A Novel YOLOv10-DECA Model for Real-Time Detection of Concrete Cracks DOI Creative Commons

Chaokai Zhang,

Ningbo Peng, Jiaheng Yan

et al.

Buildings, Journal Year: 2024, Volume and Issue: 14(10), P. 3230 - 3230

Published: Oct. 11, 2024

The You Only Look Once (YOLO) series algorithms have been widely adopted in concrete crack detection, with attention mechanisms frequently being incorporated to enhance recognition accuracy and efficiency. However, existing research is confronted by two primary challenges: the suboptimal performance of mechanism modules lack explanation regarding how these influence model’s decision-making process improve accuracy. To address issues, a novel Dynamic Efficient Channel Attention (DECA) module proposed this study, which designed YOLOv10 model effectiveness visually demonstrated through application interpretable analysis algorithms. In paper, dataset complex background used. Experimental results indicate that DECA significantly improves localization detection discontinuous cracks, outperforming (ECA). When compared similarly sized YOLOv10n model, YOLOv10-DECA demonstrates improvements 4.40%, 3.06%, 4.48%, 5.56% precision, recall, mAP50, mAP50-95 metrics, respectively. Moreover, even when larger YOLOv10s indicators are increased 2.00%, 0.04%, 2.27%, 1.12%, terms speed evaluation, owing lightweight design module, achieves an inference 78 frames per second, 2.5 times faster than YOLOv10s, thereby fully meeting requirements for real-time detection. These demonstrate optimized balance between tasks has achieved model. Consequently, study provides valuable insights future applications field.

Language: Английский

Deep learning in computational mechanics: a review DOI Creative Commons
Leon Herrmann, Stefan Kollmannsberger

Computational Mechanics, Journal Year: 2024, Volume and Issue: 74(2), P. 281 - 331

Published: Jan. 13, 2024

Abstract The rapid growth of deep learning research, including within the field computational mechanics, has resulted in an extensive and diverse body literature. To help researchers identify key concepts promising methodologies this field, we provide overview deterministic mechanics. Five main categories are identified explored: simulation substitution, enhancement, discretizations as neural networks, generative approaches, reinforcement learning. This review focuses on methods rather than applications for thereby enabling to explore more effectively. As such, is not necessarily aimed at with knowledge learning—instead, primary audience verge entering or those attempting gain discussed are, therefore, explained simple possible.

Language: Английский

Citations

28
Topology optimization methods for thermal metamaterials: A review DOI
Wei Sha, Mi Xiao, Yihui Wang

et al.

International Journal of Heat and Mass Transfer, Journal Year: 2024, Volume and Issue: 227, P. 125588 - 125588

Published: April 24, 2024

Language: Английский

Citations

23
Data-driven reliability-based topology optimization by using the extended multi scale finite element method and neural network approach DOI
Zeng Meng,

Shunsheng Lv,

Yongxin Gao

et al.

Computer Methods in Applied Mechanics and Engineering, Journal Year: 2025, Volume and Issue: 438, P. 117837 - 117837

Published: Feb. 17, 2025

Language: Английский

Citations

2
Machine learning in solid mechanics: Application to acoustic metamaterial design DOI Creative Commons
D. Yago, G. Sal‐Anglada, D. Roca

et al.

International Journal for Numerical Methods in Engineering, Journal Year: 2024, Volume and Issue: 125(14)

Published: April 5, 2024

Abstract Machine learning (ML) and Deep (DL) are increasingly pivotal in the design of advanced metamaterials, seamlessly integrated with material or topology optimization. Their intrinsic capability to predict interconnect properties across vast spaces, often computationally prohibitive for conventional methods, has led groundbreaking possibilities. This paper introduces an innovative machine approach optimization acoustic focusing on Multiresonant Layered Acoustic Metamaterial (MLAM), designed targeted noise attenuation at low frequencies (below 1000 Hz). method leverages ML create a continuous model Representative Volume Element (RVE) effective essential evaluating sound transmission loss (STL), subsequently used optimize overall configuration maximum using Genetic Algorithm (GA). The significance this methodology lies its ability deliver rapid results without compromising accuracy, significantly reducing computational overhead complete by several orders magnitude. To demonstrate versatility scalability approach, it is extended more intricate RVE model, characterized higher number parameters, optimized same strategy. In addition, underscore potential techniques synergy traditional optimization, comparative analysis conducted, comparing outcomes proposed those obtained through direct numerical simulation (DNS) corresponding full 3D MLAM model. highlights transformative combination, particularly when addressing complex topological challenges significant demands, ushering new era metamaterial component design.

