ASTREAM: Data-Stream-Driven Scalable Anomaly Detection With Accuracy Guarantee in IIoT Environment

IEEE Transactions on Network Science and Engineering, Journal Year: 2022, Volume and Issue: 10(5), P. 3007 - 3016

Published: March 8, 2022

Intrusion detection exerts a crucial influence on securing the IIoT driven by anomaly approaches. Dissimilar with static data, intrusion data is in form of dynamic stream possessing properties infiniteness, correlations, and distribution change. However, these cause some issues for current Firstly, it impractical to save whole dataset due infiniteness. Secondly, correlations are hardly considered. Thirdly, change can't be appropriately handled lack model update strategy. Thus, we propose ASTREAM ( a nomaly xmlns:xlink="http://www.w3.org/1999/xlink">stream s), novel approach that merges sliding window, update, strategies into LSHiForest achieve accurate efficient better scalability. has following characteristics: (a) window can utilized handle infiniteness streams; (b) introduced PCA consider between different attributes; (c) detect time train new model. Comprehensive experiments implemented KDDCUP99 validate performance. Experiment results reveal outperforms baselines aspects accuracy efficiency

Language: Английский

---

Anomaly Detection Using a Sliding Window Technique and Data Imputation with Machine Learning for Hydrological Time Series

Water, Journal Year: 2021, Volume and Issue: 13(13), P. 1862 - 1862

Published: July 3, 2021

Water level data obtained from telemetry stations typically contains large number of outliers. Anomaly detection and a imputation are necessary steps in monitoring system. can be detected if its values lie outside normal pattern distribution. We developed median-based statistical outlier approach using sliding window technique. In order to fill anomalies, various interpolation techniques were considered. Our proposed framework exhibited promising results after evaluating with F1-score root mean square error (RMSE) based on our artificially induced points. The present system also easily applied patterns hydrological time series diverse choices internal methods fine-tuned parameters. Specifically, the Spline method yielded superior performance non-cyclical while long short-term memory (LSTM) outperformed other distinct tidal pattern.

Language: Английский

---

LSTM-Autoencoder for Vibration Anomaly Detection in Vertical Carousel Storage and Retrieval System (VCSRS)

Sensors, Journal Year: 2023, Volume and Issue: 23(2), P. 1009 - 1009

Published: Jan. 15, 2023

Industry 5.0, also known as the "smart factory", is an evolution of manufacturing technology that utilizes advanced data analytics and machine learning techniques to optimize production processes. One key aspect 5.0 using vibration monitor detect anomalies in machinery equipment. In case a vertical carousel storage retrieval system (VCSRS), can be collected analyzed identify potential issues with system's operation. A correlation coefficient model was used accurately ascertain optimal sensor placement position. This utilized Fisher information matrix (FIM) effective independence (EFI) methods for maximum accuracy reliability. An LSTM-autoencoder (long short-term memory) training testing further enhance anomaly detection process. machine-learning technique allowed detecting patterns trends may not have been evident traditional methods. The combination resulted rate 97.70% system.

Language: Английский

---

Acoustic Anomaly Detection of Mechanical Failures in Noisy Real-Life Factory Environments

Electronics, Journal Year: 2021, Volume and Issue: 10(19), P. 2329 - 2329

Published: Sept. 23, 2021

Anomaly detection without employing dedicated sensors for each industrial machine is recognized as one of the essential techniques preventive maintenance and especially important factories with low automatization levels, a number which remain much larger than autonomous manufacturing lines. We have based our research on hypothesis that real-life sound data from working machines can be used diagnostics. However, contaminated drowned out by typical factory environmental sound, making application data-based anomaly an overly complicated process and, thus, main problem we are solving approach. In this paper, present noise-tolerant deep learning-based methodology sound-data-based within real-world machinery data. The element proposed generative adversarial network (GAN) reconstruction signal anomalies. experimental results obtained in Malfunctioning Industrial Machine Investigation Inspection (MIMII) show superiority over baseline approaches One-Class Support Vector (OC-SVM) Autoencoder–Decoder neural network. schematics using unscented Kalman Filter (UKF) mean square error (MSE) loss function L2 regularization term showed improvement Area Under Curve (AUC) noisy pump pump.

