Article
Engineering, Multidisciplinary
Suhan Kim, Hyunseong Shin
Summary: A deep learning framework is proposed for multiscale finite element analysis, in which a data-driven computational mechanics approach is adopted to overcome the inefficiency of the traditional FE2 method. Macroscopic strain and stress data are directly assigned to material points without using constitutive model. The proposed approach uses a deep neural network to enable adaptive sampling points and significantly improves the computational efficiency of offline computing.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Computer Science, Interdisciplinary Applications
Suhan Kim, Hyunseong Shin
Summary: In this paper, a data-driven multiscale finite-element method (data-driven FE2) is proposed to describe nonlinear heterogeneous materials using a deep neural network (DNN) and proper orthogonal decomposition (POD). The method solves the macroscopic problem by constructing a material genome database, which contains pre-calculated microscopic problems of the representative volume element (RVE), and assigns data to all integration points that satisfy microscopic equilibrium. This approach overcomes the computational time limitations of the classical FE2 method.
ENGINEERING WITH COMPUTERS
(2023)
Article
Materials Science, Multidisciplinary
Yang Sun, Yifeng Hu, Mabao Liu
Summary: The study investigates the influence of hard/soft interface effects on the macroscopic elastoplastic behavior of graphene reinforced nanocomposites. Theoretical framework, comparative analyses against experimental data, and detailed parametric analyses are used to expose the effects of major micromechanical variables.
MATERIALS & DESIGN
(2021)
Article
Engineering, Multidisciplinary
Stefanos Pyrialakos, Ioannis Kalogeris, Gerasimos Sotiropoulos, Vissarion Papadopoulos
Summary: This study presents a Bayesian framework for determining the mechanical properties of carbon-based nanocomposites by updating prior beliefs using measurements on large-scale structures. A surrogate modeling technique utilizing artificial neural networks is developed to predict the nonlinear stress-strain relationship of representative volume elements. This methodology is validated through numerical examples and can be applied to other physically analogous phenomena related to composite materials modeling.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Mathematics, Interdisciplinary Applications
Karl A. Kalina, Lennart Linden, Joerg Brummund, Markus Kaestner
Summary: In this paper, a new data-driven multiscale framework called FEANN is introduced, which utilizes physics-constrained artificial neural networks (ANNs) as macroscopic surrogate models and includes an autonomous data mining process. This approach enables efficient simulation of materials with complex microstructures exhibiting anisotropic and nonlinear behavior on the macroscale. The proposed framework significantly reduces the number of time-consuming microscale simulations by autonomously gathering and extending the necessary data set.
COMPUTATIONAL MECHANICS
(2023)
Article
Materials Science, Multidisciplinary
Fadi Aldakheel, Celal Soyarslan, Hari Subramani Palanisamy, Elsayed Saber Elsayed
Summary: Computational material modeling using CNN provides a solution for efficient and accurate modeling in heterogeneous materials, reducing the development costs and speeding up the design process.
MECHANICS OF MATERIALS
(2023)
Article
Mechanics
Jun-Hyok Ri, Un-Il Ri, Hyon-Sik Hong
Summary: A cluster-based nonuniform transformation field analysis method for elastic-viscoplastic composite is proposed, which reduces the number of integration points needed for the local constitutive relation and improves the prediction accuracy and computational efficiency of the effective behavior of composite materials.
COMPOSITE STRUCTURES
(2021)
Article
Mathematics
Harshit Mohan, Gopal Agrawal, Vibhu Jately, Abhishek Sharma, Brian Azzopardi
Summary: To reduce pollution and energy consumption, electric vehicles (EVs) are gaining more attention in the automotive industry. High efficiency, compactness, lightweight, low cost, and easy recyclability are desired in the electric motors used in EVs. Various motor control strategies and sensorless speed control techniques are employed to achieve better dynamic performance and increased reliability.
Article
Mechanics
Wei Huang, Rui Xu, Jie Yang, Qun Huang, Heng Hu
Summary: This paper introduces a multiscale data-driven framework for FRP composites, which collects material databases and simulates structural behavior through a data-driven approach, leading to significant cost savings and promising applications.
COMPOSITE STRUCTURES
(2021)
Article
Chemistry, Multidisciplinary
Yunfei Gao, Guogui Huang, Yinxi Li, Junyuan Zhang, Zeng Yang, Meng Wang
Summary: This study proposes a framework that utilizes Bayesian Optimized Back Propagation Artificial Neural Network (BP-ANN) to predict the homogenized mechanical parameters of soils, addressing the issue of high computational costs in existing methods.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Mechanical
Liang Li, Qian Shao, Yichen Yang, Zengtao Kuang, Wei Yan, Jie Yang, Ahmed Makradi, Heng Hu
Summary: A new method combining computational homogenization and artificial neural network (ANN) is proposed to efficiently construct elastoplastic composites database for data-driven computational mechanics (DDCM). Numerical calculations are performed to collect high-fidelity data on the representative volume element (RVE) of elastoplastic composites, which are then enhanced using ANN. The proposed method shows good accuracy and efficiency in reducing computational cost for database construction in DDCM of composites.
