Search NASA⌕ Search

SEARCH · Search NASA

Results for “Randomized methods”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 649 records · Page 36
All sourcesNASA NTRSDOE OSTI
DOE OSTI2015

Solitons and protein folding: An In Silico experiment

Protein folding [1] is the process of formation of a functional 3D structure from a random coil — the shape in which amino-acid chains leave the ribosome. Anfinsen’s dogma states that the native 3D shape of a protein is completely determined by protein’s amino acid sequence. Despite the progress in understanding the process rate and the success in folding prediction for some small proteins, with presently available physics-based methods it is not yet possible to reliably deduce the shape of a biologically active protein from its amino acid sequence. The protein-folding problem endures as one of the most important unresolved problems in science; it addresses the origin of life itself. Furthermore, a wrong fold is a common cause for a protein to lose its function or even endanger the living organism. Soliton solutions of a generalized discrete non-linear Schrödinger equation (GDNLSE) obtained from the energy function in terms of bond and torsion angles κ and τ provide a constructive theoretical framework for describing protein folds and folding patterns [2]. Here we study the dynamics of this process by means of molecular-dynamics simulations. The soliton manifestation is the pattern helix–loop–helix in the secondary structure of the protein, which explains the importance of understanding loop formation in helical proteins. We performed in silico experiments for unfolding one subunit of the core structure of gp41 from the HIV envelope glycoprotein (PDB ID: 1AIK [3]) by molecular-dynamics simulations with the MD package GROMACS. We analyzed 80 ns trajectories, obtained with one united-atom and two different all-atom force fields, to justify the side-chain orientation quantification scheme adopted in the studies and to eliminate force-field based artifacts. Our results are compatible with the soliton model of protein folding and provide first insight into soliton-formation dynamics.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2015

An adaptive multi-level simulation algorithm for stochastic biological systems

Discrete-state, continuous-time Markov models are widely used in the modeling of biochemical reaction networks. Their complexity often precludes analytic solution, and we rely on stochastic simulation algorithms (SSA) to estimate system statistics. The Gillespie algorithm is exact, but computationally costly as it simulates every single reaction. As such, approximate stochastic simulation algorithms such as the tau-leap algorithm are often used. Potentially computationally more efficient, the system statistics generated suffer from significant bias unless tau is relatively small, in which case the computational time can be comparable to that of the Gillespie algorithm. The multi-level method [Anderson and Higham, “Multi-level Monte Carlo for continuous time Markov chains, with applications in biochemical kinetics,” SIAM Multiscale Model. Simul. 10(1), 146–179 (2012)] tackles this problem. A base estimator is computed using many (cheap) sample paths at low accuracy. The bias inherent in this estimator is then reduced using a number of corrections. Each correction term is estimated using a collection of paired sample paths where one path of each pair is generated at a higher accuracy compared to the other (and so more expensive). By sharing random variables between these paired paths, the variance of each correction estimator can be reduced. This renders the multi-level method very efficient as only a relatively small number of paired paths are required to calculate each correction term. In the original multi-level method, each sample path is simulated using the tau-leap algorithm with a fixed value of τ. This approach can result in poor performance when the reaction activity of a system changes substantially over the timescale of interest. By introducing a novel adaptive time-stepping approach where τ is chosen according to the stochastic behaviour of each sample path, we extend the applicability of the multi-level method to such cases. We demonstrate the efficiency of our method using a number of examples.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2015

Medium scale carbon nanotube thin film integrated circuits on flexible plastic substrates

The present invention provides device components geometries and fabrication strategies for enhancing the electronic performance of electronic devices based on thin films of randomly oriented or partially aligned semiconducting nanotubes. In certain aspects, devices and methods of the present invention incorporate a patterned layer of randomly oriented or partially aligned carbon nanotubes, such as one or more interconnected SWNT networks, providing a semiconductor channel exhibiting improved electronic properties relative to conventional nanotubes-based electronic systems.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2025

Randomized Algorithms for Low-Rank Matrix and Tensor Decompositions

This paper surveys randomized algorithms in numerical linear algebra for low-rank decompositions of matrices and tensors. The survey begins with a review of classical matrix algorithms that can be accelerated by randomized dimensionality reduction, such as the singular value decomposition (SVD) or interpolative (ID) and CUR decompositions. Recent advances in randomized dimensionality reduction are discussed, including new methods of fast matrix sketching and sampling techniques, which are incorporated into classical matrix algorithms for fast low-rank matrix approximations. The extension of randomized matrix algorithms to tensors is then explored for several low-rank tensor decompositions in the CP and Tucker formats, including the higher-order SVD, ID, and CUR decomposition.

