Data-Driven Structural Health Monitoring and Damage Detection through Deep Learning: State-of-the-Art Review

Sensors, Journal Year: 2020, Volume and Issue: 20(10), P. 2778 - 2778

Published: May 13, 2020

Data-driven methods in structural health monitoring (SHM) is gaining popularity due to recent technological advancements sensors, as well high-speed internet and cloud-based computation. Since the introduction of deep learning (DL) civil engineering, particularly SHM, this emerging promising tool has attracted significant attention among researchers. The main goal paper review latest publications SHM using DL-based provide readers with an overall understanding various applications. After a brief introduction, overview DL (e.g., neural networks, transfer learning, etc.) presented. procedure application vibration-based, vision-based monitoring, along some technologies used for such unmanned aerial vehicles (UAVs), etc. are discussed. concludes prospects potential limitations

Language: Английский

---

Artificial intelligence based anomaly detection of energy consumption in buildings: A review, current trends and new perspectives

Applied Energy, Journal Year: 2021, Volume and Issue: 287, P. 116601 - 116601

Published: Feb. 9, 2021

Enormous amounts of data are being produced everyday by sub-meters and smart sensors installed in residential buildings. If leveraged properly, that could assist end-users, energy producers utility companies detecting anomalous power consumption understanding the causes each anomaly. Therefore, anomaly detection stop a minor problem becoming overwhelming. Moreover, it will aid better decision-making to reduce wasted promote sustainable efficient behavior. In this regard, paper is an in-depth review existing frameworks for building based on artificial intelligence. Specifically, extensive survey presented, which comprehensive taxonomy introduced classify algorithms different modules parameters adopted, such as machine learning algorithms, feature extraction approaches, levels, computing platforms application scenarios. To best authors' knowledge, first article discusses consumption. Moving forward, important findings along with domain-specific problems, difficulties challenges remain unresolved thoroughly discussed, including absence of: (i) precise definitions consumption, (ii) annotated datasets, (iii) unified metrics assess performance solutions, (iv) reproducibility (v) privacy-preservation. Following, insights about current research trends discussed widen applications effectiveness technology before deriving future directions attracting significant attention. This serves reference understand technological progress

Language: Английский

---

Deep Learning for Anomaly Detection in Time-Series Data: Review, Analysis, and Guidelines

IEEE Access, Journal Year: 2021, Volume and Issue: 9, P. 120043 - 120065

Published: Jan. 1, 2021

As industries become automated and connectivity technologies advance, a wide range of systems continues to generate massive amounts data. Many approaches have been proposed extract principal indicators from the vast sea data represent entire system state. Detecting anomalies using these on time prevent potential accidents economic losses. Anomaly detection in multivariate series poses particular challenge because it requires simultaneous consideration temporal dependencies relationships between variables. Recent deep learning-based works made impressive progress this field. They are highly capable learning representations large-scaled sequences an unsupervised manner identifying However, most them specific individual use case thus require domain knowledge for appropriate deployment. This review provides background anomaly time-series reviews latest applications real world. Also, we comparatively analyze state-of-the-art deep-anomaly-detection models with several benchmark datasets. Finally, offer guidelines model selection training strategy detection.

Language: Английский

---

Machine learning paradigm for structural health monitoring

Structural Health Monitoring, Journal Year: 2020, Volume and Issue: 20(4), P. 1353 - 1372

Published: Nov. 24, 2020

Structural health diagnosis and prognosis is the goal of structural monitoring. Vibration-based monitoring methodology has been extensively investigated. However, conventional vibration–based methods find it difficult to detect damages actual structures because a high incompleteness in information (the number sensors much fewer with respect degrees freedom structure), intense uncertainties conditions systems, coupled effects damage environmental actions on modal parameters. It truth that performance structure must be embedded data (vehicles, wind, etc.; acceleration, displacement, cable force, strain, images, videos, etc.). Therefore, there need develop completely novel based various data. Machine learning provides advanced mathematical frameworks algorithms can help discover model through deep mining Thus, machine takes an opportunity establish paradigm for theory termed This article sheds light principles some examples reviews existing challenges open questions this field.

