Sliding Time Window Research Articles

Multilevel Residual Prophet Network Time Series Model for Prediction of Irregularities on High-Speed Railway Track

Accurate prediction of the change in track irregularity plays an essential role in keeping the high-speed railway safe and stable. Regular maintenance is an important measure to guarantee track smoothness, which has a great influence on the change of track state. This paper introduces the time series anomaly detection model (AD) to detect changes by tracking the difference between the mean values of two sliding time windows. These changes are taken as one of the holidays of the prophet model which is called the AD_ Prophet model. Moreover, this paper proposes a multilevel residual prophet (Re-Prophet) prediction network which can make full use of the information of residual data to predict the results. The final predicted result is the sum of the value of each prediction. To explore the characteristics of the time series, the Hodrick-Prescott filter is used to explore the trend. A multi-month-wise box plot is used to explore the seasonal volatility and the wavelet transform is used to explore the cyclical variation. Based on these characteristics, the fitting model can be chosen. Finally, to verify the high accuracy of the model, the surface irregularities of the high-speed railway are used as the research objects. Compared with conventional prediction algorithms, the proposed model has the smallest prediction deviation and the highest accuracy based on the evaluation indicators such as mean squared error, root mean squared error, mean absolute error, and mean absolute percentage error.

Multi-GNSS Combination Multipath Reflectometry Based on IVMD Method for Sea Level Retrieval

Sea level monitoring is particularly important in coastal areas that are vulnerable to marine disasters. It was recently demonstrated that the global navigation satellite system multipath reflectometry (GNSS-MR) technique, which uses multipath signals reflected from the sea, can be applied to determine the sea level. However, this approach does not provide sufficient accuracy or equally spaced sampling to meet the actual sea level monitoring requirements for certain stations. To solve the deficiency of the traditional GNSS-MR technique, the least squares method, which is based on sliding time windows, was applied. Using the sliding windows to combine the quad-constellation multi-GNSS retrieval can effectively improve the accuracy and time resolution of sea level retrieval, but insufficient data or a lack of data in some time ranges and missing overflights in some timeframes can lead to the calculation of faults in these time windows, causing the estimated loss of corresponding sampling points. In this study, we used a robust regression solution strategy based on multi-GNSS sea level retrieval and an improved variational mode decomposition (IVMD) algorithm to process sea level retrieval after robust regression. BRST and HKQT stations are located on the western coast of France and the northern coast of Hong Kong. The two stations can both receive satellite observation data from the four satellite systems. Through the experiment, using data retrieved from the BRST and HKQT stations, the results of this study demonstrate that the IVMD method based on multi-GNSS sea level retrieval can further improve the accuracy to <10 cm and can achieve 10 min equal interval sampling. This is significant for using GNSS-MR technology to detect sea level height and monitor sea level change and could be applied to other sites.

State‐of‐Charge Prediction of Lithium‐Ion Batteries Based on Sparse Autoencoder and Gated Recurrent Unit Neural Network

Lithium‐ion batteries are widely used in daily life because of their fast charging and high energy density. The accurate prediction of state‐of‐charge (SOC), critical for the quality evaluation and long‐term planning of lithium‐ion batteries, has still been challenging owing to sophisticated battery dynamics and ever‐changing ambient conditions. Herein, a joint sparse autoencoder (SAE) and gated recurrent unit (GRU) model are developed to improve SOC prediction performance. The sliding time window method is first utilized for data reconstruction to explore the continuous time trend and discrete state characteristics in multi‐cycle data of the lithium‐ion batteries. The SAE is then used to extract features from the reconstruction data. The GRU is finally applied to realize an accurate SOC prediction. The performance of proposed method is tested on a benchmark lithium‐ion battery degradation dataset and actual battery degradation experiments. Experimental tests show that the present model has the lowest root mean square error of 0.176 and 0.271, the lowest mean absolute error of 0.1396 and 0.2412, and the lowest mean square error of 0.03098 and 0.07355 on two datasets, respectively, demonstrating that the proposed method outperforms the state‐of‐the‐art methods in terms of accuracy and robustness.

