Article
Engineering, Mechanical
Navyanth Kusampudi, Martin Diehl
Summary: This study proposes a generative machine learning model for identifying low-dimensional descriptors of microstructural features and establishing structure-property relationships. Based on this model, an integrated framework for microstructure characterization, reconstruction, and design is presented for heterogeneous materials with polycrystalline microstructures.
INTERNATIONAL JOURNAL OF PLASTICITY
(2023)
Article
Automation & Control Systems
Haifei Peng, Jian Long, Cheng Huang, Shibo Wei, Zhencheng Ye
Summary: This paper proposes a novel multi-modal hybrid modeling strategy (GMVAE-STA) that can effectively extract deep multi-modal representations and complex spatial and temporal relationships, and applies it to industrial process prediction.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2024)
Article
Computer Science, Theory & Methods
Marno Basson, Tobias M. M. Louw, Theresa R. R. Smith
Summary: We propose a variational inference-based framework for training a Gaussian process regression model with censored observational data. The framework utilizes variational sparse Gaussian process inducing variable framework and local variational methods to compute an analytically tractable lower bound for Bayesian model training and inference. The proposed framework shows comparable predictions to existing methods while reducing computational cost significantly.
STATISTICS AND COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Lin Yang, Wentao Fan, Nizar Bouguila
Summary: This article proposes a clustering method based on variational autoencoder with spherical latent embeddings. The method improves clustering accuracy by using the von Mises-Fisher mixture model prior and a dual VAE structure, and enhances model robustness through an augmented loss function.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Xiaojie Xu, Yun Zhang
Summary: This study builds Gaussian process regression models to forecast the daily price index of ten major steel products in the Chinese market. The models provide accurate forecasts for the period of April 16, 2019 to April 15, 2021, with relative root mean square errors ranging from 0.07404% to 0.22379% and correlation coefficients above 99.9%. They outperform traditional econometric models and some other machine learning models as benchmarks.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Information Systems
Yang Huang, Duen-Ren Liu, Shin-Jye Lee, Chia-Hao Hsu, Yang-Guang Liu
Summary: Resampling is commonly used for imbalanced data, but there is limited research on imbalanced regression data. This study divides regression data into rare samples and normal samples, and proposes a boosting resampling method based on a conditional variational autoencoder.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Youngju Kim, Hoyeop Lee, Chang Ouk Kim
Summary: In this paper, a robust FD model is proposed by modeling process drift using VAE. The model successfully models process drift and outperforms four comparison FD methods on actual datasets.
JOURNAL OF INTELLIGENT MANUFACTURING
(2023)
Article
Automation & Control Systems
Dennis Nieman, Botond Szabo, Harry van Zanten
Summary: This study investigates the theoretical properties of the variational Bayes method in the Gaussian Process regression model. By considering the inducing variables method introduced by Titsias (2009b), we derive sufficient conditions for obtaining contraction rates for the corresponding variational Bayes posterior. Numerical experiments demonstrate the validity of the theoretical findings, showing that the VB approach can achieve optimal contraction rates for certain covariance kernels.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Computer Science, Interdisciplinary Applications
Maria Teresa Garcia-Ordas, Carmen Benavides, Jose Alberto Benitez-Andrades, Hector Alaiz-Moreton, Isaias Garcia-Rodriguez
Summary: This paper proposes a deep learning pipeline to predict diabetic people, achieving a high accuracy rate. The use of this method shows promising results in the field of diabetes detection, outperforming existing proposals.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Chemistry, Multidisciplinary
Johanna Kleinekorte, Jonas Kleppich, Lorenz Fleitmann, Verena Beckert, Luise Blodau, Andrei Bardow
Summary: A sustainable chemical industry needs to quantify its emissions and resource consumption by life cycle assessment (LCA), but detailed mass and energy balances are usually not available at early process development stages. To address this issue, a fully automated, predictive LCA framework (APPROPRIATE) based on Gaussian Process Regression is introduced, which is applicable at Technology Readiness Level 2. By employing an encoder-decoder network combined with transfer learning, the framework achieves a condensed molecular descriptor as a latent representation to overcome limited LCA data availability. The proposed framework also integrates process descriptors, such as the stoichiometric sum of the reactants' impacts, to distinguish between process alternatives and incorporate changes in the background systems. Compared to state-of-the-art predictive LCA approaches, the APPROPRIATE framework shows increased prediction accuracy, especially in terms of the global warming impact.