Pearce, Katherine J. [The University of Texas at A↗
DOE OSTI2016

Kinetic Monte Carlo Simulation of Hydrogen Diffusion in Tungsten

Nuclear fusion is considered as the most promising sustainable energy source. If achieved, it could provide electricity to the mankind for millions of years. During the reaction, the fusion ash is continuously removed from the magnetic gap, the divertor region, and the divertor facing component is subjected to intense plasma bombardment. It is proposed to use tungsten as the divertor facing material, because it has low physical sputtering yield and superior thermomechanical properties. However, the high mobility of hydrogen in tungsten and the high bombardment flux bring the hydrogen isotopes deep into the tungsten tile. The presence of hydrogen in tungsten tends to weaken the gram boundaries and produce surface blistering. The ejection of tungsten particulates into the plasma core will cool the reaction and cease the fusion reactions. Therefore, it is important to have a detailed understanding of the hydrogen diffusion in tungsten bulk. The KMC method is a stochastic computer simulation to predict the time evolution of a certain system. Typically, at a given time, the system will evolve through a few possible paths with known probabilities. In each step, the KMC method analyzes the entire system and randomly determines the system evolving events based on all transition rates of all possible events and advance the physical time. This process is repeated till the simulation of the entire physical time is completed. In previous work, the authors developed a kinetic Monte Carlo (KMC) algorithm and program to simulate hydrogen diffusion process on tungsten reconstructed surface. In this article, this KMC method is extended to model the hydrogen diffusion in tungsten bulk. The diffusion coefficients are calculated and fitted to the Arrhenius equation. The result matches the experiment measurement very well. KMC diffusion simulation needs the locations of the hydrogen trapping sites, diffusion paths and the corresponding diffusion energy barriers. The hydrogen interstitial sites in tungsten crystal have been thoroughly studied by density function theory (DFT) and molecular dynamics (MD). Johnson and Carter used DFT to determine the preferred location of the hydrogen interstitial sites m tungsten crystal, which are the tetrahedral sites (T-sites) in body-centered-cubic (bcc) lattice surfaces. All T-sites are located on the surfaces of a lattice cell. There are four T-sites per surface and twelve T-sites per lattice cell. The hydrogen atoms diffuse by traveling from one T-site to one of its nearest four T-site neighbors. Thus, each T-site is surrounded by two neighbors on the same surface, and one neighbor on each of the two adjacent surfaces perpendicular to the surface holding the center T-site. The diffusion barrier between two neighboring T-sites was calculated to be 0.38 eV. Knowing the diffusion path, energy barrier, and the crystal structure, the KMC simulation is readily to be performed. Based on other researchers' DFT calculations, the hydrogen stable interstitial site in tungsten bulk is determined as the tetrahedral site on the surface of the bee lattice. Each T-site has 4 neighbors and the energy barrier between two adjacent T-sites are estimated to be 0.38 eV. Based on those facts and the KMC theory, a MATLAB program is developed to model the hydrogen diffusion in tungsten bulk. In the MATLAB program, substrate temperature, hydrogen concentration, physical time to be simulated, and the number of cycles are all customizable. To compare the KMC determined diffusion coefficients with experiment results in the literature, the simulations are performed for temperature 300 K - 2500 K and low hydrogen concentration (0.1%), mimicking the experiment condition. The calculated diffusion coefficients match the reference values measured by the experiment under low hydrogen concentration condition. The results are also fitted to the Arrhenius equation. The pre-exponential factor and the average activation energy in Arrhenius equation are obtained. The good agreement between simulation and experiment proves that the developed KMC algorithm truly reflect the well-established theory of interstitial hydrogens in tungsten. The developed program could easily be used to model the diffusion process and predict the diffusion coefficients under other concentrations and temperatures. This work is part of the ongoing development of hydrogen diffusion in tungsten integrating surface diffusion (previous work), bulk diffusion (this work), and adsorption/desorption on surface (future work)