Detection approach for unusable shared bikes enabled by reinforcement learning and PageRank algorithm

Existing research models can neither indicate the availability of shared bikes nor detect unusable ones owing to a lack of information on bike maintenance and failure. To improve awareness regarding the availability of shared bikes, we propose an innovative approach for detecting unusable shared bikes based on reinforcement learning and the PageRank algorithm. The proposed method identifies unusable shared bikes depending on the local travel data and provides a ranking of the shared bikes according to their availability levels. Given a sliding time window, the value function for the reinforcement learning model was determined by considering the cumulative number of unavailable shared bikes, the proportion of rental cancelations at the same stations, and the mean time between the cancelations. Reinforcement learning was then used to identify shared bikes with the worst availability. An availability ranking for the shared bikes below the reward threshold was performed using the PageRank algorithm. The proposed detection approach was applied to a trip dataset of a real-world bike-sharing system to illustrate the modeling process and its effectiveness. The detection results of unusable shared bikes in the absence of failure and feedback data can provide essential information to support the maintenance management decisions regarding shared bikes.

Drift Detection of Intelligent Sensor Networks Deployment Based on Graph Stream

The optimal deployment of sensor nodes can effectively improve the monitoring performance of the sensor network. The existing optimization deployment strategies focus on how to achieve the optimal coverage or the maximal monitoring efficiency of the sensor network deployment, and lack research on the issue of when a deployed sensor network need to update its deployment. Considering that the parameters of the monitored targets by sensor network will vary over time, making the deployment a certain degree of time drift characteristic. In this paper, we propose a deployment drift detection algorithm based on graph stream for intelligent sensor networks (ISNs) to investigate the optimization deployment update time of sensor nodes for more efficient targets monitoring. We utilize graph stream to represent the topology of sensor network at different timestamps, and propose a projection fusion clustering algorithm based on Structural Clustering Algorithm for Networks (SCAN) to construct the tensor graph summarization within a sliding time window. According to the contrast between the node-level and the graph-level queries of the summary graph in adjacent windows, we refer Hoeffding boundary and graph contrastive loss function as the criterion to judge the optimal update time of the sensor network deployment. Experiments on Intel_Lab_Data and CIMIS datasets demonstrate that our approach can effectively detect the update moment of the optimal deployment of sensor network.

Altered dynamic functional connectivity associates with post-traumatic stress disorder.

Research has been looking into neural pathophysiology of post-traumatic stress disorder (PTSD) and dynamic functioning connectivity (dFC) applying resting state functional magnetic resonance imaging (rs-fMRI). Previous studies showed that PTSD related impairments are associated with alterations distributed across different brain regions and disorganized functional connectivity, especially in Default Mode Network and the cerebellar area. In this study, we specifically looked into dFC on a whole brain level, and we focused on critical regions such as DMN and cerebellum. To explore the characteristics of dFC among patients with PTSD, we collected rs-fMRI data from 27 PTSD patients and 30 healthy controls. The study also added a control group of 33 trauma-exposed individuals to further look into trauma impact. Utilizing group spatial independent component analysis (ICA), the dynamic properties on whole brain level were detected with sliding time window approach, and k-means clustering. Two reoccurring FC "States" were identified, with connections being more concentrated on a within-network level in one state and more strongly inter-connected in the other state. Abnormalities in dFC were found within DMN, between DMN and cerebellum, and between DMN and visual network for PTSD patients. The findings were in accordance with the study hypothesis that the dFC alterations might point to deficits in emotional modulation and dysfunctional self-referential thought. Abnormalities in dFC among PTSD patients might also be indicators of PTSD symptoms including depression and anxiety, hypervigilance, impaired cognitive functioning and self-referential information processing.

