Deep Multiscale Soft-Threshold Support Vector Data Description for Enhanced Heavy-Duty Gas Turbine Generator Sets’ Anomaly Detection DOI Creative Commons
Zhang Kun, Li Hongren, Wang Xin

et al.

Shock and Vibration, Journal Year: 2024, Volume and Issue: 2024, P. 1 - 16

Published: April 29, 2024

This paper introduces an innovative approach, Deep Multiscale Soft-Threshold Support Vector Data Description (DMS-SVDD), designed for the detection of anomalies and prediction faults in heavy-duty gas turbine generator sets (GENSETs). The model combines a support vector data description (SVDD) with deep autoencoder backbone network framework, integrating multiscale convolutional neural (M) soft-threshold activation (S) into Deep-SVDD framework. In comparison conventional methods, such as One-Class Machine (OCSVM) (AE), DMS-SVDD demonstrates improvements accuracy (by 22.94%), recall 32%), F1 score 12.02%), smoothness 39.15%). excels particularly feature extraction, denoising, early fault detection, offering proactive strategy maintenance. Furthermore, demonstrated enhanced training efficiency reduction convergence rounds by 66% overall times 34.13%. study concludes that presents robust efficient solution anomaly practical advantages decision Future research could explore additional refinements applications across diverse industrial contexts.

Language: Английский

Advances in natural language processing for healthcare: A comprehensive review of techniques, applications, and future directions DOI

Fatmah Alafari,

Maha Driss, Asma Cherif

et al.

Computer Science Review, Journal Year: 2025, Volume and Issue: 56, P. 100725 - 100725

Published: Feb. 6, 2025

Language: Английский

Citations

1
Detection of Anomalies in Data Streams Using the LSTM-CNN Model DOI Creative Commons
Agnieszka Duraj, Piotr S. Szczepaniak,

Artur Sadok

et al.

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1610 - 1610

Published: March 6, 2025

This paper presents a comparative analysis of selected deep learning methods applied to anomaly detection in data streams. The results obtained on the popular Yahoo! Webscope S5 dataset are used for computational experiments. two commonly and recommended models literature, which basis this analysis, following: LSTM its more complicated variant, autoencoder. Additionally, usefulness an innovative LSTM-CNN approach is evaluated. indicate that can successfully be streams as performance compares favorably with mentioned standard models. For evaluation, F1score used.

Language: Английский

Citations

1
An Innovative IoT and Edge Intelligence Framework for Monitoring Elderly People Using Anomaly Detection on Data from Non-Wearable Sensors DOI Creative Commons
Amir Ali, Teodoro Montanaro, Ilaria Sergi

et al.

Sensors, Journal Year: 2025, Volume and Issue: 25(6), P. 1735 - 1735

Published: March 11, 2025

The aging global population requires innovative remote monitoring systems to assist doctors and caregivers in assessing the health of elderly patients. Doctors often lack access continuous behavioral data, making it difficult detect deviations from normal patterns when patients arrive for a consultation. Without historical insights into common behaviors potential anomalies detected with unobtrusive techniques (e.g., non-wearable devices), timely informed medical interventions become challenging. To address this, we propose an edge-based Internet Things (IoT) framework that enables real-time anomaly detection using sensors By processing data locally, system minimizes privacy concerns ensures immediate availability, allowing healthcare professionals unusual early. employs advanced machine learning (ML) models identify may indicate risks. A prototype our has been developed test its feasibility demonstrate, through application two most frequently used ML models, i.e., isolation forest Long Short-Term Memory (LSTM) networks, can provide scalability, efficiency, reliability context care. Further, provided dashboard alerts longitudinal trends, facilitating proactive interventions. proposed approach improves responsiveness by providing instant patient behavior, more accurate diagnoses This study lays groundwork future advancements field offers valuable research community harness full combining edge computing, artificial intelligence (AI), IoT

Language: Английский

Citations

0
Analysis of long-term trends and 15-year predictions of smoking-related bladder cancer burden in china across different age and sex groups from 1990 to 2021 DOI Creative Commons

Jieming Zuo,

Junhao Chen, Zhiyong Tan

et al.

Discover Oncology, Journal Year: 2025, Volume and Issue: 16(1)

Published: March 27, 2025

Tobacco is a significant risk factor for bladder cancer, with notable disparities in smoking rates and cancer prevalence between sex. Our objective to assess the sex- age-specific burden of attributable China from 1990 2021, predict its future trends over next 15 years using GBD study data. All data were extracted 2021 study, utilizing metrics such as mortality rates, disability-adjusted life (DALYs), age-standardized (ASMR), DALY (ASDR) describe smoking-attributable China. We employed joinpoint age-period-cohort (APC) analysis methods elucidate epidemiological characteristics cancer. Frontier was used visually demonstrate potential reduction based on development level each country or region. applied ARIMA model fit years. From number deaths DALYs due significantly increased. However, ASMR ASDR decreased both sexs but males experiencing higher burden. Population aging drove decline ASDR, despite rising absolute DALYs. Joinpoint regression yielded average annual percentage changes (AAPC) - 1.23 1.38 rate change being lower than females. The impact age, period, cohort varied. There slight increase relative health inequality among countries different income levels. By 2036, smoking-related are expected continue decreasing, this trend more pronounced males. Over past three decades, has increased across age groups, while have shown declining trend, reflecting certain public progress. This especially evident primarily driven by population demographic effects. levels slightly projected particularly Therefore, precise prevention intervention strategies targeting groups essential further alleviate

Language: Английский

Citations

0
Adaptive Artificial Intelligence for Students with Specific Learning Disabilities in Reading Science Content DOI
Richard Lamb, Danielle Malone, Tosha L. Owens

et al.