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2012

THE EFFECTS OF VIEWING ANGLE ON THE MASS DISTRIBUTION OF EXOPLANETS

We present a mathematical method to statistically decouple the effects of unknown inclination angles on the mass distribution of exoplanets that have been discovered using radial-velocity (RV) techniques. The method is based on the distribution of the product of two random variables. Thus, if one assumes a true mass distribution, the method makes it possible to recover the observed distribution. We compare our prediction with available RV data. Assuming that the true mass function is described by a power law, the minimum mass function that we recover proves a good fit to the observed distribution at both mass ends. In particular, it provides an alternative explanation for the observed low-mass decline, usually explained as sample incompleteness. In addition, the peak observed near the low-mass end arises naturally in the predicted distribution as a consequence of imposing a low-mass cutoff in the true distribution. If the low-mass bins below 0.02 M{sub J} are complete, then the mass distribution in this regime is heavily affected by the small fraction of lowly inclined interlopers that are actually more massive companions. Finally, we also present evidence that the exoplanet mass distribution changes form toward low mass, implying that a single power law may not adequately describe the sample population.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2019

Evaluating Optimization Strategies for Engine Simulations Using Machine Learning Emulators

This work evaluates different optimization algorithms for computational fluid dynamics (CFD) simulations of engine combustion. Due to the computational expense of CFD simulations, emulators built with machine learning algorithms were used as surrogates for the optimizers. Two types of emulators were used: a Gaussian process (GP) and a weighted variety of machine learning methods called SuperLearner (SL). The emulators were trained using a dataset of 2048 CFD simulations that were run concurrently on a supercomputer. The design of experiments (DOE) for the CFD runs was obtained by perturbing nine input parameters using a Monte-Carlo method. The CFD simulations were of a heavy duty engine running with a low octane gasoline-like fuel at a partially premixed compression ignition mode. Ten optimization algorithms were tested, including types typically used in research applications. Each optimizer was allowed 800 function evaluations and was randomly tested 100 times. The optimizers were evaluated for the median, minimum, and maximum merits obtained in the 100 attempts. Some optimizers required more sequential evaluations, thereby resulting in longer wall clock times to reach an optimum. The best performing optimization methods were particle swarm optimization (PSO), differential evolution (DE), GENOUD (an evolutionary algorithm), and micro-genetic algorithm (GA). These methods found a high median optimum as well as a reasonable minimum optimum of the 100 trials. Moreover, all of these methods were able to operate with less than 100 successive iterations, which reduced the wall clock time required in practice. Two methods were found to be effective but required a much larger number of successive iterations: the DIRECT and MALSCHAINS algorithms. A random search method that completed in a single iteration performed poorly in finding optimum designs but was included to illustrate the limitation of highly concurrent search methods. The last three methods, Nelder–Mead, bound optimization by quadratic approximation (BOBYQA), and constrained optimization by linear approximation (COBYLA), did not perform as well.

Probst, Daniel M.↗
DOE OSTI2025

Microscopic calculations with noniterative finite amplitude methods and the application to neutron radiative captures and inelastic scatterings

We derive the fully self-consistent quasiparticle random-phase approximation (QRPA) equations with noniterative finite amplitude methods and calculate the transition strengths of giant resonances. Then, we apply the QRPA results to both neutron radiative capture calculations based on the statistical Hauser-Feshbach theory and inelastic scattering calculations based on distorted-wave Born approximation (DWBA). We compare the calculated results with available experimental data and demonstrate how our approach can reproduce giant resonances and various nuclear reactions.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2022

Random strain induced correlations in materials with intertwined nematic and magnetic orders

