Hierarchical Multiscale Research Articles

Crack Detection via Hierarchical Multiscale Feature Learning and Densely Connected Conditional Random Field

Crack Detection via Hierarchical Multiscale Feature Learning and Densely Connected Conditional Random Field

HSG-MGAF Net: Heterogeneous subgraph-guided multiscale graph attention fusion network for interpretable prediction of whole-slide image

Background and objectivePathological whole slide image (WSI) prediction and region of interest (ROI) localization are important issues in computer-aided diagnosis and postoperative analysis in clinical applications. Existing computer-aided methods for predicting WSI are mainly based on multiple instance learning (MIL) and its variants. However, most of the methods are based on instance independence and identical distribution assumption and performed at a single scale, which not fully exploit the hierarchical multiscale heterogeneous information contained in WSI. MethodsHeterogeneous Subgraph-Guided Multiscale Graph Attention Fusion Network (HSG-MGAF Net) is proposed to build the topology of critical image patches at two scales for adaptive WSI prediction and lesion localization. The HSG-MGAF Net simulates the hierarchical heterogeneous information of WSI through graph and hypergraph at two scales, respectively. This framework not only fully exploits the low-order and potential high-order correlations of image patches at each scale, but also leverages the heterogeneous information of the two scales for adaptive WSI prediction. ResultsWe validate the superiority of the proposed method on the CAMELYON16 and the TCGA- NSCLC, and the results show that HSG-MGAF Net outperforms the state-of-the-art method on both datasets. The average ACC, AUC and F1 score of HSG-MGAF Net can reach 92.7 %/0.951/0.892 and 92.2 %/0.957/0.919, respectively. The obtained heatmaps can also localize the positive regions more accurately, which have great consistency with the pixel-level labels. ConclusionsThe results demonstrate that HSG-MGAF Net outperforms existing weakly supervised learning methods by introducing critical heterogeneous information between the two scales. This approach paves the way for further research on light weighted heterogeneous graph-based WSI prediction and ROI localization.

Building Physics-based Hierarchical Multiscale Models Using Machine Learning

Building Physics-based Hierarchical Multiscale Models Using Machine Learning

A multi‐stage model for bird's eye view prediction based on stereo‐matching model and RGB‐D semantic segmentation

AbstractBird's‐Eye‐View (BEV) map is a powerful and detailed scene representation for intelligent vehicles that provides both the location and semantic information about nearby objects from a top‐down perspective. BEV map generation is a complex multi‐stage task, and the existing methods typically perform poorly for distant scenes. Thus, the authors introduce a novel multi‐stage model that infers to obtain more accurate BEV map. First, the authors propose the Adaptive Aggregation with Stereo Mixture Density (AA‐SMD) model, which is an improved stereo matching model that eliminates bleeding artefacts and provides more accurate depth estimation. Next, the authors employ the RGB‐Depth (RGB‐D) semantic segmentation model to improve the semantic segmentation performance and connectivity of their model. The depth map and semantic segmentation maps are then combined to create an incomplete BEV map. Finally, the authors propose a Multi Strip Pooling Unet (MSP‐Unet) model with a hierarchical multi‐scale (HMS) attention and strip pooling (SP) module to improve prediction with BEV generation. The authors evaluate their model with a Car Learn to Act (CARLA)‐generated synthetic dataset. The experiment results demonstrate that the authors’ model generates a highly accurate representation of the surrounding environment achieving a state‐of‐the‐art result of 61.50% Mean Intersection‐over‐Union (MIoU) across eight classes.

Promotion of Wound Healing Using Nanoporous Silk Fibroin Sponges.

