Article
Computer Science, Artificial Intelligence
Ryno Laubscher, Pieter Rousseau
Summary: A novel integrated deep learning approach for data-driven surrogate modelling of combustion CFD simulations is proposed, combining VAEs with DNNs to predict detailed cell-by-cell two-dimensional distributions of temperature, velocity, and species mass fractions from high level inputs. Regularization is found to be necessary during all training phases, and sufficiently accurate results were achieved with mean average errors below 0.3% for species mass fractions. Validation mean average percentage errors for temperature and velocity fields are 1.7% and 7.1% respectively, indicating the ability to predict detailed two-dimensional contours of CFD solution data with adequate generalizability and accuracy.
APPLIED SOFT COMPUTING
(2021)
Article
Chemistry, Multidisciplinary
Enea Prifti, James P. Buban, Arashdeep Singh Thind, Robert F. Klie
Summary: Identifying point defects and other structural anomalies using scanning transmission electron microscopy (STEM) is crucial for understanding the properties of materials. Recent advancements in computer vision and machine learning have enabled automated analysis of atomic structures and defects.
Article
Computer Science, Artificial Intelligence
Paraskevi Nousi, Styliani-Christina Fragkouli, Nikolaos Passalis, Panagiotis Iosif, Theocharis Apostolatos, George Pappas, Nikolaos Stergioulas, Anastasios Tefas
Summary: This study investigates the underlying structures in the interpolation coefficients of gravitational waves and proposes a spiral module for improved speed-accuracy trade-off. Experimental results demonstrate the module's excellent performance in various neural network architectures and its ability to evaluate millions of input parameters in a short time.
Article
Engineering, Aerospace
Oscar Ulises Espinosa Barcenas, Jose Gabriel Quijada Pioquinto, Ekaterina Kurkina, Oleg Lukyanov
Summary: The aircraft conceptual design step involves a significant number of aerodynamic configuration evaluations. Surrogate modeling is used to simplify complex models and reduce computational time. A multilayer perceptron (MLP) is implemented to predict aerodynamic coefficients with high accuracy.
Article
Computer Science, Artificial Intelligence
Guillaume Salha, Romain Hennequin, Jean-Baptiste Remy, Manuel Moussallam, Michalis Vazirgiannis
Summary: This paper introduces FastGAE, a general framework to scale graph AE and VAE to large graphs, which significantly speeds up training and improves performance through an effective stochastic subgraph decoding scheme. FastGAE demonstrates effectiveness on various real-world graphs, outperforming existing approaches by a wide margin.
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
Energy & Fuels
Ramesh Kanthasamy, Eydhah Almatrafi, Imtiaz Ali, Hani Hussain Sait, Mohammed Zwawi, Faisal Abnisa, Leo Choe Peng, Bamidele Victor Ayodele
Summary: This study uses Bayesian optimized multilayer perceptron neural network to model the prediction of biochar and syngas from biomass-derived waste pyrolysis. The performance of the neural networks is influenced by the number of connecting layers and the size of the hidden neurons. The best-performing neural network architecture for predicting biochar yield is 3-2-10-10-1 with R2 of 0.984 and RMSE of 0.34, while for predicting syngas yield it is 3-7-10-3-1 with R2 of 0.999.
Article
Computer Science, Artificial Intelligence
Tal Mezheritsky, Liset Vazquez Romaguera, William Le, Samuel Kadoury
Summary: A deep learning-based motion modelling framework for ultrasound IGRT is presented, showing potential for target tracking with a mean tracking error of 3.5 +/- 2.4 mm, surpassing comparable methodologies in both metric categories.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Computer Science, Artificial Intelligence
Qien Yu, Muthu Subash Kavitha, Takio Kurita
Summary: This study proposes a method for anomaly detection using a mixture of CVAEs models, which learns the multi-manifold relationships of data through an ensemble of experts. The model shows superior performance on multiple datasets compared to existing methods for image anomaly detection tasks.
APPLIED INTELLIGENCE
(2021)
Article
Computer Science, Information Systems
Yongliang Zhang, Xiaoli Wang, Yanxing Wang, Ningchaoran Yan, Linping Feng, Lu Zhang
Summary: This paper proposes an adaptive online updating 1D convolutional autoencoders (AOU-1D-CAE) surrogate model for filter optimization problems, which involve time-consuming simulations and many variables in the design. The experimental results demonstrate that the proposed surrogate model improves data collection efficiency and prediction performance. Furthermore, the surrogate model assists particle swarm optimization (PSO) in finding the global optimal solution, leading to enhanced optimization efficiency.
Article
Biodiversity Conservation
Guoli Zhang, Ming Wang, Kai Liu
Summary: This paper compares and analyzes the application of two feedforward neural network models (CNNs and MLPs) in global wildfire susceptibility prediction, and explores the interpretability of the CNNs model. By constructing response variables and monthly wildfire predictors, four MLPs and CNNs architectures were built, and five statistical measures were used to evaluate the prediction performance of the models. The contextual-based CNN-2D model was found to have the highest accuracy, while the MLPs model was more suitable for pixel-based classification, and the performance ranking of the four models was CNN-2D > MLP-1D > MLP-2D > CNN-1D.