Electronic nematicity is rarely observed as an isolated instability of a correlated electron system. Instead, in iron pnictides and in certain cuprates and heavy-fermion materials, nematicity is intertwined with an underlying spin-stripe or charge-stripe state. As a result, random strain, ubiquitous in any real crystal, creates both random-field disorder for the nematic degrees of freedom and random-bond disorder for the spin or charge ones. Here, we put forward an Ashkin-Teller model with random Baxter fields to capture the dual role of random strain in nematic systems for which nematicity is a composite order arising from a stripe state. Using Monte Carlo to simulate this random Baxter-field model, we find not only the expected break-up of the system into nematic domains, but also the emergence of nontrivial disorder-promoted magnetic correlations. Such correlations enhance and tie up the fluctuations associated with the two degenerate magnetic stripe states from which nematicity arises, leaving characteristic signatures in the spatial profile of the magnetic domains, in the configurational space of the spin variables, and in the magnetic noise spectrum. We discuss possible experimental manifestations of these effects in iron-pnictide superconductors. Furthermore, our work establishes the random Baxter-field model as a more complete alternative to the random-field Ising model to describe complex electronic nematic phenomena in the presence of disorder.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2022

Proposal and application of ROM-Lasso method for sensitivity coefficient evaluation

We propose a novel method for evaluating sensitivity coefficients of neutronics parameters to cross sections, so-called the reduced-order modeling technique ROM-Lasso. In this method, cross sections of interest are randomly sampled, and corresponding perturbed core analyses are performed. Then, the sensitivity coefficient vector of the higher-level model is expanded via the active subspace bases obtained with the lower-level model whose dimensional complexity is smaller than that of the higher-level model, and the expansion coefficients are estimated by the Lasso regression. A unique feature of the ROM-Lasso method allows the use of different bases optimized for each neutronics parameter. We conducted a verification calculation for an accelerator-driven system and demonstrated that the ROM-Lasso method can reproduce the sensitivity coefficients with a much smaller number of forward calculations than the direct method. The proposed method can be used to practically evaluate sensitivity coefficients. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2014

Cosmological constraints from Sunyaev-Zeldovich cluster counts: An approach to account for missing redshifts

The accumulation of redshifts provides a significant observational bottleneck when using galaxy cluster surveys to constrain cosmological parameters. We propose a simple method to allow the use of samples where there is a fraction of the redshifts that are not known. The simplest assumption is that the missing redshifts are randomly extracted from the catalog, but the method also allows one to take into account known selection effects in the accumulation of redshifts. We quantify the reduction in statistical precision of cosmological parameter constraints as a function of the fraction of missing redshifts for simulated surveys, and also investigate the impact of making an incorrect assumption for the distribution of missing redshifts.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2012

Randomized, Multicenter Trial on the Effect of Radiation Therapy on Plantar Fasciitis (Painful Heel Spur) Comparing a Standard Dose With a Very Low Dose: Mature Results After 12 Months' Follow-Up

Purpose: To conduct a randomized trial of radiation therapy for painful heel spur, comparing a standard dose with a very low dose. Methods and Materials: Sixty-six patients were randomized to receive radiation therapy either with a total dose of 6.0 Gy applied in 6 fractions of 1.0 Gy twice weekly (standard dose) or with a total dose of 0.6 Gy applied in 6 fractions of 0.1 Gy twice weekly (low dose). In all patients lateral opposing 4- to 6-MV photon beams were used. The results were measured using a visual analogue scale, the Calcaneodynia score, and the SF12 health survey. The fundamental phase of the study ended after 3 months, and the follow-up was continued up to 1 year. Patients with insufficient pain relief after 3 months were offered reirradiation with the standard dosage at any time afterward. Results: Of 66 patients, 4 were excluded because of withdrawal of consent or screening failures. After 3 months the results in the standard arm were highly significantly superior compared with those in the low-dose arm (visual analogue scale, P=.001; Calcaneodynia score, P=.027; SF12, P=.045). The accrual of patients was stopped at this point. Further evaluation after 12 months' follow-up showed the following results: (1) highly significant fewer patients were reirradiated in the standard arm compared with the low-dose arm (P<.001); (2) the results of patients in the low-dose arm who were reirradiated were identical to those in the standard arm not reirradiated (reirradiation as a salvage therapy if the lower dose was ineffective); (3) patients experiencing a favorable result after 3 months showed this even after 12 months, and some results even improved further between 3 and 12 months. Conclusions: This study confirms the superior analgesic effect of radiation therapy with 6-Gy doses on painful heel spur even for a longer time period of at least 1 year.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2020