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: May 2, 2025

Abstract The growing integration of generative artificial intelligence (AI) technologies, including systems such as ChatGPT, into educational environments in science presents new opportunities to support learning. However, mainstream AI tools often fail adequately assist students with specific learning disabilities reading, dyslexia. Students reading require specialized instruction tailored the unique challenges posed by difficulties comprehension, decoding, and retaining multi-step directions present complex texts. While current technologies can provide basic explanations, they lack real-time, adaptive guidance step-by-step feedback personalized individual learners. Additionally, predominantly text-based does not suit needs who benefit from interactive, multimodal strategies visual aids. To better serve neurodiverse learners classrooms, must evolve a focus on inclusivity. Potential improvements include algorithms based upon use neurological data, enhanced formative assessment techniques, incorporation graphics other multisensory features. With innovative designs that align principles universal learning, AI-based could individualized skill development for all students. This will sustained efforts develop is responsive diverse needs.

Language: Английский

Citations

0
A Comparative Study of Deep-Learning Autoencoders (DLAEs) for Vibration Anomaly Detection in Manufacturing Equipment DOI Open Access
Seonwoo Lee, Akeem Bayo Kareem,

Jang-Wook Hur

et al.

Electronics, Journal Year: 2024, Volume and Issue: 13(9), P. 1700 - 1700

Published: April 27, 2024

Speed reducers (SR) and electric motors are crucial in modern manufacturing, especially within adhesive coating equipment. The motor mainly transforms electrical power into mechanical force to propel most machinery. Conversely, speed vital elements that control the torque of rotating machinery, ensuring optimal performance efficiency. Interestingly, variations chamber temperatures machines use specific adhesives can lead defects chains jigs, causing possible breakdowns reducer its surrounding components. This study introduces novel deep-learning autoencoder models enhance production efficiency by presenting a comparative assessment for anomaly detection would enable precise predictive insights modeling complex temporal relationships vibration data. data acquisition framework facilitated adherence governance principles maintaining quality consistency, storage processing operations, aligning with management standards. here capture attention practitioners involved data-centric processes, industrial engineering, advanced manufacturing techniques.

Language: Английский

Citations

3
Enhancing Cyberattack Detection Using Dimensionality Reduction With Hybrid Deep Learning on Internet of Things Environment DOI Creative Commons
Salahaldeen Duraibi, Abdullah Mujawib Alashjaee

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 84752 - 84762

Published: Jan. 1, 2024

Language: Английский

Citations

2
Isolation Forest Anomaly Detection in Vital Sign Monitoring for Healthcare DOI
Kanchan Yadav,

Upendra Singh Aswal,

V. Saravanan

et al.

Published: Dec. 29, 2023

The use of the isolate forest technique for recognizing anomalies in monitoring vital signs healthcare is examined this work. A deductive approach, based on interpretivism, uses secondary data along with a descriptive design. procedure's strong metrics performance are demonstrated by results, wherein effective identification indicated high precision, recollection, and AUC numbers. Its advantage over conventional methods comparisons. impact parameter tuning discussed, highlighting careful balancing act between mathematical efficiency accuracy. Opportunities issues can be qualitatively understood through assessment that have been detected. Improvements to interpretability, validation results medical professionals, refinement among suggestions put forward. Parameter improvement, understanding, real-world verification, combined models should main areas future research.

Language: Английский

Citations

4
Digital technology in occupational health of manufacturing industries: a systematic literature review DOI Creative Commons

Luping Jiang,

Jingdong Zhang, Yiik Diew Wong

et al.

Deleted Journal, Journal Year: 2024, Volume and Issue: 6(12)

Published: Nov. 22, 2024

In this study, we fill the gap of limited effort on systematic literature review into field digital technology for occupational health manufacturing industries. Upon reviewing 53 publications selected by combined bibliometric and classical methods, present an integrated overview major research areas hot topics critically identify prevalent technologies application modes, enablers barriers to implementation, as well agenda in health. The results show that, with increasing popularity penetration items like wearable devices sensors, human–robot collaboration, deep learning analytics, identified implementation are: intelligent manufacturing, competitive condition, data-driven decision-making tool, considerations welfare health; technological gap, privacy data security, culture acceptance, cost consideration. Additionally, propositions three aspects six perspectives are recommended future field. Overall, study provides insights through analysis synthesis, offers means achieve Sustainable Development Goals (SDGs) exploring efficient protect labor rights improve

Language: Английский

Citations

1
Deep Multiscale Soft-Threshold Support Vector Data Description for Enhanced Heavy-Duty Gas Turbine Generator Sets’ Anomaly Detection DOI Creative Commons
Zhang Kun, Li Hongren, Wang Xin

et al.

Shock and Vibration, Journal Year: 2024, Volume and Issue: 2024, P. 1 - 16

Published: April 29, 2024

This paper introduces an innovative approach, Deep Multiscale Soft-Threshold Support Vector Data Description (DMS-SVDD), designed for the detection of anomalies and prediction faults in heavy-duty gas turbine generator sets (GENSETs). The model combines a support vector data description (SVDD) with deep autoencoder backbone network framework, integrating multiscale convolutional neural (M) soft-threshold activation (S) into Deep-SVDD framework. In comparison conventional methods, such as One-Class Machine (OCSVM) (AE), DMS-SVDD demonstrates improvements accuracy (by 22.94%), recall 32%), F1 score 12.02%), smoothness 39.15%). excels particularly feature extraction, denoising, early fault detection, offering proactive strategy maintenance. Furthermore, demonstrated enhanced training efficiency reduction convergence rounds by 66% overall times 34.13%. study concludes that presents robust efficient solution anomaly practical advantages decision Future research could explore additional refinements applications across diverse industrial contexts.

Language: Английский

Citations

0