Journal of the Mechanics and Physics of Solids, Год журнала: 2025, Номер unknown, С. 106122 - 106122
Опубликована: Март 1, 2025
Язык: Английский
Journal of the Mechanics and Physics of Solids, Год журнала: 2025, Номер unknown, С. 106122 - 106122
Опубликована: Март 1, 2025
Язык: Английский
Advanced Engineering Informatics, Год журнала: 2023, Номер 56, С. 102007 - 102007
Опубликована: Апрель 1, 2023
Язык: Английский
Процитировано
164Computers in Biology and Medicine, Год журнала: 2023, Номер 158, С. 106734 - 106734
Опубликована: Март 1, 2023
Язык: Английский
Процитировано
66Neural Networks, Год журнала: 2023, Номер 162, С. 472 - 489
Опубликована: Март 13, 2023
Язык: Английский
Процитировано
59Thin-Walled Structures, Год журнала: 2024, Номер 205, С. 112495 - 112495
Опубликована: Сен. 24, 2024
Язык: Английский
Процитировано
47Engineering Applications of Artificial Intelligence, Год журнала: 2023, Номер 127, С. 107302 - 107302
Опубликована: Ноя. 8, 2023
Glacio-hydrological modeling is a key task for assessing the influence of snow and glaciers on water resources, essential resources management. The present study aims to enhance conceptual hydrological model (namely Glacial Snow Melt (GSM)) by data-driven swarm computing enhancing accuracy rainfall runoff prediction. proposed framework combines (i.e. GSM) with time series predictor (SVR) optimization-driven parameter tuning firefly algorithm (SVR-FFA). This integration uniquely captures complex interplay between meteorological variables, glacier processes, responses. Applying hybrid proved better results than standalone GSM ordinary SVR in simulating series. performance integrated metaheuristic-based (W-SG-SVR-FFA) demonstrated several enhancements over model. During calibration (validation) period, evaluation metric coefficient determination (R2) was 0.77 (0.77) 0.98 (0.91) W-SG-SVR-FFA Kling-Gupta Efficiency (KGE) values were 0.81 0.97 (0.87), respectively. method glacierized catchments underscores its importance areas undergoing swift climate change glacial melting. approach enables readers witness intricate equilibrium model's complexity simulation outcomes.
Язык: Английский
Процитировано
28npj Computational Materials, Год журнала: 2023, Номер 9(1)
Опубликована: Дек. 13, 2023
Abstract The design of materials and identification optimal processing parameters constitute a complex challenging task, necessitating efficient utilization available data. Bayesian Optimization (BO) has gained popularity in due to its ability work with minimal However, many BO-based frameworks predominantly rely on statistical information, the form input-output data, assume black-box objective functions. In practice, designers often possess knowledge underlying physical laws governing material system, rendering function not entirely black-box, as some information is partially observable. this study, we propose physics-informed BO approach that integrates physics-infused kernels effectively leverage both decision-making process. We demonstrate method significantly improves efficiency enables more data-efficient BO. applicability showcased through NiTi shape memory alloys, where are identified maximize transformation temperature.
Язык: Английский
Процитировано
28Materials Science and Engineering A, Год журнала: 2023, Номер 880, С. 145211 - 145211
Опубликована: Июнь 1, 2023
Язык: Английский
Процитировано
25Engineering Applications of Artificial Intelligence, Год журнала: 2024, Номер 133, С. 108055 - 108055
Опубликована: Фев. 26, 2024
In many cutting-edge applications, high-fidelity computational models prove to be too slow for practical use and are therefore replaced by much faster surrogate models. Recently, deep learning techniques have increasingly been utilized accelerate such predictions. To enable on large-dimensional complex data, specific neural network architectures developed, including convolutional graph networks. this work, we present a novel encoder–decoder geometric framework called MAgNET, which extends the well-known networks accommodate arbitrary graph-structured data. MAgNET consists of innovative Multichannel Aggregation (MAg) layers pooling/unpooling layers, forming U-Net architecture that is analogous U-Nets. We demonstrate predictive capabilities in modeling non-linear finite element simulations mechanics solids.
Язык: Английский
Процитировано
17Acta Materialia, Год журнала: 2024, Номер 266, С. 119645 - 119645
Опубликована: Янв. 2, 2024
Язык: Английский
Процитировано
16Engineering Applications of Artificial Intelligence, Год журнала: 2024, Номер 133, С. 108085 - 108085
Опубликована: Фев. 18, 2024
This paper proposes a novel framework for simulating the dynamics of beams on elastic foundations. Specifically, partial differential equations modeling Euler–Bernoulli and Timoshenko Winkler foundation are simulated using causal physics-informed neural network (PINN) coupled with transfer learning. Conventional PINNs encounter challenges in handling large space–time domains, even problems closed-form analytical solutions. A causality-respecting PINN loss function is employed to overcome this limitation, effectively capturing underlying physics. However, it observed that lacks generalizability. We propose solutions similar instead training from scratch by employing learning while adhering causality accelerate convergence ensure accurate results across diverse scenarios. The primary contribution lies introducing context structural engineering coupling enhance generalizability Numerical experiments beam highlight efficacy proposed approach various initial conditions, including those noise data. Furthermore, potential method demonstrated an extended spatial temporal domain. Several comparisons suggest accurately captures inherent dynamics, outperforming state-of-the-art methods under standard L2-norm metric accelerating convergence.
Язык: Английский
Процитировано
12