Structure–activity relationship-based chemical classification of highly imbalanced Tox21 datasets

Abstract The specificity of toxicant-target biomolecule interactions lends to the very imbalanced nature of many toxicity datasets, causing poor performance in Structure–Activity Relationship (SAR)-based chemical classification. Undersampling and oversampling are representative techniques for handling such an imbalance challenge. However, removing inactive chemical compound instances from the majority class using an undersampling technique can result in information loss, whereas increasing active toxicant instances in the minority class by interpolation tends to introduce artificial minority instances that often cross into the majority class space, giving rise to class overlapping and a higher false prediction rate. In this study, in order to improve the prediction accuracy of imbalanced learning, we employed SMOTEENN, a combination of Synthetic Minority Over-sampling Technique (SMOTE) and Edited Nearest Neighbor (ENN) algorithms, to oversample the minority class by creating synthetic samples, followed by cleaning the mislabeled instances. We chose the highly imbalanced Tox21 dataset, which consisted of 12 in vitro bioassays for > 10,000 chemicals that were distributed unevenly between binary classes. With Random Forest (RF) as the base classifier and bagging as the ensemble strategy, we applied four hybrid learning methods, i.e., RF without imbalance handling (RF), RF with Random Undersampling (RUS), RF with SMOTE (SMO), and RF with SMOTEENN (SMN). The performance of the four learning methods was compared using nine evaluation metrics, among which F 1 score, Matthews correlation coefficient and Brier score provided a more consistent assessment of the overall performance across the 12 datasets. The Friedman’s aligned ranks test and the subsequent Bergmann-Hommel post hoc test showed that SMN significantly outperformed the other three methods. We also found that a strong negative correlation existed between the prediction accuracy and the imbalance ratio (IR), which is defined as the number of inactive compounds divided by the number of active compounds. SMN became less effective when IR exceeded a certain threshold (e.g., > 28). The ability to separate the few active compounds from the vast amounts of inactive ones is of great importance in computational toxicology. This work demonstrates that the performance of SAR-based, imbalanced chemical toxicity classification can be significantly improved through the use of data rebalancing.

Idakwo, Gabriel↗
DOE OSTI2016

Predicting Correlation Coefficients for Monte Carlo Eigenvalue Simulations

Monte Carlo methods are most often considered as a reference for neutron transport simulations since very limited approximations are made about nuclear data and system geometry. To report uncertainty of any tally evaluated as generation averages, the sample variance is divided by the number of active generations, which is based on the assumption that the neutron generations are independent. Correlation effects between neutrons in multiplying systems, particularly when performing power iteration to evaluate eigenvalues have been observed in previous work. Neglecting the correlation effect results in an underestimate of uncertainty reported by Monte Carlo calculations. Previous work has also proposed methods to predict the underestimation ratio. Yamamoto et al expanded the fission source distribution with diffusion equation modes, performed numerical simulation of the AR (autoregressive) process of the expansion coefficients and used the correlation of the AR process to predict that of the Monte Carlo eigenvalue simulation. Sutton applied the discretized phase space (DPS) approach to predict the underestimation ratio but the method cannot predict the ratio when one neutron generates offspring in different phase space regions or generates a random number of offspring. This paper presents a method to predict the correlation effect with the model of multi-type branching processes (MBP). The method requires simulations for one generation of neutrons without knowing the source distribution and can predict the underestimation ratio for the cases where the traditional DPS approach does not work. The generation-to-generation correlation determines the convergence rate of active generations, the bias of variance estimator for each generation and the underestimation ratio of variance estimator for tallies averaged over active generations. The generation-to-generation correlation is characterized by the Auto-Correlation Coefficients (ACC) between tallies from different generations. (authors)