Language: Английский

Citations

9
Dynamically configured physics-informed neural network in topology optimization applications DOI
Jichao Yin, Ziming Wen, Shuhao Li

et al.

Computer Methods in Applied Mechanics and Engineering, Journal Year: 2024, Volume and Issue: 426, P. 117004 - 117004

Published: April 26, 2024

Language: Английский

Citations

9
Data-driven topology optimization of mechanical metamaterials via deep neural network and material-field function DOI

Zhengtong Han,

Ze Xu, Yang Zhou

et al.

Mechanics of Advanced Materials and Structures, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 17

Published: Feb. 12, 2025

Data-driven methods offer an innovative way to explore high-performance mechanical metamaterials, accelerating their engineering applications. However, most existing approaches use image pixel values (e.g. element densities) as input, leading the curse of dimensionality, resulting in high storage, memory demands, computational costs, and long training times. This article presents a novel lightweight data-driven approach using material field series expansion (MFSE) function deep neural network (DNN) non-iteratively design optimal metamaterials. By describing distribution with material-field instead elemental densities, number variables is significantly reduced. A multi-layer perceptron was trained map coefficients boundary conditions, principal component analysis (PCA) applied reduce output dimensions. Once trained, instantly generates topology optimization designs for optimizing bulk modulus, shear or minimizing Poisson's ratio (PR), demonstrated through numerical examples. The proposed method achieves accuracy minimal amount data. Compared density-based models, MFSE-DNN reduces time, allowing on personal PCs lower resources. not limited studied metamaterial can be further extended various metamaterials extreme specific functionalities.

Language: Английский

Citations

1
Machine learning in design for additive manufacturing: A state-of-the-art discussion for a support tool in product design lifecycle DOI Creative Commons
Michele Trovato, Luca Belluomo, Michele Bici

et al.

The International Journal of Advanced Manufacturing Technology, Journal Year: 2025, Volume and Issue: unknown

Published: March 5, 2025

Language: Английский

Citations

1
Beyond the limits of parametric design: Latent space exploration strategy enabling ultra-broadband acoustic metamaterials DOI

Min Woo Cho,

Seok Hyeon Hwang,

Jun-Young Jang

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 133, P. 108595 - 108595

Published: May 15, 2024

Language: Английский

Citations

8
Nature’s Load-Bearing Design Principles and Their Application in Engineering: A Review DOI Creative Commons
Firas Breish, Christian Hamm, Simone Andresen

et al.

Biomimetics, Journal Year: 2024, Volume and Issue: 9(9), P. 545 - 545

Published: Sept. 9, 2024

Biological structures optimized through natural selection provide valuable insights for engineering load-bearing components. This paper reviews six key strategies evolved in nature efficient mechanical load handling: hierarchically structured composites, cellular structures, functional gradients, hard shell–soft core architectures, form follows function, and robust geometric shapes. The also discusses recent research that applies these to design, demonstrating their effectiveness advancing technical solutions. challenges of translating nature’s designs into applications are addressed, with a focus on how advancements computational methods, particularly artificial intelligence, accelerating this process. need further development innovative material characterization techniques, modeling approaches heterogeneous media, multi-criteria structural optimization advanced manufacturing techniques capable achieving enhanced control across multiple scales is underscored. By highlighting holistic approach designing components, advocates adopting similarly comprehensive methodology practices shape the next generation

Language: Английский

Citations

6
Establishing Efficacy of Machine Learning Techniques for Vulnerability Information of Tubular Buildings DOI Open Access
Muhammad Zain, Suraparb Keawsawasvong, Chanachai Thongchom

et al.

Engineered Science, Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 1, 2023

During recent times, the emergence of artificial intelligence in structural engineering has rendered researchers to work on reducing overall computational effort required for producing vulnerability information infrastructural facilities.However, supertall and tubular building analysis lacks substantial research due their intricate behavior aleatory uncertainties.This paper establishes feasibility using versatile Machine Learning (ML) algorithms fragility relationships high-rise structures by considering a 55-story tall building, located high seismicity area.Initially, vibrational modes are decoupled, Incremental Dynamic Analysis (IDA) been conducted each individual mode discreetly.The initial four were considered analysis, constituting modal mass participation more than 90%.Inference drawn between efficacy employed ML techniques establish grounds rapid assessment buildings ground motion features along with characteristics.Testing datasets have suggested adequacy substantiating successful prediction Engineering Demand Parameter (EDP), applicability establishing structures.

Language: Английский

Citations

15