Data driven intrusion detection for 6LoWPAN based IoT systems

Wide adoption of Internet of Things (IoT) devices and their limitations in terms of hardware cause them to be easy targets for attackers. This, in turn, requires monitoring such systems using intrusion detection systems and take mitigative actions against insider and outsider attackers. Recent studies have explored that machine learning based intrusion detection systems are quite successful in detecting different types of cyber threats targeting IoT systems. However, the proposed systems in these studies incurred limitations in terms of the characteristics of their datasets and detection models. Specifically, a big proportion of the proposed models were developed using simulation-based data generated through specific simulators. Some of these studies also used previously published testbed data that contain the samples of outdated IoT attacks and vulnerabilities. Furthermore, they focused on a lower attack variety and proposed binary classifiers which do not scale in multi-attack scenarios. In this study, we propose a machine learning based multi-class classifier that can classify 6 attack types together with the benign traffic. Our node based feature extraction and detection methodology allows locating the network addresses of the attackers, rather than a rough network level attack existence information, by modeling their traffic characteristics over a sliding time window. For training and testing our models, we also propose an intrusion detection dataset generated using the traffic data collected from real IoT devices running with 6LoWPAN and RPL protocols. Besides having RPL routing attacks in the dataset, we leverage Mirai botnet, employed frequently to target IoT devices. The results show that the proposed intrusion detection system can detect 6 attack types with high recall scores ranging from 79% to 100%. We also illustrate the practicality of the developed model via deployment in a proof of concept implementation over a testbed.

Model-free super-twisting terminal sliding mode controller using sliding mode disturbance observer for n-DOF upper-limb rehabilitation exoskeleton with backlash hysteresis

This paper proposes a model-free super-twisting terminal sliding mode controller (MF-STTSMC) for the n-DOF upper-limb rehabilitation exoskeleton in the presence of external disturbance and backlash hysteresis. This ultra-local model-based control scheme consists of a sliding mode disturbance observer (SMDO) and a super-twisting non-singular fast terminal sliding mode controller (ST-NFTSMC). The ultra-local model is used to approximate the origin complex exoskeleton system in a short sliding time window, which avoids the requirement for accurate modelling and decreases the difficulty of controller design. To compensate for the lumped uncertainties, the SMDO is constructed to achieve finite-time observation and zero estimation error. Besides, to realise finite-time convergence of sliding surface and tracking error with less chattering, the ST-NFTSMC is proposed by employing a super-twisting algorithm. Furthermore, the stability of a closed-loop system with MF-STTSMC is ensured by using the Lyapunov theorem. Finally, to validate the proposed strategy, co-simulations of 7-DOF iReHave upper-limb exoskeleton are realised by using SolidWorks, SimMechanics, and Matlab/Simulink. Compared to time-delay estimation-based intelligent proportional-derivative control (TDE-iPD), α-variable adaptive model-free controller (α-AMFC), and inclusive and enhanced time delay controller (IETDC), the higher performances of the proposed MF-STTSMC are demonstrated.

Online State-of-Health Estimation Method for Lithium-Ion Battery Based on CEEMDAN for Feature Analysis and RBF Neural Network

The state-of-health (SOH) estimation is of great importance in the battery management system. However, the accurate real-time SOH estimation is still a challenge because the capacity of the battery cannot be precisely monitored and measured in real-time. In response to this problem, this article proposes an online SOH monitoring method for lithium-ion battery (LIB) based on the charge–discharge feature of battery in real-time. First, the relationship between the charge–discharge feature and SOH through principal component analysis (PCA) is analyzed, and the appropriate feature which can precisely describe the battery aging process is then selected. Second, as to decompose and denoise the collected charge–discharge feature data, the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is applied. The major trend obtained is predicted by logistic regression based on the sliding time window (LR-STW), and the minor fluctuation is predicted by Kalman filter (KF). Finally, an online SOH estimation method based on RBF neural network is proposed to establish the mapping relationship between feature and SOH. The proposed method was verified by the NASA experimental data. The experimental results show that the proposed method can accurately predict the LIB SOH in real-time, and the robustness of the proposed method is strengthened and promised.

An Energy Sensitive and Congestion Balance Routing Scheme for Non-Terrestrial-Satellite-Network (NTSN)

Recently, non-terrestrial-satellite-networks (NTSN) have attracted more attention due to excellent performance. NTSN specifically refers to a mega satellite constellation network. Global coverage, low latency and high data rate make global interconnection possible. However, the lack of on-board energy degrades the efficiency of the network system. To solve those pressing problems, an energy sensitive and congestion balance (ESCB) routing scheme is proposed in this paper. To compensate for the limited energy of satellites, a tubular sliding time window (TSTW) model is proposed. The residual energy can be predicted by it and a multi-objective algorithm is applied to calculate the routing path. Energy constraints, network congestion and latency are taken into account in the specific algorithm. A traffic model consisting of two different services is introduced as a verification system. Simulations demonstrate that the specific routing scheme achieves energy balance and congestion avoidance. In addition, comparison results demonstrate that ESCB is superior to the virtual-topology-based shortest path (VT-SP) routing algorithm.