ACS SUSTAINABLE CHEMISTRY & ENGINEERING
(2023)
Article
Mathematical & Computational Biology
Hongyu Duan, Feng Li, Junliang Shang, Jinxing Liu, Yan Li, Xikui Liu
Summary: The development of single-cell technologies has led to a surge in research, particularly in the analysis of chromatin accessibility differences at the single-cell level using scATAC-seq. However, challenges in distinguishing cell types have emerged due to the increasing number of cells and data characteristics. We propose a method called scVAEBGM, which combines a Variational Autoencoder (VAE) with a Bayesian Gaussian-mixture model (BGM) to process and analyze scATAC-seq data. This method can estimate the number of cell types without prior knowledge and is more robust to noise and better represents single-cell data in lower dimensions.
INTERDISCIPLINARY SCIENCES-COMPUTATIONAL LIFE SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Francesco Camastra, Angelo Casolaro, Gennaro Iannuzzo
Summary: The paper introduces a novel manifold learning algorithm called the deep Gaussian process autoencoder (DPGA) based on deep Gaussian processes. The algorithm has two main characteristics: a bottleneck structure borrowed from variational autoencoders and the use of doubly stochastic variational inference for deep Gaussian processes architecture (DSVI). The main contributions of the paper are the DGPA algorithm itself and the introduction of the manifold learning performance protocol (MLPP) for evaluating it. Experimental tests on synthetic and real datasets demonstrate that the deep Gaussian process autoencoder compares favorably with other manifold learning algorithms.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Yi Shan Lee, Junghui Chen
Summary: This paper introduces a method called source-aided variational state-space autoencoder (SA-VSSAE) for monitoring industrial processes with sparse target data. The method enhances the reliability of the target model by performing information sharing. Compared to traditional state-space models, SA-VSSAE extracts dynamic and nonlinear features and incorporates process uncertainty in a probabilistic form.
Article
Computer Science, Artificial Intelligence
Borui Cai, Shuiqiao Yang, Longxiang Gao, Yong Xiang
Summary: This paper introduces a novel hybrid variational autoencoder (HyVAE) for forecasting time series by jointly learning the local patterns and temporal dynamics. Experimental results demonstrate that the proposed HyVAE achieves better results compared to other methods.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Daniele Gammelli, Kasper Pryds Rolsted, Dario Pacino, Filipe Rodrigues
Summary: In this paper, a new model is proposed to deal with censored observations. By exploiting correlations between multiple outputs, the model combines the flexibility of Gaussian process with the ability to leverage information from correlated outputs, and achieves better estimation of the true process.
PATTERN RECOGNITION
(2022)
Article
Materials Science, Multidisciplinary
Hyung Keun Park, Yongju Kim, Jeong Min Park, Kei Ameyama, Hyoung Seop Kim
Summary: A harmonic structured material (HSM) with coarse-grained core and fine-grained shell microstructure was designed using a numerical model to simulate the deformation behavior of heterostructured materials. The model considered the pile-up of geometrically necessary dislocations (GNDs) near grain boundaries to implement the hetero-deformation induced (HDI) strengthening feature. The finite element analysis supported the experimental results and described the strain partitioning near the core-shell boundaries in the HSM. The optimal harmonic microstructure was investigated using machine learning (ML), and the correlation between microstructures and mechanical properties was established through Bayesian inference. A new HSM, predicted to have the best combination of strength and ductility, was successfully manufactured and exhibited superior mechanical performance compared to previous designs.
Article
Materials Science, Multidisciplinary
Gang Hee Gu, Yongju Kim, Min Hong Seo, Hyoung Seop Kim
Summary: Plastic strain ratio, a material property that estimates the mechanical anisotropy and drawability of metal sheets, is needed even in large deformation regimes.
