A survey on machine learning approaches for uncertainty quantification of engineering systems

Machine learning for computational science and engineering, Год журнала: 2025, Номер 1(1)

Опубликована: Янв. 30, 2025

Язык: Английский

---

Multivariate engineering formulas discovery with knowledge‐based neural network

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Фев. 26, 2025

Abstract Multivariate engineering formulas are the foundation of various standards worldwide for constructing complex systems. Traditional formula discovery methods suffer from low efficiency, curse dimensionality, and physical interpretability. To address these limitations, this study proposes a knowledge‐based method efficiently generating multivariate directly data. The consists four components: (1) deep generative model considering dimensional homogeneity, (2) physics‐adaptive normalization multiple variables with different units, (3) feature merging algorithm grounded in dimensionality theory, (4) machine learning‐based data segmentation piecewise formulas. Experiments on two ground‐truth datasets demonstrate that our proposed improves accuracy generated by 35.6% (measured mean absolute error), compared to Eureqa program. Additionally, it enhances mechanistic interpretability results, both emerging physics‐informed neural network‐based equation methods. successfully capture implicit mechanisms experimental data, consistent theoretical analysis. Overall, holds great promise improving efficiency discovering interpretable generalizable formulas, facilitating transformation new techniques testing applications.

Язык: Английский

---

Geometry physics neural operator solver for solid mechanics

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Янв. 3, 2025

Abstract This study developed Geometry Physics neural Operator (GPO), a novel solver framework to approximate the partial differential equation (PDE) solutions for solid mechanics problems with irregular geometry and achieved significant speedup in simulation time compared numerical solvers. GPO leverages weak form of PDEs based on principle least work, incorporates information, imposes exact Dirichlet boundary conditions within network architecture attain accurate efficient modeling. focuses applying model behaviors complicated bodies without any guided or labeled training data. adopts modified Fourier operator as backbone achieve significantly improved convergence speed learn solution field problems. Numerical experiments involved two‐dimensional plane hole three‐dimensional building structure constraints. The results indicate that layer constraints contribute accuracy speed, outperforming previous benchmark simulations geometry. comparison also showed can converge fields faster than commercial structural examples. Furthermore, demonstrates stronger performance solvers when mesh size is smaller, it achieves over 3 2 large degree freedom examples, respectively. limitations nonlinearity structures are further discussed prospective developments. remarkable suggest potential modeling applications large‐scale infrastructures.

Язык: Английский

---

Hybrid physics‐informed neural network with parametric identification for modeling bridge temperature distribution

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Фев. 18, 2025

Abstract This paper introduces a novel hybrid multi‐model thermo‐temporal physics‐informed neural network (TT‐PINN) framework for thermal loading prediction in composite bridge decks. Unlike the existing PINN applications heat transfer that focus on simple geometries, this uniquely addresses multi‐material domains and realistic boundary conditions through dual‐network architecture designed structures. The further incorporates environmental of natural convection solar radiation into loss function employs learning efficient adaptation to varying conditions. Moreover, mechanism enables rapid new states, thus markedly reducing computations as compared conventional finite element method (FEM). Through noise‐augmented training parameter identification, TT‐PINN effectively handles real‐world monitoring data uncertainties allows material property calibration with limited sensor data. framework's ability capture complex behavior is validated by studying cable‐stayed bridge. It significantly reduces computational costs traditional FEM approaches.

Язык: Английский

---

Damage detection for railway bridges using time‐frequency decomposition and conditional generative model

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 11, 2024

Abstract A novel damage detection model, which utilizes the spatiotemporal characteristics of acceleration data, is proposed to assess structural integrity railway bridges. For this, measured data are decomposed into several intrinsic mode functions (IMFs) using sparse random decomposition model. The generated IMFs subsequently integrated enhanced time series conditional generative adversarial network model identify possible in bridges across various frequency bands. influence environmental and operational variables (EOVs), particularly temperature fluctuations, was also investigated. verified both numerical experimental from a plate girder bridge. Further validation conducted Z24 bridge dataset, cases under EOVs were successfully predicted. Throughout process, anomaly metrics introduced establish threshold value, covariance‐based domain metric proven be most effective our cases.

Язык: Английский

---

Mainshock–aftershock sequence simulation via latent space encoding of generative adversarial networks

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2024, Номер unknown

Опубликована: Сен. 29, 2024

Abstract Aftershocks (ASs) following strong mainshocks (MSs) can exacerbate structural damage or lead to collapse. However, the scarcity of recorded data necessitates reliance on artificial sequences, which have difficulty in characterizing time‐frequency correlation between MSs and ASs. This study innovatively converts AS time history prediction into an image translation task, exploiting invertible transformation accelerograms representations. An encoder–decoder neural network is developed encode MS information latent space a pre‐trained generative adversarial network, enabling accurate predictions through decoder. The integration seismic parameters further improves performance. Comparative analyses demonstrate that proposed method outperforms traditional ones accuracy robustness reproduces non‐stationarity

Язык: Английский

---

Automatic classification of near‐fault pulse‐like ground motions

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2024, Номер unknown

Опубликована: Дек. 24, 2024

Abstract This study presents an automated, quantitative classification method for near‐fault pulse‐like ground motions, distinguishing between forward‐directivity and fling‐step (FS) motions. The introduces two novel parameters—the pulse velocity ratio area ratio—which transform the standard from a qualitative to framework. Combined with enhanced extraction technique that captures permanent displacement characteristics, these parameters significantly improve efficiency repeatability. automated approach overcomes limitations of manual classification, providing reproducible results. identified FS motions can be applied dynamic analysis cross‐fault structures, enhancing reliability seismic hazard assessments.

Язык: Английский

---