INTERNATIONAL JOURNAL OF MECHANICAL SCIENCES
(2023)
Article
Environmental Studies
Yuche Chen, Yunteng Zhang, Ruixiao Sun
Summary: This study developed machine learning models to estimate electric bus energy consumption based on data from Chattanooga in 2019 and 2020, achieving predicted mean absolute percentage error rates of 3% for LSTM and 5% for ANN. The research proposed a data-partitioning algorithm and conducted K-fold cross-validation to select optimal model structures and input variables, demonstrating the predictability of the models when compared with existing literature.
TRANSPORTATION RESEARCH PART D-TRANSPORT AND ENVIRONMENT
(2021)
Article
Chemistry, Physical
Ke Chen, Xuke Tang, Binbin Jia, Cezhou Chao, Yan Wei, Junyu Hou, Leiting Dong, Xuliang Deng, Ting-Hui Xiao, Keisuke Goda, Lin Guo
Summary: Inspired by the heterophase structure of nacre, a centimetre-sized bulk material consisting of graphene oxide (GO) and amorphous/crystalline MnO2 nanosheets adhered together with polymer-based crosslinkers was prepared, exhibiting high flexural strength and fracture toughness. Experimental and numerical analyses revealed that the ordered heterophase structure and synergistic crosslinking interactions across multiscale interfaces contribute to the superior mechanical properties of the material.
Article
Energy & Fuels
Ahmed G. Saad, Ahmed Emad-Eldeen, Wael Z. Tawfik, Ahmed G. El-Deen
Summary: Graphene-based nanocomposites have strong potential as high-capacity supercapacitor electrodes in energy storage systems. Developing an accurate and effective prediction technique using machine learning models is crucial for reducing the time needed to design and test electrode materials. Experimental data from over two hundred research papers was examined to predict the specific capacitance of graphene-based electrode structures, with the artificial neural network (ANN) model demonstrating the highest accuracy in predictions.
JOURNAL OF ENERGY STORAGE
(2022)
Article
Engineering, Multidisciplinary
Jan Niklas Fuhg, Christoph Boehm, Nikolaos Bouklas, Amelie Fau, Peter Wriggers, Michele Marino
Summary: Computational multiscale methods have been used in engineering problems to analyze and derive constitutive responses, but their application in a nonlinear framework may be limited by high computational costs, numerical difficulties, and inaccuracies. A hybrid methodology combining classical constitutive laws, data-driven correction, and computational multiscale approaches is presented in this paper. This approach improves the versatility and accuracy compared to classical model-driven or purely data-driven techniques.
INTERNATIONAL JOURNAL OF ENGINEERING SCIENCE
(2021)
Article
Materials Science, Composites
Haitang Yang, Yanping Yang, Yu Liu, Delong He, Jinbo Bai
Summary: By studying original CNTs and carboxylic acid-modified CNTs in epoxy composites, this research found that the modified CNTs exhibited better mechanical properties and dispersion state, while confirming the importance of interface analysis.
COMPOSITE INTERFACES
(2021)
Article
Engineering, Multidisciplinary
Dimitrios Tsapetis, Gerasimos Sotiropoulos, George Stavroulakis, Vissarion Papadopoulos, Manolis Papadrakakis
Summary: This study extends isogeometric thin shell formulations to incorporate constitutive laws generated by stochastic multiscale analyses, demonstrating the impact of material and inclusion variability on structural response through the use of stochastic processes. The consideration of spatial material variability in shell structures makes this formulation an ideal candidate for simulating composite material shell structures.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Mechanics
Serafeim Bakalakos, Ioannis Kalogeris, Vissarion Papadopoulos
Summary: This paper introduces an XFEM formulation for heat transfer analysis of multi-phase materials, which simplifies the analysis process by capturing discontinuities in temperature field with appropriate discontinuous functions. Through validation and simulation, the method's effectiveness in heat conduction problems is demonstrated.
COMPOSITE STRUCTURES
(2021)
Article
Computer Science, Interdisciplinary Applications
Michael D. Shields, Dimitris G. Giovanis, V. S. Sundar
Summary: The paper proposes the use of an affine invariant ensemble MCMC sampler for conditional sampling to address extreme cases where subset simulation breaks down. The algorithm automatically varies step size and is particularly effective for estimating failure probabilities in strongly non-Gaussian and lower effective dimension scenarios.
COMPUTERS & STRUCTURES
(2021)
Article
Engineering, Multidisciplinary
Ioannis Kalogeris, Vissarion Papadopoulos
Summary: This work introduces a surrogate modeling strategy based on diffusion maps manifold learning and artificial neural networks to efficiently predict responses of complex systems. By utilizing dimensionality reduction with diffusion maps, the method overcomes the curse of dimensionality and improves prediction accuracy and training efficiency.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Materials Science, Composites
Xiaoxin Lu, Fabrice Detrez, Julien Yvonnet, Jinbo Bai
Summary: The study developed the ALIAS method to identify the local stiffness tensor of graphene polymer nanocomposites, revealing the modeling of graphene at continuum scale and its effects on the overall elastic properties. Results showed a significant softening effect due to interfaces, while wrinkles increased the stiffness of nanocomposites.