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2018

Split‐dose 4‐L polyethylene glycol regimen for patients with previous colorectal surgery in bowel preparation before colonoscopy: A randomized, controlled, single‐blind study

OBJECTIVE We aimed to investigate the efficacy of a split‐dose 4‐L polyethylene glycol (PEG) regimen for the quality of bowel preparation in Asian patients with previous colorectal surgery for colorectal cancer (CRC). METHODS This was a prospective, single‐center, randomized controlled, endoscopist‐blinded study. Patients with previous colorectal surgery for CRC were randomly allocated to a routine, morning‐only 2‐L PEG (2‐MO) group or a split‐dose 4‐L PEG (4‐SD) group. The primary outcome was a successful bowel preparation rate. Secondary outcomes were polyp detection rate (PDR), adenoma detection rate (ADR), patient compliance, satisfaction, tolerance, willingness to repeat the preparation and difficulty of the bowel preparation process. RESULTS In total, 187 patients were included (93 in the 2‐MO group, 94 in the 4‐SD group) in this study. The rate of successful bowel preparation in the 4‐SD group was higher than in the 2‐MO group (89.4% vs 66.7%, P < 0.001) in an intention‐to‐treat analysis according to the Aronchick score. Patients’ satisfaction with the bowel preparation process in the 4‐SD group was superior to that in the 2‐MO group (93.4% vs 82.2%, P = 0.021). No significant differences were detected in PDR, ADR, patient compliance, tolerance, willingness to repeat the preparation or difficulty of the bowel preparation process. CONCLUSIONS The 4‐SD PEG regimen was superior to a routine, morning‐only 2‐L PEG preparation for bowel preparation in an Asian population with previous colorectal surgery.

Liu, Zhu↗
DOE OSTI2016

Photonic mesophases from cut rod rotators

The photonic band properties of random rotator mesophases are calculated using supercell methods applied to cut rods on a hexagonal lattice. Inspired by the thermodynamic mesophase for anisotropic building blocks, we vary the shape factor of cut fraction for the randomly oriented basis. We find large, stable bandgaps with high gap isotropy in the inverted and direct structures as a function of cut fraction, dielectric contrast, and filling fraction. Bandgap sizes up to 34.5% are maximized at high dielectric contrast for rods separated in a matrix. The bandgaps open at dielectric contrasts as low as 2.0 for the transverse magnetic polarization and 2.25 for the transverse electric polarization. Additionally, the type of scattering that promotes the bandgap is correlated with the effect of disorder on bandgap size. Slow light properties are investigated in waveguide geometry and slowdown factors up to 5 × 10{sup 4} are found.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2016

Methods for resistive switching of memristors

The present invention is directed generally to resistive random-access memory (RRAM or ReRAM) devices and systems, as well as methods of employing a thermal resistive model to understand and determine switching of such devices. In particular example, the method includes generating a power-resistance measurement for the memristor device and applying an isothermal model to the power-resistance measurement in order to determine one or more parameters of the device (e.g., filament state).

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2016

3D Representative Volume Element Reconstruction of Fiber Composites via Orientation Tensor and Substructure Features

To provide a seamless integration of manufacturing processing simulation and fiber microstructure modeling, two new stochastic 3D microstructure reconstruction methods are proposed for two types of random fiber composites: random short fiber composites, and Sheet Molding Compounds (SMC) chopped fiber composites. A Random Sequential Adsorption (RSA) algorithm is first developed to embed statistical orientation information into 3D RVE reconstruction of random short fiber composites. For the SMC composites, an optimized Voronoi diagram based approach is developed for capturing the substructure features of SMC chopped fiber composites. The proposed methods are distinguished from other reconstruction works by providing a way of integrating statistical information (fiber orientation tensor) obtained from material processing simulation, as well as capturing the multiscale substructures of the SMC composites.

Li, Yi↗