Streaming Data Processing

Aims: The data which is continuously being produced by hundreds of thousands of data sources is recognized as streamed data. The data which is processed via this kind of source is relatively smaller in size and is being sent at the same time it is generated.
 Study Design: In streaming data, the data range is so wide like the telemetry from interconnected devices or other such forms of data with the inclusion of certain web applications. This information should be handled consecutively and steadily on a record-by-record premise or throughout sliding time windows and utilized for a wide assortment of examinations including relationships, totals, separating, and inspecting.
 Place and Duration of Study: Service usage (for metering and billing), server activity, website clicks, and the geo-location of devices, people, and physical goods are just a few of the many aspects of a company's business and customer activity that can be seen through this type of analysis. It also enables companies to respond quickly to new arising situations.
 Methodology: The research methodology is used for the current research work is the qualitative method through which the research studies of a similar domain are studied thoroughly. It has been analyzed that the specified changes in the large volumes of data can better be managed through stream data processing.
 Results: The flaws of batch data processing are better dealt with through the usage of streaming data processing agenda. Real-time monitoring as well as response functionality are the keys to success in the given method of data processing.
 Conclusion: Stream data processing connects analytics and applications. Because multiple systems can be constructed using the same architecture, this makes the construction of the infrastructure a similar architecture. It additionally allows designers to fabricate applications that utilize scientific outcomes to straightforwardly answer information experiences and make a move.

Predicting aging trajectories of decline in brain volume, cortical thickness and fractional anisotropy in schizophrenia

Brain-age prediction is a novel approach to assessing deviated brain aging trajectories in different diseases. However, most studies have used an average brain age gap (BAG) of individuals with schizophrenia of different illness durations for comparison with healthy participants. Therefore, this study investigated whether declined brain structures as reflected by BAGs may be present in schizophrenia in terms of brain volume, cortical thickness, and fractional anisotropy across different illness durations. We used brain volume, cortical thickness, and fractional anisotropy as features to train three models from the training dataset. Three models were applied to predict brain ages in the hold-out test and schizophrenia datasets and calculate BAGs. We divided the schizophrenia dataset into multiple groups based on the illness duration using a sliding time window approach for ANCOVA analysis. The brain volume and cortical thickness models revealed that, in comparison with healthy controls, individuals with schizophrenia had larger BAGs across different illness durations, whereas the BAG in terms of fractional anisotropy did not differ from that of healthy controls after disease onset. Moreover, the BAG at the initial stage of schizophrenia was the largest in the cortical thickness model. In contrast, the BAG from approximately two decades after disease onset was the largest in the brain volume model. Our findings suggest that schizophrenia differentially affects the decline of different brain structures during the disease course. Moreover, different trends of decline in thickness and volume-based measures suggest a differential decline in dimensions of brain structure throughout the course of schizophrenia.

Real-time water quality detection based on fluctuation feature analysis with the LSTM model

AbstractSignal analysis and anomaly detection for water pollution early warning systems are important and necessary. In view of the nonlinear and volatile characteristics of water quality time series, this paper proposes a new method for water anomaly detection based on fluctuation feature analysis. The method has two steps. First, the water quality time series data are used to calculate the residuals between the observed value and the predicted value with the long short-term memory (LSTM) network. Second, the dynamic features are extracted by sliding time window and described by the Approximate Entropy (ApEn) which are input to the anomaly detection model with Isolation Forest. Compared with traditional anomaly detection methods, the results obtained by the proposed method show better performance in distinguishing water quality anomalies. The proposed method can be applied to real-time water quality anomaly detection and early warning.