Language: Английский

---

Deep learning for data anomaly detection and data compression of a long‐span suspension bridge

Computer-Aided Civil and Infrastructure Engineering, Journal Year: 2019, Volume and Issue: 35(7), P. 685 - 700

Published: Dec. 27, 2019

Abstract As intelligent sensing and sensor network systems have made progress low‐cost online structural health monitoring has become possible widely implemented, large quantities of highly heterogeneous data can be acquired during the monitoring. This resulted in exceeding capacity traditional analytics techniques, especially large‐scale or critical civil structures. In particular, storage a big challenge, hence, resulting emergence compression reconstruction as new area (SHM) infrastructure systems. SHM generally include anomalies that disturb analysis assessment. The fundamental reasons for abnormality are extremely complex. Therefore, abnormal is difficult poses serious challenges to achieve high‐accuracy after been compressed. Considering these significant challenges, this paper, novel deep‐learning‐enabled framework proposed divided into two phases: (a) one‐dimensional Convolutional Neural Network (CNN) extracts features directly from input signals designed detect with validated high accuracy; (b) method based on Autoencoder structure further developed, which recover under such low ratio. To validate approach, acceleration system long‐span bridge China employed. detection phase, results show anomaly accuracy. Subsequently, smaller errors achieved even by using only 10% ratio normal data.

Language: Английский

---

Vibration feature extraction using signal processing techniques for structural health monitoring: A review

Mechanical Systems and Signal Processing, Journal Year: 2022, Volume and Issue: 177, P. 109175 - 109175

Published: May 4, 2022

Language: Английский

---

Integrated structural health monitoring in bridge engineering

Automation in Construction, Journal Year: 2022, Volume and Issue: 136, P. 104168 - 104168

Published: Feb. 23, 2022

Language: Английский

---

Structural health monitoring using extremely compressed data through deep learning

Computer-Aided Civil and Infrastructure Engineering, Journal Year: 2019, Volume and Issue: 35(6), P. 597 - 614

Published: Nov. 22, 2019

Abstract This study introduces a novel convolutional neural network (CNN)‐based approach for structural health monitoring (SHM) that exploits form of measured compressed response data through transfer learning (TL)‐based techniques. The implementation the proposed methodology allows damage identification and localization within realistic large‐scale system. To validate method, first, well‐known benchmark model is numerically simulated. Using acceleration histories, as well in terms discrete histograms, CNN models are trained, robustness architectures evaluated. Finally, pretrained CNNs fine‐tuned to be adaptable three‐parameter, extremely data, based on mean, standard deviation, scale factor. performance each assessed using training accuracy histories confusion matrices, along with other metrics. In addition numerical study, method demonstrated experimental vibration verification validation. results indicate deep TL can implemented effectively SHM similar systems different types sensors.

Language: Английский

---

Toward data anomaly detection for automated structural health monitoring: Exploiting generative adversarial nets and autoencoders

Structural Health Monitoring, Journal Year: 2020, Volume and Issue: 20(4), P. 1609 - 1626

Published: June 7, 2020

Damage detection is one of the most important tasks for structural health monitoring civil infrastructure. Before a damage algorithm can be applied, integrity data must ensured; otherwise results may misleading or incorrect. Indeed, sensor system malfunction, which in anomalous (often called faulty data), serious problem, as sensors usually operate extremely harsh environments. Identifying and eliminating anomalies crucial to ensuring that reliable achieved. Because vast amounts typically collected by system, manual removal prohibitive. Machine learning methods have potential automate process anomaly detection. Although supervised been proven effective detecting anomalies, two unresolved challenges reduce accuracy detection: (1) class imbalance (2) incompleteness patterns training dataset. Unsupervised address these challenges, but improvements are required deal with data. In this article, generative adversarial networks combined widely applied unsupervised method, is, autoencoders, improve performance existing methods. addition, time-series transformed Gramian Angular Field images so advanced computer vision included network. Two datasets from full-scale bridge, including examples caused malfunctions, utilized validate proposed methodology. Results show methodology successfully identify good robustness, hence overcome key difficulties achieving automated monitoring.

Language: Английский

---

Lost data reconstruction for structural health monitoring using deep convolutional generative adversarial networks

Structural Health Monitoring, Journal Year: 2020, Volume and Issue: 20(4), P. 2069 - 2087

Published: Oct. 10, 2020

In the application of structural health monitoring, measured data might be temporarily or permanently lost due to sensor fault transmission failure. The with a high loss ratio undermine its ability for modal identifications and condition evaluations. To reconstruct in field this study proposes deep convolutional generative adversarial network which includes generator encoder–decoder structure an discriminator. proposed model needs understand content complete signals, as well produce realistic hypotheses signals. Given stably before occurrence loss, is trained extract features maintained set signals using responses remaining functional sensors alone. discriminator feeds back distinguished results improve reconstruction accuracy. When training model, are employed better handle low-frequency high-frequency effectiveness efficiency method validated by two case studies. As number epoch increases, reconstructed learn from high-frequency, amplitude gradually increases. It can seen that final match real time domain frequency domain. further demonstrate applicability analysis, acceleration used accurately identify parameters numerical case, vehicle-induced precisely decomposed strain case. Finally, capacity also investigated different numbers faulted gauges.

Language: Английский

---