Language: Английский

---

Detecting outliers in a univariate time series dataset using unsupervised combined statistical methods: A case study on surface water temperature

Ecological Informatics, Journal Year: 2022, Volume and Issue: 69, P. 101672 - 101672

Published: May 11, 2022

Language: Английский

---

Anomaly Detection and Repairing for Improving Air Quality Monitoring

Sensors, Journal Year: 2023, Volume and Issue: 23(2), P. 640 - 640

Published: Jan. 6, 2023

Clean air in cities improves our health and overall quality of life helps fight climate change preserve environment. High-resolution measures pollutants' concentrations can support the identification urban areas with poor raise citizens' awareness while encouraging more sustainable behaviors. Recent advances Internet Things (IoT) technology have led to extensive use low-cost sensors for hyper-local monitoring. As a result, public administrations citizens increasingly rely on information obtained from make decisions their daily lives mitigate pollution effects. Unfortunately, most sensing applications, are known be error-prone. Thanks Artificial Intelligence (AI) technologies, it is possible devise computationally efficient methods that automatically pinpoint anomalies those data streams real time. In order enhance reliability we believe highly important set up data-cleaning process. this work, propose AIrSense, novel AI-based framework obtaining reliable pollutant raw collected by network sensors. It enacts an anomaly detection repairing procedure measurements before applying calibration model, which converts concentration gasses. There very few studies sensor (millivolts). Our approach first proposes detect repair they calibrated considering temporal sequence correlations between different features. If at least some previous available not anomalous, trains model uses prediction observations; otherwise, exploits observation. Firstly, majority voting system based three algorithms detects data. Then, repaired avoid missing values measurement time series. end, provides concentrations. Experiments conducted dataset 12,000 observations produced 12 demonstrated importance process improving algorithms' performances.

Language: Английский

---

Anomaly detection for streaming data based on grid-clustering and Gaussian distribution

Information Sciences, Journal Year: 2023, Volume and Issue: 638, P. 118989 - 118989

Published: April 23, 2023

Language: Английский

---

Recent advances in wireless sensor networks for structural health monitoring of civil infrastructure

Journal of Infrastructure Intelligence and Resilience, Journal Year: 2023, Volume and Issue: 3(1), P. 100066 - 100066

Published: Nov. 12, 2023

Wireless Smart Sensor Networks (WSSN) have seen significant advancements in recent years. They act as a core part of structural health monitoring (SHM) systems by facilitating efficient measurement, assessment, and hence maintenance civil infrastructure. This paper presents the latest technology developments WSSN last ten years, including ones for single sensor node those network nodes. Focus is placed on critical aspects such advancements, event-triggered sensing, multimeric edge/cloud computing, time synchronization, real-time data acquisition, decentralized processing, long-term reliability. In addition, full-scale applications demonstrations SHM are also summarized. Finally, remaining challenges future research directions discussed to promote further development applications.

Language: Английский

---

Detection of possible hydrological precursor anomalies using long short-term memory: A case study of the 1996 Lijiang earthquake

Journal of Hydrology, Journal Year: 2021, Volume and Issue: 599, P. 126369 - 126369

Published: April 26, 2021

Language: Английский

---

Machine Learning-Assisted, Process-Based Quality Control for Detecting Compromised Environmental Sensors

Environmental Science & Technology, Journal Year: 2023, Volume and Issue: 57(46), P. 18058 - 18066

Published: Aug. 15, 2023

Machine learning (ML) techniques promise to revolutionize environmental research and management, but collecting the necessary volumes of high-quality data remains challenging. Environmental sensors are often deployed under harsh conditions, requiring labor-intensive quality assurance control (QAQC) processes. The need for manual QAQC is a major impediment scalability these sensor networks. Existing automated make strong assumptions about noise profiles in they filter that do not necessarily hold broadly sensors, however. Toward goal increasing volume data, we introduce an ML-assisted methodology robust low signal-to-noise ratio data. Our approach embeds measurements into dynamical feature space trains binary classification algorithm (Support Vector Machine) detect deviation from expected process dynamics, indicating whether has become compromised requires maintenance. This strategy enables detection wide variety nonphysical signals. We apply three novel sets produced by 136 low-cost (stream level, drinking water pH, electroconductivity), our group across 250,000 km2 Michigan, USA. proposed achieved accuracy scores up 0.97 consistently outperformed state-of-the-art anomaly techniques.

Language: Английский

---