A Reliable Online Dictionary Learning Denoising Strategy for Noisy Microseismic Data

Improving the quality of microseismic recordings is a critical step in microseismic data processing. We introduce a wavelet-weighted online dictionary learning (WWODL) denoising strategy for noisy microseismic recordings. We develop an adaptive parameters estimation approach for tunable Q-factor wavelet transform, which provides accurate periodic and non-stationary subband information from microseismic data for online dictionary learning (ODL). A sliding time window is employed to divide the obtained subbands into a series of patches of equal length, which are then assembled into a matrix and fed into the ODL. The subband kurtosis information is weighted to the constraint function of the ODL, further enhancing the sparse coding ability for each subband. Shortening the oscillation duration, an improved ODL is developed with a faster convergence speed in calculating sparse coefficients. Our results confirm that the WWODL can suppress high-frequency, low-frequency, and shared-bandwidth noises and has a minimal impact on the first arrival. The time consumption and signal-to-noise ratio of the WWODL are average 1/8.675 and 56.82% higher than ODL, respectively.

Online Tracking of Small-Signal Stability Rightmost Eigenvalue Based on Reference Point

Machine learning technologies have been applied to improve the real-time performance of small-signal stability (SSS) assessment, while achieving a high accuracy requires numerous samples, and model performance may degrade if not updated over time. Furthermore, single models tend to learn general features without analyzing specific characteristics within the samples, which may lead to a high frequency of large errors, particularly at operating points (OPs) where the eigenvalue trajectories have sudden changes. Facing such issues, this paper introduces the concept of reference points (RPs) for accurate online tracking of the rightmost eigenvalue (RE), as the RP information reflects the characteristics among its surrounding OPs. The performance of this model is sensitive to RPs, so affinity propagation (AP) clustering is employed to determine the number of RPs and generate corresponding groups, accommodating diverse OP characteristics. This paper generates data-driven networks for each group and combines them into a multi-network for precise RE prediction. Case studies show that the use of RPs improves accuracy by nearly 2% compared to methods without them, with even greater improvements in mixed load type scenarios. To mitigate computational stress, this paper proposes an adaptive partial update strategy based on the dynamic time warping algorithm, avoiding the need to update all networks within each sliding time window. Experimental results verify that the total running time is reduced by more than 10%. Online tracking demonstrates that RPs help decrease the frequency of large errors by 40%, especially at sudden change points of RE trajectories.

Research on Stock Index Prediction Based on Stock Correlation Network and Deep Learning

Aiming at the problem of stock index prediction, constructing a time series stock correlation network based on the fundamentals and technology of stock index components, then using the depth map neural network to learn the hierarchical representation of stock correlation network, and obtaining the candidate prediction signal in an end-to-end way. The architecture composed of depth map neural network method and end-to-end strategy is called DIFFPOOL architecture. Taking the CSI 300 index as the research object, combining the DIFFPOOL architecture with softmax classifier, long-term and short-term memory neural network (LSTM), linear regression, and logical regression, respectively, and uses the sliding time window method to obtain the corresponding prediction accuracy of stock index. The accuracy of the combined model under the optimal parameters fluctuates in the interval [0.56, 0.62]. Ultimately, the first mock exam is based on the mean absolute error (MAE) and the root mean square error (RMSE). The regression models of DIFFPOOL and regression models are compared with LSTM, recurrent neural network (RNN), and back propagation neural network (BP). Compared with the single model, the MAE and RMSE of the combined model are smaller, 0.0061 and 0.0081, respectively. Experiments show that by aggregating the node attribute information of the stock association network hierarchically, we can dynamically capture the impact of different industry sectors on stock index price fluctuations and further improve prediction accuracy.

