AlloyGAN: Domain-Promptable Generative Adversarial Network for Generating Aluminum Alloy Microstructures DOI
Biao Yin, Yangyang Fan, Nicholas Josselyn

et al.

Published: Dec. 15, 2023

The global metal market, expected to exceed $18.5 trillion by 2030, faces costly inefficiencies from defects in alloy manufacturing. Although microstructure analysis has improved performance, current numerical models struggle accurately simulate solidification. In this research, we thus introduce AlloyGAN - the first domain-driven Conditional Generative Adversarial Network (cGAN) involving domain prior for generating microstructures of previously not considered chemical and manufactural compositions. improves cGAN process factors solidification reaction generate scientifically valid images given basic manufacturing It achieves a faster equally accurate alternative traditional material science methods assessing microstructures. We contribute (1) novel Alloy-GAN design rapid optimization; (2) unique that inject knowledge into cGAN-based models; (3) metrics machine learning chemistry generation evaluation. Our approach highlights promise GAN-based scientific discovery materials. successfully transitioned an AIGC startup with core focus on model-generated metallography. open its interactive demo at: https://deepalloy.com/

Language: Английский

Microstructure reconstruction using diffusion-based generative models DOI
Kang‐Hyun Lee, Gun Jin Yun

Mechanics of Advanced Materials and Structures, Journal Year: 2023, Volume and Issue: 31(18), P. 4443 - 4461

Published: April 20, 2023

AbstractThis paper proposes a microstructure reconstruction framework with denoising diffusion models for the first time. The novelty and strength of proposed model lie in its universality generality characterization (MCR) that can be applied to various types composite materials. applicability diffusion-based is validated several microstructures (e.g., polycrystalline alloy, carbonate, ceramics, copolymer, fiber composite, etc.) have different morphological characteristics. Moreover, an implicit probabilistic (which yields non-Markovian processes) formulated accelerate sampling process, thereby controlling computational cost considering practicability reliability.Keywords: reconstructiondiffusion modeldenoising modelneural networkcomposite materials Data availability statementNo data was used research described article.Additional informationFundingThis material based upon work supported by Air Force Office Scientific Research under award number FA2386-22-1-4001 Institute Engineering at Seoul National University. authors are grateful their support.

Language: Английский

Citations

41
Geologically Constrained Convolutional Neural Network for Mineral Prospectivity Mapping DOI

Fanfan Yang,

Renguang Zuo

Mathematical Geosciences, Journal Year: 2024, Volume and Issue: 56(8), P. 1605 - 1628

Published: April 29, 2024

Language: Английский

Citations

13
Machine Learning in Petrology: State-of-the-Art and Future Perspectives DOI Creative Commons
Maurizio Petrelli

Journal of Petrology, Journal Year: 2024, Volume and Issue: 65(5)

Published: March 28, 2024

Abstract This article reports on the state-of-the-art and future perspectives of machine learning (ML) in petrology. To achieve this goal, it first introduces basics ML, including definitions, core concepts, applications. Then, starts reviewing ML Established applications mainly concern so-called data-driven discovery involve specific tasks like clustering, dimensionality reduction, classification, regression. Among them, clustering reduction have been demonstrated to be valuable for decoding chemical record stored igneous metamorphic phases enhance data visualization, respectively. Classification regression find applications, example, petrotectonic discrimination geo-thermobarometry, The main manuscript consists depicting emerging trends directions petrological investigations. I propose a scenario where methods will progressively integrate support established automating time-consuming repetitive tasks, improving current models, boosting discovery. In framework, promising include (1) acquisition new multimodal petrologic data; (2) development fusion techniques, physics-informed ML-supported numerical simulations; (3) continuous exploration potential boost contribution petrology, our challenges are: improve ability models capture complexity processes, link algorithms with physical thermodynamic nature investigated problems, start collaborative effort among researchers coming from different disciplines, both research teaching.

Language: Английский

Citations

11
Artificial intelligence in paleontology DOI Creative Commons
Congyu Yu, Fangbo Qin, Akinobu Watanabe

et al.