Wound dressings are important for wound repair. The morphology of the biomaterials used in these dressings, and in particular, the pore structure affects tissue regeneration by facilitating attachment and proliferation of cells due to the hierarchical multiscale, water absorbance, and nutrient transport. In the present study, silk fibroin (SF) sponges with walls containing nanopores (SFNS) were prepared from SF nanoparticles generated during the autoclaving of SF solutions, followed by leaching the SF nanoparticles from the freeze-dried sponges of SF. The nano/microporous structure, biofluid absorbance, and porosity of the SF sponges with and without nanopores were characterized. In vitro cell proliferation, in vivo biocompatibility, and wound healing were evaluated with the sponges. The results demonstrated that SFNS had significantly increased porosity and water permeability, as well as cell attachment and proliferation when compared with SF sponges without the nanopores (SFS). Wound dressings were assessed in a rat skin wound model, and SFNS was superior to SFS in accelerating wound healing, supported by vascularization, deposition of collagen, and increased epidermal thickness over 21 days. Hence, such a dressing material with a hierarchical multiscale pore structure could promote cell migration, vascularization, and tissue regeneration independently without adding any growth factor, which would offer a new strategy to design and engineer better-performed wound dressing.

Hierarchical transfer learning for deep learning velocity model building

Deep learning is a promising approach to velocity model building because it has the potential of processing large seismic surveys with minimal resources. By leveraging large quantities of model-gather pairs, neural networks (NNs) can automatically map data to the model space, directly providing a solution to the inverse problem. Such mapping requires big data, which proves prohibitive for 2D and 3D surveys of realistic size. We have developed a transfer learning (TL) strategy. A network is first trained on a smaller subproblem, which then becomes the starting solution to a larger, more difficult data set, akin to the hierarchical multiscale strategy for full-waveform inversion. We perform TL by having subobjectives that escalate in complexity and by first training an NN at estimating horizontally layered velocity models and then proceeding to train an augmented network at estimating 2D dipping layered models. TL improves convergence and allows using a lesser quantity of 2D models for training. For synthetic tests, the structural similarity index measure of 2D interval velocity models in the time domain is [Formula: see text] and the root-mean-square (rms) error is [Formula: see text] m/s. We benchmark our algorithm on the Marmousi2 model and observe that our method can apply to velocity models with continuous deformed layers with dips up to 35°. We benchmark our algorithm on 2D marine field data and produce an rms velocity model that leads to coherent stacking and a time interval velocity model that reproduces salient features of the stacked section. TL expedites and regularizes training and data-driven techniques may be applied to field data with minimal preprocessing even though we lack real target velocity models.

Hierarchical Multiscale Modeling of Thaw-Induced Landslides in Permafrost

Hierarchical Multiscale Modeling of Thaw-Induced Landslides in Permafrost

Grain‐Slip Derived Network Topology to Remarkable Strength–Toughness Combination of Perovskite Film for Flexible Solar Cells

AbstractThe toughness and strength are generally mutually exclusive for most materials. Although the biological materials in nature such as wood, bone, and nacre exhibit outstanding toughness by forming hierarchical multiscale (nano to macro) structures, it is a huge challenge to simultaneously obtain excellent strength and toughness from material synthesis. Here, one kind of network topology is observed by introducing sodium hyaluronate into organometallic halide perovskite film to greatly improve its strength and toughness. The grain slip and grain subdivision under tensile stress are schemed to dissipate the system energy and endow the perovskite film with remarkable toughness. Meanwhile, the subdivided grains linked by sodium hyaluronate through strong interaction result in high strength of perovskite film. As a result, the perovskite films exhibit robust enhancements with the elongation at break from 1.58% to 5.02% and the fracture strength from 23.13 to 55.25 MPa. It is worth noting that the efficiency of inverted flexible perovskite solar cells reaches 20.01% as well as maintains 90% of the initial efficiency after 6000 cycles of bending at a 2 mm curvature radius. This work devises a topology structure to overcome the conflict between toughness and strength of perovskite films for wearable electronics.

Application Analysis of 5G Intelligent Communication Technologies in the Field of Sports Distance Education.