Research on Indoor 3D Positioning Algorithm Based on WiFi Fingerprint

Indoor 3D positioning is useful in multistory buildings, such as shopping malls, libraries, and airports. This study focuses on indoor 3D positioning using wireless access points (AP) in an environment without adding additional hardware facilities in large-scale complex places. The integration of a deep learning algorithm into indoor 3D positioning is studied, and a 3D dynamic positioning model based on temporal fingerprints is proposed. In contrast to the traditional positioning models with a single input, the proposed method uses a sliding time window to build a temporal fingerprint chip as the input of the positioning model to provide abundant information for positioning. Temporal information can be used to distinguish locations with similar fingerprint vectors and to improve the accuracy and robustness of positioning. Moreover, deep learning has been applied for the automatic extraction of spatiotemporal features. A temporal convolutional network (TCN) feature extractor is proposed in this paper, which adopts a causal convolution mechanism, dilated convolution mechanism, and residual connection mechanism and is not limited by the size of the convolution kernel. It is capable of learning hidden information and spatiotemporal relationships from the input features and the extracted spatiotemporal features are connected with a deep neural network (DNN) regressor to fit the complex nonlinear mapping relationship between the features and position coordinates to estimate the 3D position coordinates of the target. Finally, an open-source public dataset was used to verify the performance of the localization algorithm. Experimental results demonstrated the effectiveness of the proposed positioning model and a comparison between the proposed model and existing models proved that the proposed model can provide more accurate three-dimensional position coordinates.

Industrial internet security evaluation technology based on digital twin

In the process of industrial Internet security situation analysis, affected by the perception model, the average approximation error of security assessment results is high. Therefore, an industrial Internet security evaluation technology based on digital twin is proposed. Design the industrial Internet assessment mode, design the industrial Internet technical framework according to the digital twin, and clarify the elements of security situation assessment. According to the sliding time window, the fuzzy knowledge base is established. The security situation awareness model is designed based on digital twin, combined with the calculation results of fuzzy closeness to improve the accuracy of automatic perception results. Establish a multi port access protocol and realize the industrial Internet security evaluation technology under the digital twin with the help of the backed up coded key file. The experimental results show that the accuracy of the evaluation results of the design method is always above 95%, the tamper-proof success rate is always above 95%, and the physical value difference between the final values is 0.10. Therefore, this method effectively improves the evaluation accuracy. and tamper-proof success rate, reducing the difference value, it effectively controls the deviation of Internet operation data, provides security, is feasible and effective.

EFFECT OF SYNCHRONIZATION OF CONTINUOUS ACOUSTIC EMISSION STATISTICAL PROPERTIES DURING STRUCTURALLY HETEROGENEOUS MATERIALS DEFORMATION

In this paper, a correlation analysis of the statistical properties of continuous acoustic emission recorded in various parts of marble and glass fiber laminate samples during their quasi-static deformation is carried out. The spectral measure of coherence, which is a generalization of the square modulus of the coherence spectrum to the case of multidimensional series, is chosen as a correlation measure. The measure of coherence was estimated for the width of the multifractal spectrum and the spectrum carrier realizing its maximum, calculated in a sliding time window for acoustic emission signals. It is shown that the preparation of a macrofracture site is accompanied by synchronization of the statistical properties of acoustic emission in the selected frequency intervals. Based on the analysis of changes in the frequency-averaged measure of coherence for both types of materials, four characteristic stages are distinguished, the boundaries of which are individual for each of the materials. The onset of the fourth stage, which is characterized by a monotonic increase in the average measure of the coherence of the statistical properties of the AE, can be chosen as a possible criterion for the transition of the material to the limiting state.

Remaining Useful Life Prediction Based on Multi-Representation Domain Adaptation

All current deep learning-based prediction methods for remaining useful life (RUL) assume that training and testing data have similar distributions, but the existence of various operating conditions, failure modes, and noise lead to insufficient data with similar distributions during the training process, thereby reducing RUL prediction performance. Domain adaptation can effectively solve this problem by learning the cross-domain invariant features of the source domain and target domain to reduce the distribution difference. However, most domain adaptive methods extract the source domain and target domain features into a single space for feature alignment, which may leave out effective information and affect the accuracy of prediction. To address this problem, we propose a data-driven approach named long short-term memory network and multi-representation domain adaptation (LSTM-MRAN). We standardize and process the degraded sensor data with a sliding time window, use LSTM to extract features from the degraded data, and mine the time series information between the data. Then, we use multiple substructures in multi-representation domain adaptation to extract features of the source domain and target domain from different spaces and align features by minimizing conditional maximum mean difference (CMMD) loss functions. The effectiveness of the method is verified by the CMAPSS dataset. Compared with methods without domain adaptation and other transfer learning methods, the proposed method provides more reliable RUL prediction results under datasets with different operating conditions and failure modes.