Earth-Science Reviews, Journal Year: 2024, Volume and Issue: 252, P. 104765 - 104765

Published: April 2, 2024

The accumulation of large datasets and increasing data availability have led to the emergence data-driven paleontological studies, which reveal an unprecedented picture evolutionary history. However, fast-growing quantity complication modalities make processing laborious inconsistent, while also lacking clear benchmarks evaluate collection generation, performances different methods on similar tasks. Recently, artificial intelligence (AI) has become widely practiced across scientific disciplines, but not so much date in paleontology where traditionally manual workflows been more usual. In this study, we review >70 AI studies since 1980s, covering major tasks including micro- macrofossil classification, image segmentation, prediction. These feature a wide range techniques such as Knowledge-Based Systems (KBS), neural networks, transfer learning, many other machine learning automate variety research workflows. Here, discuss their methods, datasets, performance compare them with conventional studies. We attribute recent increase most lowering entry bar training deployment models rather than innovations fossil compilation methods. present recently developed implementations diffusion model content generation Large Language Models (LLMs) that may interface future. Even though yet significant part paleontologist's toolkit, successful implementation is growing shows promise for paradigm-transformative effects years come.

Language: Английский

Citations

9
Synthesis of batik motifs using a diffusion - generative adversarial network DOI

One Octadion,

Novanto Yudistira, Diva Kurnianingtyas

et al.

Multimedia Tools and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 22, 2025

Language: Английский

Citations

1
FaciesViT: Vision transformer for an improved core lithofacies prediction DOI Creative Commons
Ardiansyah Koeshidayatullah, Sadam Al-Azani, E. E. Baraboshkin

et al.

Frontiers in Earth Science, Journal Year: 2022, Volume and Issue: 10

Published: Oct. 4, 2022

Lithofacies classification is a fundamental step to perform depositional and reservoir characterizations in the subsurface. However, such often hindered by limited data availability biased time-consuming analysis. Recent work has demonstrated potential of image-based supervised deep learning analysis, specifically convolutional neural networks (CNN), optimize lithofacies interpretation using core images. While most works have used transfer overcome datasets simultaneously yield high-accuracy prediction. This method raises some serious concerns regarding how CNN model learns makes prediction as was originally trained with entirely different datasets. Here, we proposed an alternative approach adopting vision transformer model, known FaciesViT , mitigate this issue provide improved We also experimented various architectures baseline models two compare evaluate performance our model. The experimental results show that significantly outperform established architecture for both all cases, achieving f1 score weighted average tested metrics 95%. For first time, study highlights application Vision Transformer geological dataset. Our findings several advantages over conventional models, including (i) no hyperparameter fine-tuning exhaustive augmentation required match accuracy models; (ii) it can datasets; (iii) better generalize new, unseen shows could further image recognition geosciences issues related generalizability explainability models. Furthermore, implementation been shown improve overall reproducibility which significant subsurface characterization basins worldwide.

Language: Английский

Citations

25
Generative Adversarial Network for Synthetic Image Generation Method: Review, Analysis, and Perspective DOI
Christine Dewi

Springer eBooks, Journal Year: 2024, Volume and Issue: unknown, P. 91 - 116

Published: Jan. 1, 2024

Language: Английский

Citations

4
Liquefaction susceptibility mapping using artificial neural network for offshore wind farms in Taiwan DOI
Chih‐Yu Liu, Cheng‐Yu Ku,

Ting-Yuan Wu

et al.

Engineering Geology, Journal Year: 2025, Volume and Issue: unknown, P. 108013 - 108013

Published: March 1, 2025

Language: Английский

Citations

0
Exploring the role of ore texture in comminution and approaches to identify and promote non-random breakage DOI Creative Commons

Carolina Månbro,

Mehdi Parian, Jan Rosenkranz

et al.

Minerals Engineering, Journal Year: 2025, Volume and Issue: 230, P. 109405 - 109405

Published: May 16, 2025

Language: Английский

Citations

0
Hierarchical multi-label taxonomic classification of carbonate skeletal grains with deep learning DOI

Madison Ho,

Sidhant Idgunji,

Jonathan L. Payne

et al.

Sedimentary Geology, Journal Year: 2022, Volume and Issue: 443, P. 106298 - 106298

Published: Nov. 18, 2022

Language: Английский

Citations

11