China has entered a new fifth-generation (5G) technological period, college students are relatively fast to absorb new things, and conventional teaching methods are incapable of properly stimulating students' interest in sporting activities. As a result, the 5G integrated teaching paradigm, which is based on communication technology, is an unavoidable reform of sports remote education. For a long time, technology and sports have been inextricably linked. However, new possibilities on the Internet are fast developing, resulting in massive volumes of data. In this paper, we propose a novel 5G framework for efficient sports distance education. Initially, the sports dataset is preprocessed using normalization, and the features are extracted using Principal Component Analysis (PCA). Following feature selection, a Hierarchical Multiscale Convolutional Neural Network (HM-CNN) is used to categorize and initialize the 5G utilizing the Enhanced Transfer Control Protocol (E-TCP) for efficient data transmission. The Elevated Ant Colony Optimization Algorithm improves the performance of the suggested system even more (EACO). The experimental results indicate that 5G integrated education reform based on communication technology may successfully enhance the number of college students' sports population.

Enhancing Hierarchical Multiscale Off-Road Mobility Model by Neural Network Surrogate Model

Abstract A hierarchical multiscale off-road mobility model is enhanced through the development of an artificial neural network (ANN) surrogate model that captures the complex material behavior of deformable terrain. By exploiting the learning capability of neural networks, the incremental stress and strain relationship of granular terrain is predicted by the ANN representative volume elements (RVE) at various states of the stress and strain. A systematic training procedure for ANN RVEs is developed with a virtual tire test rig model for producing training data from the discrete-element (DE) RVEs without relying on computationally intensive full vehicle simulations on deformable terrain. The ANN surrogate RVEs are then integrated into the hierarchical multiscale computational framework as a lower-scale model with the scalable parallel computing capability, while the macroscale terrain deformation is described by the finite element (FE) approach. It is demonstrated with several numerical examples that off-road vehicle mobility performances predicted by the proposed FE-ANN multiscale terrain model are in good agreement with those of the FE-DE multiscale model while achieving a substantial computational time reduction. The accuracy and robustness of the ANN RVE for fine-grain sand terrain are discussed for scenarios not considered in training datasets. Furthermore, a drawbar pull test simulation is presented with the ANN RVE developed with data in the cornering scenario and validated against the full-scale vehicle test data. The numerical results confirm the predictive ability of the FE-ANN multiscale terrain model for off-road mobility simulations.

A hierarchical multiscale strategy for analyzing the impact response of 3D braided composites

A hierarchical multiscale framework has been developed to predict the impact response of 3D braided composites, in which a bridging model(unit cell model), was employed as a bridge between micro- and macro- scale. A paraboloidal elastic-to-plastic damage model for the epoxy, including the hydrostatic effect, was conducted, where a robust algorithm has been implemented to enhance the convergence and accuracy. At the microscale, nine RVEs (representative volume elements) with different fibers distributions were generated in order to enhance reliability and to reduce the size effect for the micromechanical analysis; At the macroscale, firstly, a unit cell model was proposed as a virtual meso-scale based on the geometric topology of 3D braided composites, and then the unit cell model was implicitly introduced into the macromodel based on the transformation matrix. The homogenization method was adopted for transferring the mechanical properties between scales. These constitutive models were implemented by two user defined subroutines (UMAT and VUMAT) in ABAQUS/Standard and ABAQUS/Explicit for micromechanical analysis and impact simulation, respectively. Furthermore, the failure envelop was carried out for matching a better strength criteria for the braiding yarn. Finally, the high-speed impact simulation was employed to validate the effectiveness of the multiscale framework. The results show that the numerical results are agreed with the experiments well.