COMPOSITES SCIENCE AND TECHNOLOGY
(2021)
Article
Construction & Building Technology
Zacharias Fasoulakis, Dimitrios Vamvatsikos, Vissarion Papadopoulos
Summary: This paper focuses on probabilistic estimation of the buckling capacity of single-bolted members with stochastic geometric imperfections from plain or lipped angle sections. An experimental-stochastic mechanics approach is adopted, and it is found that imperfections have a lower influence compared to material properties and lateral loading for typical lattice tower angle members.
JOURNAL OF STRUCTURAL ENGINEERING
(2021)
Article
Chemistry, Physical
Xiaoxin Lu, Julien Yvonnet, Leonidas Papadopoulos, Ioannis Kalogeris, Vissarion Papadopoulos
Summary: A stochastic data-driven multilevel finite-element (FE2) method is introduced for random nonlinear multiscale calculations, which uses a hybrid NN-I scheme to construct a surrogate model of the macroscopic nonlinear constitutive law and significantly reduces computational time.
Article
Engineering, Multidisciplinary
Stefanos Pyrialakos, Ioannis Kalogeris, Gerasimos Sotiropoulos, Vissarion Papadopoulos
Summary: This study presents a Bayesian framework for determining the mechanical properties of carbon-based nanocomposites by updating prior beliefs using measurements on large-scale structures. A surrogate modeling technique utilizing artificial neural networks is developed to predict the nonlinear stress-strain relationship of representative volume elements. This methodology is validated through numerical examples and can be applied to other physically analogous phenomena related to composite materials modeling.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Materials Science, Multidisciplinary
Serafeim Bakalakos, Ioannis Kalogeris, Vissarion Papadopoulos, Manolis Papadrakakis, Panagiotis Maroulas, Dimitrios A. Dragatogiannis, Costas A. Charitidis
Summary: This paper investigates the thermal properties of carbon nanotube reinforced polyethylene as a highly conductive material. An integrated approach combining numerical and experimental procedures is proposed. The interfacial thermal conductance parameter value is inferred by calibrating the numerically predicted effective conductivity to the series of experimental measurements, demonstrating the potential of the composite as a highly conductive material.
MODELLING AND SIMULATION IN MATERIALS SCIENCE AND ENGINEERING
(2022)
Article
Engineering, Multidisciplinary
Gerasimos Sotiropoulos, Vissarion Papadopoulos
Summary: In this work, a formulation and modeling scheme for the non-linear multi-scale analysis of thin shells is presented. This method is capable of dealing with large deformations and heterogeneous micro-structures composed of non-linear materials and cohesive interfaces. By utilizing an attached coordinate system, the projection of strain measures allows for the elimination of large rotations, simplifying the boundary value problem at the micro-structural level. The resulting methodology has been tested against popular benchmarks and successfully integrated in existing FE2 codes, providing countless simulation possibilities and wide applicability in engineering fields.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Automation & Control Systems
Stefanos Nikolopoulos, Ioannis Kalogeris, Vissarion Papadopoulos
Summary: This paper presents a novel non-intrusive surrogate modeling scheme based on deep learning for predictive modeling of complex systems described by parametrized time-dependent partial differential equations. The proposed method utilizes a convolutional autoencoder and a feed forward neural network to establish a mapping from the problem's parametric space to its solution space. The surrogate model is capable of predicting the entire time history response simultaneously with remarkable computational gains and very high accuracy.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Engineering, Civil
Stefanos Nikolopoulos, Ioannis Kalogeris, Vissarion Papadopoulos
Summary: This paper presents a non-intrusive surrogate modeling scheme for transient response analysis of nonlinear structures involving random parameters using a two-level neural network architecture. The proposed scheme combines feed-forward neural networks with convolutional autoencoders to deliver an accurate and inexpensive emulator of the structural system under investigation.
ENGINEERING STRUCTURES
(2022)
Article
Engineering, Multidisciplinary
Ketson R. M. dos Santos, Dimitris G. Giovanis, Katiana Kontolati, Dimitrios Loukrezis, Michael D. Shields
Summary: A novel surrogate model based on Grassmannian diffusion maps and geometric harmonics is developed for predicting the response of complex physical phenomena. The model utilizes low-dimensional representation and mapping techniques to reconstruct the full solution. The performance of the model is verified through various examples.
INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING
(2022)
Article
Engineering, Mechanical
Dimitrios G. Giovanis, Michael D. Shields
Summary: The objective of this study is to quantify the uncertainty in probability of failure estimates resulting from incomplete knowledge of the probability distributions for the input random variables. The study proposes a framework that combines Subset Simulation (SuS) with Bayesian/information theoretic multi-model inference, and through methods such as multi-model inference and importance sampling, empirical probability distributions of failure probabilities that provide direct estimates of the uncertainty in failure probability estimates are obtained.
PROBABILISTIC ENGINEERING MECHANICS
(2022)
