DeepOKAN: Deep operator network based on Kolmogorov Arnold networks for mechanics problems

Computer Methods in Applied Mechanics and Engineering, Год журнала: 2025, Номер 436, С. 117699 - 117699

Опубликована: Янв. 14, 2025

Язык: Английский

---

Sequential Deep Operator Networks (S-DeepONet) for predicting full-field solutions under time-dependent loads

Engineering Applications of Artificial Intelligence, Год журнала: 2023, Номер 127, С. 107258 - 107258

Опубликована: Окт. 11, 2023

Язык: Английский

---

Prediction of thermal runaway for a lithium-ion battery through multiphysics-informed DeepONet with virtual data

eTransportation, Год журнала: 2024, Номер 21, С. 100337 - 100337

Опубликована: Май 9, 2024

Язык: Английский

---

Rethinking materials simulations: Blending direct numerical simulations with neural operators

npj Computational Materials, Год журнала: 2024, Номер 10(1)

Опубликована: Июль 4, 2024

Abstract Materials simulations based on direct numerical solvers are accurate but computationally expensive for predicting materials evolution across length- and time-scales, due to the complexity of underlying equations, nature multiscale spatiotemporal interactions, need reach long-time integration. We develop a method that blends with neural operators accelerate such simulations. This methodology is integration community solver U-Net operator, enhanced by temporal-conditioning mechanism enable extrapolation efficient time-to-solution predictions dynamics. demonstrate effectiveness this hybrid framework microstructure via phase-field method. Such exhibit high spatial gradients co-evolution different material phases simultaneous slow fast establish coupled large speed-up compared DNS depending strategy utilized. generalizable broad range simulations, from solid mechanics fluid dynamics, geophysics, climate, more.

Язык: Английский

---

Virtual sensing-enabled digital twin framework for real-time monitoring of nuclear systems leveraging deep neural operators

npj Materials Degradation, Год журнала: 2025, Номер 9(1)

Опубликована: Март 6, 2025

Язык: Английский

---

A thermodynamics-informed deep learning approach for lightweight modeling of gas turbine performance

Engineering Applications of Artificial Intelligence, Год журнала: 2025, Номер 143, С. 110022 - 110022

Опубликована: Янв. 15, 2025

Язык: Английский

---

StressD: 2D Stress estimation using denoising diffusion model

Computer Methods in Applied Mechanics and Engineering, Год журнала: 2023, Номер 416, С. 116343 - 116343

Опубликована: Авг. 22, 2023

Finite element analysis (FEA), a common approach for simulating stress distribution given geometry, is generally associated with high computational cost, especially when mesh resolution required. Furthermore, the non-adaptive nature of FEA requires entire model to be solved even minor geometric variations creating bottleneck during iterative design optimization. This necessitates framework that can efficiently predict in geometries based on boundary and loading conditions. In this paper, we present StressD, predicting von Mises fields denoising diffusion model. The StressD involves two models, U-net-based an auxiliary network generate structures. generates normalized map conditions condition, while used determine scaling information needed un-normalize generated map. We evaluate cantilever structures show it able accurately significantly reducing cost compared traditional FEA.

Язык: Английский

---

A finite element-based physics-informed operator learning framework for spatiotemporal partial differential equations on arbitrary domains

Engineering With Computers, Год журнала: 2024, Номер unknown

Опубликована: Авг. 2, 2024

Abstract We propose a novel finite element-based physics-informed operator learning framework that allows for predicting spatiotemporal dynamics governed by partial differential equations (PDEs). The Galerkin discretized weak formulation is employed to incorporate physics into the loss function, termed (FOL), along with implicit Euler time integration scheme temporal discretization. A transient thermal conduction problem considered benchmark performance, where FOL takes temperature field at current step as input and predicts next step. Upon training, network successfully evolution over any initial high accuracy compared solution element method (FEM) even heterogeneous conductivity arbitrary geometry. advantages of can be summarized follows: First, training performed in an unsupervised manner, avoiding need large data prepared from costly simulations or experiments. Instead, random patterns generated Gaussian process Fourier series, combined constant fields, are used cover possible cases. Additionally, shape functions backward difference approximation exploited domain discretization, resulting purely algebraic equation. This enhances efficiency, one avoids time-consuming automatic differentiation optimizing weights biases while accepting discretization errors. Finally, thanks interpolation power FEM, geometry microstructure handled FOL, which crucial addressing various engineering application scenarios.

Язык: Английский

---

Prediction of 4D stress field evolution around additive manufacturing-induced porosity through progressive deep-learning frameworks

Machine Learning Science and Technology, Год журнала: 2024, Номер 5(1), С. 015038 - 015038

Опубликована: Фев. 13, 2024

Abstract This study investigates the application of machine learning models to predict time-evolving stress fields in complex three-dimensional structures trained with full-scale finite element simulation data. Two novel architectures, multi-decoder CNN (MUDE-CNN) and multiple encoder–decoder model transfer (MTED-TL), were introduced address challenge predicting progressive spatial evolutional distributions around defects. The MUDE-CNN leveraged a shared encoder for simultaneous feature extraction employed decoders distinct time frame predictions, while MTED-TL progressively transferred knowledge from one block another, thereby enhancing prediction accuracy through learning. These evaluated assess their accuracy, particular focus on temporal an additive manufacturing (AM)-induced isolated pore, as understanding such defects is crucial assessing mechanical properties structural integrity materials components fabricated via AM. evaluation demonstrated MTED-TL’s consistent superiority over MUDE-CNN, owing learning’s advantageous initialization weights smooth loss curves. Furthermore, autoregressive training framework was improve consistently outperforming both MTED-TL. By accurately AM-induced defects, these can enable real-time monitoring proactive defect mitigation during fabrication process. capability ensures enhanced component quality enhances overall reliability additively manufactured parts.

Язык: Английский

---

Advanced deep operator networks to predict multiphysics solution fields in materials processing and additive manufacturing

Additive manufacturing, Год журнала: 2024, Номер 88, С. 104266 - 104266

Опубликована: Май 1, 2024

Unlike classical artificial neural networks, which require retraining for each new set of parametric inputs, the Deep Operator Network (DeepONet), a lately introduced deep learning framework, approximates linear and nonlinear solution operators by taking functions (infinite-dimensional objects) as inputs mapping them to complete fields. In this paper, two newly devised DeepONet formulations with sequential Residual U-Net (ResUNet) architectures are trained first time simultaneously predict thermal mechanical fields under variable loading, loading histories, process parameters, even geometries. Two real-world applications demonstrated: 1- coupled thermo-mechanical analysis steel continuous casting multiple visco-plastic constitutive laws 2- sequentially direct energy deposition additive manufacturing. Despite highly challenging spatially target distributions, DeepONets can infer reasonably accurate full-field temperature stress solutions several orders magnitude faster than traditional optimized finite-element (FEA), when FEA simulations run on latest high-performance computing platforms. The proposed model's ability provide field predictions almost instantly unseen input parameters opens door future preliminary evaluation design optimization these vital industrial processes.

Язык: Английский

---