METALS AND MATERIALS INTERNATIONAL
(2023)
Article
Materials Science, Multidisciplinary
Soung Yeoul Ahn, Sang Guk Jeong, Eun Seong Kim, Suk Hoon Kang, Jungho Choe, Joo Young Ryu, Dae Woon Choi, Jin Seok Lee, Jung-Wook Cho, Takayoshi Nakano, Hyoung Seop Kim
Summary: Interest in recycling Zircaloy-4 scrap has grown recently due to its high melting temperature and powder reactivity challenges for metal additive manufacturing (MAM). This study successfully built Zircaloy-4 parts using laser powder bed fusion MAM for the first time while suppressing powder reactivity. The mechanical properties of the printed Zircaloy-4 were evaluated at room temperature and reactor operating temperature, showing remarkable yield strengths at both temperatures. This study serves as a starting point for safe Zircaloy-4 additive manufacturing.
METALS AND MATERIALS INTERNATIONAL
(2023)
Article
Medicine, General & Internal
Dougho Park, Mun-Chul Kim, Daeyoung Hong, Yong-Suk Jeong, Hyoung Seop Kim, Jong Hun Kim
Summary: This study aimed to explore the long-term risk of recurrence and mortality in patients with acute ischemic stroke (AIS), acute myocardial infarction (AMI), or acute hemorrhagic stroke (AHS). The results showed that the risk of developing secondary AIS was significantly lower in the AMI and AHS groups compared to the AIS group. The risk of developing secondary AMI was also significantly lower in the AMI and AHS groups compared to the AIS group. Initial AHS was a decisive risk factor for developing secondary AHS. Furthermore, AMI and AHS were more closely related to long-term mortality than AIS.
JOURNAL OF CLINICAL MEDICINE
(2023)
Article
Materials Science, Multidisciplinary
Jihye Kwon, Olivier Bouaziz, Hyoung Seop Kim, Yuri Estrin
Summary: Crumpled metallic thin foils have great potential as weight-saving and energy-absorption materials, but further research is needed on their crumpling process and mechanical behavior, considering their complex internal structures. This study presents two possible computational strategies for simulating crumpled materials under closed-die compression. The analysis shows that the continuum-based approach is more suitable for representing the macroscopic mechanical behavior of crumpled materials within a certain range of relative densities. The porous continuum approach also offers the benefits of low computational cost and high efficacy. However, the direct method is preferable when accurately reproducing internal structural pattern changes is necessary, such as for predicting mechanical response under complex loading conditions.
ADVANCED ENGINEERING MATERIALS
(2023)
Article
Materials Science, Multidisciplinary
Jae Heung Lee, Hyeonseok Kwon, Praveen Sathiyamoorthi, Sujung Son, Hyoung Seop Kim
Summary: To enhance the tensile properties of Al0.3CoCrNi (at%) medium-entropy alloys at room temperature, high-pressure torsion (HPT) and appropriate annealing are employed. Severe deformation induced by HPT leads to the formation of fine-grained structure and nanometer-scale B2 and sigma precipitates, which result in a synergistic combination of high yield strength (>1.0 GPa) and large uniform elongation (>25%) by utilizing grain refinement and precipitation strengthening simultaneously.
ADVANCED ENGINEERING MATERIALS
(2023)
Article
Materials Science, Multidisciplinary
D. H. Chung, J. Lee, Q. F. He, Y. K. Kim, K. R. Lim, H. S. Kim, Y. Yang, Y. S. Na
Summary: The study investigates the toughening/strengthening mechanisms of heterostructured eutectic high-entropy alloys (EHEAs) and discovers that fully eutectic HEAs show superior performance in both yield stress and fracture toughness due to the high hetero-deformation-induced (HDI) strengthening/toughening associated with a high misorientation angle at the grain/phase boundaries.
JOURNAL OF MATERIALS SCIENCE & TECHNOLOGY
(2023)
Article
Materials Science, Multidisciplinary
Jungwan Lee, Olivier Bouaziz, Yuri Estrin, Hyoung Seop Kim
Summary: A physically based model is used to study the strain-hardening behavior of a metastable medium-entropy alloy Fe-61(CoNi)(29)Cr-10, which is controlled by a deformation-induced martensitic transformation. The model accurately predicts the stress-strain curves and shows the variation of phase composition with strain due to the phase transformation.