Holistic multitask regression network for multiapplication shape regression segmentation

A holistic multitask regression approach was implemented to tackle the limitations of clinical image analysis. Standard practice requires identifying multiple anatomic structures in multiple planes from multiple anatomic regions using multiple modalities. The proposed novel holistic multitask regression network (HMR-Net) formulates organ segmentation as a multitask learning problem. Multitask learning leverages the strength of joint task problem solving from capturing task correlations. HMR-Net performs multitask regression by estimating an organ's class, regional location, and precise contour coordinates. The estimation of each coordinate point also corresponds to another regression task. HMR-Net leverages hierarchical multiscale and fused organ features to handle nonlinear relationships between image appearance and distinct organ properties. Simultaneously, holistic shape information is captured by encoding coordinate correlations. The multitask pipeline enables the capturing of holistic organ information (e.g. class, location, shape) to perform shape regression for medical image segmentation. HMR-Net was validated on eight representative datasets obtained from a total of 222 subjects. A mean average precision and dice score reaching up to 0.81 and 0.93, respectively, was achieved on the representative multiapplication database. The generalized model demonstrates comparable or superior performance compared to state-of-the-art algorithms. The high-performance accuracy demonstrates our model as an effective general framework to perform organ shape regression in multiple applications. This method was proven to provide high-contrast sensitivity to delineate even the smallest and oddly shaped organs. HMR-Net's flexible framework holds great potential in providing a fully automatic preliminary analysis for multiple types of medical images.

Synthesis of biomass-based carbon aerogels in energy and sustainability

Carbon aerogels are 3D hierarchical multiscale porous materials with outstanding physicochemical properties such as high specific surface area, low density, high porosity, excellent electrical conductor, good chemical stability, hydrophobicity, and adjustable surface chemistry among others. Unlike conventional carbon aerogels, biomass-based carbon aerogels are economical, environmentally friendly and have nigh inexhaustible precursors, which have generated extensive interest and exhibited outstanding electrocatalysis and adsorption/absorption performance. In this review, we mainly summarized the four main kinds of biomass (cellulose, chitosan, lignin and tannin) as carbon aerogel precursor, and discussed in detail their resource, constitute and optimized synthesis mechanism. Further advice was also given for better utilization of biomass as carbon aerogel precursors.

Modified Hierarchical Multiscale Dispersion Entropy and its Application to Fault Identification of Rotating Machinery

The rotating machinery possesses complicated structures and various fault types, whose health state monitoring is essential for the normal production and operation of the equipment. To distinguish different working states of rotating machinery efficiently and accurately, this article presents a novel approach for extracting fault features of vibration signals called modified hierarchical multiscale dispersion entropy (MHMDE). And on this basis, an innovative approach for fault diagnosis of rotating machinery based on MHMDE, multi-cluster feature selection (MCFS) and particle swarm optimization kernel extreme learning machine (PSO-KELM) is developed. Firstly, MHMDE is employed to extract the high-dimensional fault features of rotating machinery. This approach can effectively overcome the shortcomings that multi-scale entropy only focuses on the information in the low-frequency components but discards the high-frequency information, as well as the significant dropping of efficiency if the number of hierarchical layers of hierarchical entropy is large. Then MCFS is employed to screen the sensitive features from the high-dimensional fault features. Finally, the sensitive feature vectors are input into the PSO-KELM-based fault classifier to complete the rotating machinery fault diagnosis. It is proved that the presented approach can effectively identify different fault states of rotating machinery through three typical examples. Meanwhile, the presented approach is compared with multi-scale dispersion entropy (MDE) and hierarchical dispersion entropy (HDE), etc. The results show that the presented approach possesses more superior performance.