ECOLOGICAL INDICATORS
(2021)
Article
Computer Science, Artificial Intelligence
Xiangru Chen, Yanan Sun, Mengjie Zhang, Dezhong Peng
Summary: In this article, a novel method for automatically designing optimal architectures of VAEs for image classification, called EvoVAE, based on a genetic algorithm, is proposed. Experiment results demonstrate the superiority of the EvoVAE algorithm over peer competitors on three benchmark datasets.
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION
(2021)
Article
Computer Science, Artificial Intelligence
Hamidreza Eivazi, Soledad Le Clainche, Sergio Hoyas, Ricardo Vinuesa
Summary: Modal-decomposition techniques are important for capturing dominant flow features. This research proposes a deep probabilistic-neural-network architecture for learning non-linear modes from turbulent-flow data. The method extracts non-linear modes and encourages the learning of independent latent variables. By constraining the shape of the latent space, a set of parsimonious modes can be extracted for high-quality flow reconstruction.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Forestry
Mehmet Ismail Guersoy, Osman Orhan, Senem Tekin
Summary: Considering the increase in wildfire disasters due to climate change, this study used Convolutional Neural Network (CNN) and Multilayer Perceptron (MLP) methods to create wildfire susceptibility models in six provinces in the Mediterranean region of Turkey. Seventeen environmental variables were analyzed, and the Synthetic Minority Oversampling Technique (SMOTE) was used to balance the limited fire inventory data. The CNN model showed superior performance in wildfire susceptibility assessment, indicating its better prediction capability compared to the MLP model. The produced susceptibility maps can assist decision-makers and government organizations in preventing future natural disasters in the Mediterranean region.
FOREST ECOLOGY AND MANAGEMENT
(2023)
Article
Engineering, Multidisciplinary
Jian Tang, Siddhant Kumar, Laura De Lorenzis, Ehsan Hosseini
Summary: We propose Neural Cellular Automata (NCA) for simulating microstructure development in the solidification process of metals. NCA, based on convolutional neural networks, can learn essential features of solidification and are much faster than conventional Cellular Automata (CA). Notably, NCA can make reliable predictions beyond their training range, indicating their understanding of the physics of solidification. While CA data is used for training in this study, NCA can be trained on any microstructural simulation data, such as phase-field models.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Thermodynamics
Mahsa Taghavi, Swapnil Sharma, Vemuri Balakotaiah
Summary: This study investigates the natural convection effects in the insulation layers of spherical storage tanks and their impact on the tanks' performance. The permeability and Rayleigh number of the insulation material are considered as key factors. The results show that as the Rayleigh number increases, new convective cells emerge and cause the cold boundary to approach the external hot boundary. In the case of large temperature differences, multiple solutions may coexist.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Jinyang Xu, Fangjun Hong, Chaoyang Zhang
Summary: This study introduces a self-induced jet impingement device for enhancing pool boiling performance in high power electronic cooling. Through visualization and parametric investigations, the effects of this device on pool boiling performance are studied, revealing the promotion of additional liquid supply and vapor exhausting. The flow rate of the liquid jet is found to positively impact boiling performance.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Wenchao Ke, Yuan Liu, Fissha Biruke Teshome, Zhi Zeng
Summary: Underwater wet laser welding (UWLW) is a promising and labor-saving repair technique. A thermal multi-phase flow model was developed to study the heat transfer, fluid dynamics, and phase transitions during UWLW. The results show that UWLW creates a water keyhole, making the welding environment similar to in air laser welding.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Xingrong Lian, Lin Tian, Zengyao Li, Xinpeng Zhao
Summary: This study investigates the heat transfer mechanisms in natural fiber-derived porous structures and finds that thermal radiation has a significant impact on the thermal conductivity in low-density regions, while natural convection rarely occurs. Insulation materials derived from micron-sized natural fibers can achieve minimum thermal conductivity at specific densities. Strategies to lower the thermal conductivity include increasing porosity and incorporating nanoscale pores using nanosize fibers.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Yasir A. Malik, Kilian Koebschall, Stephan Bansmer, Cameron Tropea, Jeanette Hussong, Philippe Villedieu
Summary: Ice crystal icing is a significant hazard in aviation, and accurate modeling of sticking efficiency is essential. In this study, icing wind tunnel experiments were conducted to quantify the volumetric liquid water fraction, sticking efficiency, and maximum thickness of ice layers. Two measurement techniques, calorimetry and capacitive measurements, were used to measure the liquid water content and distribution in the ice layers. The experiments showed that increasing wet bulb temperatures and substrate heat flux significantly increased sticking efficiency and maximum ice layer thickness.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Jinqi Hu, Tongtong Geng, Kun Wang, Yuanhong Fan, Chunhua Min, Hsien Chin Su