METALLURGICAL AND MATERIALS TRANSACTIONS A-PHYSICAL METALLURGY AND MATERIALS SCIENCE
(2023)
Article
Materials Science, Multidisciplinary
M. Farvizi, M. Bahamirian, A. Faraji, H. S. Kim
Summary: By adding YSZ microparticles with a similar thermal expansion coefficient to NiTi alloy as reinforcement, the wear resistance and mismatch stress in the matrix can be improved. The use of YSZ instead of monoclinic zirconia as reinforcement reduces mismatch stress and increases the fraction of austenitic phase, leading to better shape memory and wear resistance properties.
Article
Mechanics
Shin-Yeong Lee, Jin-Hwan Kim, Frederic Barlat, Hyoung Seop Kim
Summary: Fracture of martensitic steel in hole expansion and stretch bending was predicted using finite element analysis with shell elements. Mechanical experiments were conducted to characterize mechanical properties related to fracture and plasticity. Two fracture models (HC and GTN-shear) were selected and calibrated with load-displacement curves of fracture tests. The GTN-shear model predicted fracture well in the conical hole expansion test, while the HC model predicted fracture well in the stretch bending test. Differences between the two models may result from a non-linear loading path and description of localized necking, as analyzed through triaxiality analysis.
ENGINEERING FRACTURE MECHANICS
(2023)
Article
Materials Science, Multidisciplinary
Gang Hee Gu, Hyeonseok Kwon, Yongju Kim, Farahnaz Haftlang, Yoon-Uk Heo, Hyoung Seop Kim
Summary: Carbon-added equi-atomic CoCrFeMnNi high-entropy alloys exhibit bake hardening effect and show better performance compared to conventional bake-hardenable materials, providing possibilities for the industrialization of various interstitial HEA systems as well as carbon-added CoCrFeMnNi HEAs.
MATERIALS & DESIGN
(2023)
Article
Materials Science, Multidisciplinary
Hyojin Park, Jungwan Lee, Rae Eon Kim, Sujung Son, Soung Yeoul Ahn, Hyoung Seop Kim
Summary: This study investigates the impact of warm rolling on the microstructure and tensile properties of Fe60(CoNi)30Cr10 Fe-based medium entropy alloy (Fe-MEA). The results show that the warm-rolled sample has coarse elongated grains, deformation-induced martensite, and a high density of dislocations. Tensile tests demonstrate that the warm-rolled Fe-MEA exhibits enhanced strength and a similar level of elongation compared to the annealed sample. The study suggests that warm rolling shows potential as a processing technique to improve the mechanical properties of Fe-MEA for broader applications.
METALS AND MATERIALS INTERNATIONAL
(2023)
Article
Materials Science, Multidisciplinary
Hamed Shahmir, Shabnam Kazemi, Mohammad Sajad Mehranpour, Hyoung Seop Kim
Summary: The effect of Al addition on the microstructure of a CoCrFeNiMn high-entropy alloy was investigated using a novel approach. The formation and stability of phases were examined through heat treatment and CALPHAD predictions. This method is significant for alloy design as it allows for a quick study of phase evolution in complex multi-component alloys.
MATERIALS TODAY COMMUNICATIONS
(2023)
Article
Nanoscience & Nanotechnology
Jeong Min Park, Hyeonseok Kwon, Jungho Choe, Kyung Tae Kim, Ji-Hun Yu, Yoon-Uk Heo, Hyoung Seop Kim
Summary: This study presents a guideline for alloy design in additive manufacturing, aiming to produce high-quality products with excellent mechanical performance by utilizing the unique segregation engineering of LPBF-driven microstructures. Mo-doping enhances the strength and ductility of ferrous medium-entropy alloys.
SCRIPTA MATERIALIA
(2023)
Article
Chemistry, Physical
Farahnaz Haftlang, Alireza Zargaran, Jongun Moon, Soung Yeoul Ahn, Jae Bok Seol, Hyoung Seop Kim
Summary: In this study, the concept of maraging characteristics was manipulated by a multistep thermomechanical approach. A novel metastable maraging Fe68Ni10Mn10Co10Ti1.5Si0.5 (at%) medium entropy alloy was designed and microstructurally engineered with heterogeneities to achieve ultra-high yield strength and total elongation. This provides a prospective direction for the development of high-performance materials for structural applications.
JOURNAL OF ALLOYS AND COMPOUNDS
(2023)