Forging of Hierarchical Multiscale Capabilities for Simulation of Energetic Materials

AbstractWe present new capabilities for investigation of microstructure in energetic material response for both explicit large‐scale and multiscale simulations. We demonstrate the computational capabilities by studying the effect of porosity on the reactive shock response of a coarse‐grain (CG) model of the energetic material cyclotrimethylene trinitramine (RDX), the non‐reactive equation of state for a porous representative volume element (RVE) of CG RDX, and utilization of available supercomputing resources for speculative sampling to accelerate hierarchical multiscale simulations. Small amounts of porosity (up to 4 %) are shown to have significant effect on the initiation of reactive CG RDX using large‐scale reactive dissipative particle dynamics simulations. Non‐reactive RVEs are shown to undergo a porosity‐dependent pore collapse at hydrostatic conditions, and an existing automation framework is shown to be easily modified for the incorporation of microstructure while retaining reliable convergence properties. A novel predictive sampling method based on use of kernel density estimators is shown to effectively accelerate time‐to‐solution in a multiscale simulation, scaling with free CPU cores, while making no assumptions about the underlying physics for the data being analyzed. These multidisciplinary studies of distinct yet connected problems combine to provide methodological insights for high‐fidelity modeling of reactive systems with microstructure.

Hierarchical Multiscale Modeling of Tire–Soil Interaction for Off-Road Mobility Simulation

A high-fidelity computational terrain dynamics model plays a crucial role in accurate vehicle mobility performance prediction under various maneuvering scenarios on deformable terrain. Although many computational models have been proposed using either finite element (FE) or discrete element (DE) approaches, phenomenological constitutive assumptions in FE soil models make the modeling of complex granular terrain behavior very difficult and DE soil models are computationally intensive, especially when considering a wide range of terrain. To address the limitations of existing deformable terrain models, this paper presents a hierarchical FE–DE multiscale tire–soil interaction simulation capability that can be integrated in the monolithic multibody dynamics solver for high-fidelity off-road mobility simulation using high-performance computing (HPC) techniques. It is demonstrated that computational cost is substantially lowered by the multiscale soil model as compared to the corresponding pure DE model while maintaining the solution accuracy. The multiscale tire–soil interaction model is validated against the soil bin mobility test data under various wheel load and tire inflation pressure conditions, thereby demonstrating the potential of the proposed method for resolving challenging vehicle-terrain interaction problems.

A Hierarchical Multiscale Modeling Investigation on the Behavior of Microtextured Regions in Ti-6242 α/β Processing

Ti-6242 is a near alpha titanium alloy, which has excellent high-temperature creep resistance and is widely used in jet engine compressors. This alloy is susceptible to creep fatigue failure under dwell loading below 473 K. The existence of microtextured regions (MTRs) contributes significantly to this fast crack propagation. Mechanical processing in the alpha + beta region has been employed to eliminate MTRs, but the efficiency depends significantly on the applied strain path. Previous investigations based on crystal plasticity finite element (CPFE) simulations have demonstrated the relationship between breakdown efficiency and loading direction. Therein, MTRs with regular geometry and pure initial orientation were used to isolate the effect of loading direction from initial microstructure. In this paper, the behavior of MTRs with realistic initial microstructure was investigated using a hierarchical multiscale modeling framework, and the microscale results were analyzed in detail to understand the behavior of MTRs under different loading conditions. It was shown that a hierarchical multiscale model with realistic initial microstructure at the microscale can reflect the influences from different strain paths, initial orientation distributions, and positions of the region simultaneously. The combined effect of initial orientation distribution and loading direction on the MTR breakdown efficiency is discussed in detail.

Experimental validation and effect of modelling assumptions in the hierarchical multi-scale simulation of the cup drawing of AA6016 sheets

An essential step in the improvement of design strategies for a wide range of industrial deep drawing applications is the development of methods which allow for the precise prediction of shape and processing parameters. Earlier work has demonstrated, in a clear but qualitative manner, the capabilities of the hierarchical multiscale (HMS) model, which predicts the anisotropic plastic properties of metallic materials based on a statistical analysis of microstructure-based anisotropy and a continuous description of the yield locus. The method is implemented into the ABAQUS finite-element software but, until recently, little attention had been paid to other factors which determine the accuracy of a finite element prediction in general, such as mesh size, friction coefficient and rigid/elastic modelling of the tools. Through the analysis of cup drawing, which is a well-established laboratory-scale test relevant to industrial applications, a quantitative comparison is provided between measured cup geometry and punch force and modelling results for commercial AA6016T4 aluminium sheets. The relatively weak earing behaviour of these materials serves to emphasise the small differences still found between model and experiment, which may be addressed by future refinement of the micromechanical component of the HMS. Average cup height and punch force, which is an important process parameter omitted in earlier studies, depend primarily on the friction coefficient and assumptions in the modelling of the tools. Considering the balance between accuracy and precision, it is concluded that the proposed methodology has matured sufficiently to be used as a design tool at industrial level.

