoring model based on neural network. According to the design goal of the model, the basic framework of the system is given, which provides accurate detection basis for network intrusion detection. The experimental results show that the detection accuracy of the model is high.

Dilated-Dense Network for traffic congestion prediction Model

It is very important to use the convolutional neural network model for urban short-term traffic congestion forecasting. However, pooling operation process is prone to cause data structure information loss when using CNN model to predict short-term traffic flow. Complete data feature information cannot be transmitted, which reduces the prediction ability of the model. To solve these problems, an innovative model based on dilated convolutional-dense networks is proposed in this paper. Firstly, the model can use the dilated convolution to obtain the characteristics of the larger receptive field with fewer network parameters, and fully extract the complex and variable data features. Then, through up-sampling and densely connected, the problem of parameter degradation in the process of increasing layers of neural network can be solved. Finally, the actual urban road average speed data blocks are taken to verify the validity of the model. Experimental results show that compared with traditional CNN model, the average absolute error of the network prediction structure is reduced by 3% to 23%.

Deep Principal Component Analysis Based on Layerwise Feature Extraction and Its Application to Nonlinear Process Monitoring

In order to deeply exploit intrinsic data feature information hidden among the process data, an improved kernel principal component analysis (KPCA) method is proposed, which is referred to as deep principal component analysis (DePCA). Specifically, motivated by the deep learning strategy, we design a hierarchical statistical model structure to extract multilayer data features, including both the linear and nonlinear principal components. To reduce the computation complexity in nonlinear feature extraction, the feature-samples’ selection technique is applied to build the sparse kernel model for DePCA. To integrate the monitoring statistics at each feature layer, Bayesian inference is used to transform the monitoring statistics into fault probabilities, and then, two probability-based DePCA monitoring statistics are constructed by weighting the fault probabilities at all the feature layers. Two case studies involving a simulated nonlinear system and the benchmark Tennessee Eastman process demonstrate the superior fault detection performance of the proposed DePCA method over the traditional KPCA-based methods.

Research on gesture recognition of smart data fusion features in the IoT

With the rapid development of Internet of things technology, the interaction between people and things has become increasingly frequent. Using simple gestures instead of complex operations to interact with the machine, the fusion of smart data feature information and so on has gradually become a research hotspot. Considering that the depth image of the Kinect sensor lacks color information and is susceptible to depth thresholds, this paper proposes a gesture segmentation method based on the fusion of color information and depth information; in order to ensure the complete information of the segmentation image, a gesture feature extraction method based on Hu invariant moment and HOG feature fusion is proposed; and by determining the optimal weight parameters, the global and local features are effectively fused. Finally, the SVM classifier is used to classify and identify gestures. The experimental results show that the proposed fusion features method has a higher gesture recognition rate and better robustness than the traditional method.

Hyperspectral Data Feature Extraction Using Deep Learning Hybrid Model

The original hyperspectral data served as the initial features has the characteristics of high dimension and redundancy, which is not suitable for the subsequent analysis, so extracting feature information is needed. The deep learning model has a strong ability in feature learning, but if the model has too many layers which will lead to the original information loss in the process of layer-by-layer feature learning and reduce the subsequent classification accuracy. To solve this problem, the paper proposed a deep learning model of hybrid structure with the contractive autoencoder and restricted boltzmann machine to extract the hyperspectral data feature information. First, through pre-processing the spectral data, the 2d spectrum data is converted into a one dimensional vector. Then, a hybrid model is constructed for unsupervised training and supervised learning for the hyperspectral data, and features are extracted from bottom to top gradually according to the hybrid model. Finally, the SVM classifier is adopted to enhance the classification ability of spectral data. The paper uses the hybrid model proposed to test for extracting features with two sets of AVIRIS data and compares with PCA and GCA methods. The experiment results show that the feature extraction algorithm based on hybrid depth model can get the better features, and have strong distinguish performance, and can get better classification accuracy by the SVM algorithm.