Summary: This study experimentally examined the heat dissipation of vibrating fans and demonstrated its inherent mechanism through numerical simulation. The results showed that the flow fields induced by the vibrating blades exhibited pulsating features and formed large-scale and small-scale vortical structures, significantly improving heat dissipation. The study also identified the impacts of different blade structures and developed a trapezoidal-folding blade, which effectively reduced the maximum temperature of the heat source and alleviated high-temperature failure crisis.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Dan-Dan Su, Xiao-Bin Li, Hong-Na Zhang, Feng-Chen Li
Summary: The boiling heat transfer of low-boiling-point working fluid is a common heat dissipation technology in electronic equipment cooling. This study analyzed the interfacial boiling behavior of R134a under different conditions and found that factors such as the initial thickness of the liquid film, solid-liquid interaction force, and initial temperature significantly affect the boiling mode and thermal resistance.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Jinyi Wu, Dongke Sun, Wei Chen, Zhenhua Chai
Summary: A unified lattice Boltzmann-phase field scheme is proposed to simulate dendrite growth of binary alloys in the presence of melt convection. The effects of various factors on the growth are investigated numerically, and the model is validated through comparisons and examinations.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Shaokun Ge, Ya Ni, Fubao Zhou, Wangzhaonan Shen, Jia Li, Fengqi Guo, Bobo Shi
Summary: This study investigated the temperature distribution of main cables in a suspension bridge during fire scenarios and proposed a prediction model for the maximum temperature of cables in different lane fires. The results showed that vehicle fires in the emergency lane posed a greater thermal threat to the cables.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Shuang-Ying Wu, Shi-Yao Zhou, Lan Xiao, Jia Luo
Summary: This paper investigates the two-phase flow and heat transfer characteristics of low-velocity jet impacting on a cylindrical surface. The study reveals that the heat transfer regimes are non-phase transition and nucleate boiling with the increase of heat transfer rate. The effects of jet impact height and outlet velocity on local surface temperatures are pronounced at the non-phase transition stage. The growth rates of heat transfer rate and liquid loss rate increase significantly from the non-phase transition to nucleate boiling stage.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Emad Hasani Malekshah, Wlodzimierz Wlodzimierz, Miros law Majkut
Summary: Cavitation has significant practical importance and can be controlled by air injection. This study investigates the natural to ventilated cavitation process around a hydrofoil through numerical and experimental methods. The results show that the location and rate of air injection have a meaningful impact on the characteristics of cavitation.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Feriel Yahiat, Pascale Bouvier, Antoine Beauvillier, Serge Russeil, Christophe Andre, Daniel Bougeard
Summary: This study explores the enhancement of mixing performance in laminar flow equipment by investigating the generation of chaotic advection using wall deformations in annular geometries. The findings demonstrate that the combined geometry can achieve perfect mixing at various Reynolds numbers.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Hui He, Ning Lyu, Caihua Liang, Feng Wang, Xiaosong Zhang
Summary: This study investigates the condensation, frosting, and defrosting processes on superhydrophobic surfaces with millimeter-scale structures. The results reveal that the structures can influence the growth and removal of frost crystals, with the bottom grooves creating a frost-free zone and conical edges promoting higher frost crystal heights. Two effective methods for defrosting are observed: hand-lifting the groove and airfoil retraction contraction on protruding structures. This research provides valuable insights into frost formation and defrosting on millimeter-structured superhydrophobic surfaces, with potential applications in anti-frost engineering.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Thiwanka Arepolage, Christophe Verdy, Thibaut Sylvestre, Aymeric Leray, Sebastien Euphrasie
Summary: This study developed two thermal concentrators, one with a 2D design of uniform thickness and another with a 3D design, using the coordinate transformation technique and metamaterials. By structuring the thermal conductor, the desired local density-heat capacity product and anisotropic thermal conductivities were achieved. The homogenized thermal conductivities were obtained from finite element simulations and cylindrical symmetry consideration. A 3D concentrator was fabricated using 3D metal printing and characterized using a thermal camera. Compared to devices that solely consider anisotropic conductivities, the time evolution characteristics of the metadevice designed with coordinate transformation were closer to those of an ideal concentrator.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)
Article
Thermodynamics
Liangyuan Cheng, Qingyang Wang, Jinliang Xu
Summary: In this study, we investigated the supercritical heat transfer of CO2 in a horizontal tube with a diameter of 10.0 mm, covering a wide range of pressures, mass fluxes, and heat fluxes. The study revealed a non-monotonic increase in wall temperatures along the flow direction and observed both positive and negative wall temperature differences between the bottom and top tube. The findings were explained by the thermal conduction in the solid wall interacting with the stratified-wavy flow in the tube.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2024)