Accelerating multi-scale sheet forming simulations by exploiting local macroscopic quasi-homogeneities

Multi-scale simulations are computationally expensive if a two-way coupling is employed. In the context of sheet metal forming simulations, a fine-scale representative volume element (RVE) crystal plasticity (CP) model would supply the Finite Element analysis with plastic properties, taking into account the evolution of crystallographic texture and other microstructural features. The main bottleneck is that the fine-scale model must be evaluated at virtually every integration point in the macroscopic FE mesh. We propose to address this issue by exploiting a verifiable assumption that fine-scale state variables of similar RVEs, as well as the derived properties, subjected to similar macroscopic boundary conditions evolve along nearly identical trajectories. Furthermore, the macroscopic field variables primarily responsible for the evolution of fine-scale state variables often feature local quasi-homogeneities. Adjacent integration points in the FE mesh can be then clustered together in the regions where the field responsible for the evolution shows low variance. This way the fine-scale evolution is tracked only at a limited number of material points and the derived plastic properties are propagated to the surrounding integration points subjected to similar deformation. Optimal configurations of the clusters vary in time as the local deformation conditions may change during the forming process, so the clusters must be periodically adapted. We consider two operations on the clusters of integration points: splitting (refinement) and merging (unrefinement). The concept is tested in the Hierarchical Multi-Scale (HMS) framework [1] that computes macroscopic deformations by means of the FEM, whereas the micro-structural evolution at the individual FE integration points is predicted by a CP model. The HMS locally and adaptively approximates homogenized stress responses of the CP model by means of analytical plastic potential or yield criterion function. Our earlier work investigated simple test cases [2]. In this contribution we present a deep drawing process simulated using the HMS framework improved to exploit local quasi-homogeneities. We conclude that large performance gains (e.g. a speedup of 25) are obtained at the expense of introducing only a minor (e.g. below 1%) modelling error compared to the HMS simulation with no clustering of integration points.

A systematic study of pool boiling heat transfer on structured porous surfaces: From nanoscale through microscale to macroscale

An experimental study has been conducted to examine the effects of macroscale, microscale, and nanoscale surface modifications in water pool boiling heat transfer and to determine the different heat transfer enhancing mechanisms at different scales. Nanostructured surfaces are created by acid etching, while microscale and macroscale structured surfaces are synthesized through a sintering process. Six structures are studied as individual and collectively integrated surfaces from nanoscale through microscale to macroscale: polished plain, flat nanostructured, flat porous, modulated porous, nanostructured flat porous, and nanostructured modulated porous. Boiling performance is measured in terms of critical heat flux (CHF) and heat transfer coefficient (HTC). Both HTC and CHF have been greatly improved on all modified surfaces compared to the polished baseline. Hierarchical multiscale surfaces of integrated nanoscale, microscale, and macroscale structures have been proven to have the most significant improvements on HTC and CHF. The CHF and HTC of the hierarchical multiscale modulated porous surface have achieved the most significant improvements of 350% and 200% over the polished plain surface, respectively. Experimental results are compared to the predictions of a variety of theoretical models with an attempt to reveal the different heat transfer enhancing mechanisms at different scales. It is concluded that models for the structured surfaces at all scales need to be further developed to be able to have good quantitative predictions of CHFs on structured surfaces.