Wide Filter Research Articles

Invariance embedded physics-infused deep neural network-based sub-grid scale models for turbulent flows

In this paper, we present two novel physics-infused neural network (NN) architectures that satisfy various invariance conditions for constructing efficient and robust Deep Learning (DL)-based sub-grid scale (SGS) turbulence models for use in Large Eddy Simulation (LES) procedures widely used in various fluid engineering applications. The first architecture, called tensor basis neural networks (TBNN), recently proposed in the context of Reynolds-averaged Navier–Stokes (RANS) modeling, and introduced herein for SGS modeling of wall-bounded turbulence, embeds the analytical expansion of the SGS turbulence stresses into integrity basis tensors composed of the symmetric and anti-symmetric parts of the resolved velocity gradient tensor, thus incorporating the Galilean, rotational and reflectional invariances. In our second approach, a relatively simple yet powerful architecture, called Galilean Invariance embedded Neural Network (GINN) incorporates the Galilean invariance, and takes in as inputs the independent components of the integrity basis tensors in addition to the invariant inputs in a single input layer. Explicit filtering of data from direct simulations of the canonical channel flow at two friction Reynolds numbers Reτ≈395 and 590 provided accurate data for training and testing these models. Both sets of models are used to predict the SGS stresses for feature datasets generated with different filter widths, and at different Reynolds numbers. The GINN model yields less prediction error on test datasets (mean squared error ∼0.4) compared to the TBNN model (mean squared error ∼0.5). Upon comparison, it is revealed that GINN has better feature extraction capacity owing to its ability to establish relations between and extract information from cross-components of the integrity basis tensors and the SGS stresses. Based on their predictive performance, both models proposed herein have shown great promise for use in a posteriori actual LESs. The present work illustrates the significance of physics infusion as well as invariance embedding into NN architecture for constructing efficient and robust DL-based SGS turbulent models in achieving superior predictive efficacy.

Multi-constraint point set registration with redundant point removal for the registration of coronary arteries

BackgroundCoronary artery disease (CAD) is the leading cause of death worldwide. The registration of the coronary artery at different phases can help radiologists explore the motion patterns of the coronary artery and assist in the diagnosis of CAD. However, there is no automatic and easy-to-execute method to solve the missing data problem that occurs at the endpoints of the coronary artery tree. This paper proposed a non-rigid multi-constraint point set registration with redundant point removal (MPSR-RPR) algorithm to tackle this challenge. MethodsFirstly, the MPSR-RPR algorithm roughly registered two coronary artery point sets with the pre-set smoothness regularization parameter and Gaussian filter width value. The moving coherent, local feature, and the corresponding relationship between bifurcation point pairs were exploited as the constraints. Next, the spatial geometry information of the coronary artery was utilized to automatically recognize the vessel endpoints and to delete the redundant points of the coronary artery. Finally, the algorithm continued carrying out the multi-constraint registration with another group of the pre-set parameters to improve the alignment performance. ResultsThe experimental results demonstrated that the MPSR-RPR algorithm achieved a significantly lower mean value of the modified Hausdorff distance (MHD) compared to the other state-of-the-art methods for addressing the serious missing data in the left and right coronary arteries. ConclusionThis study demonstrated the effectiveness of the proposed algorithm in aligning coronary arteries, providing significant value in assisting in the diagnosis of coronary artery and myocardial lesions.

A parameter-free LES model for anisotropic mesh adaptivity

Balancing accuracy and computational cost is a challenge in the modelling of turbulent flows. A widely used method for turbulence modelling is large–eddy simulation (LES). LES allows one to describe large scale flow features at a reasonable computational cost compared to the more accurate direct numerical simulation (DNS), making it a popular choice for engineering applications. One strategy to balance accuracy and cost with LES is through the use of mesh adaptivity, which allows the degrees of freedom in a problem to be reduced by changing spatial discretisation. However, mesh adaptivity can affect accuracy when using the standard Smagorinsky LES model with an implicit filter, considering that the parameter Cs is highly dependent on the filter width, which depends on mesh resolution. This work is aimed to develop an LES model that does not require any user–defined parameters and is suitable for mesh adaptivity with implicit filter. In this study we introduce a parameter–free LES model incorporating an anisotropic eddy–viscosity formulation combined with anisotropic mesh adaptivity. In our model, the parameter Cs in the eddy–viscosity formulation of the Smagorinsky model, is replaced by a function that evaluates the relative location of turbulence fluctuations in each element with respect to the turbulence spectrum inertial range. The anisotropic formulation of the eddy–viscosity allows for the application of an appropriate filter width in different directions, improving accuracy. Additionally, the mesh adaptivity algorithm assesses the local turbulence fluctuations via local Reynolds number and vortex identification criteria. This assessment leads to the refinement of regions with higher turbulence fluctuations down to the smallest scale limit in the inertial range in the corresponding direction, and also leads to the coarsening of regions without turbulence fluctuation up to largest scale limit in the inertial range. This method is tested using a flow past a sphere test case. The results are compared both qualitatively and quantitatively to results obtained with the standard Smagorinsky model, and demonstrate the better performance of our method with lower computational cost. This allows us to simulate large Reynolds number cases and compare our quantitative results to experimental results found in the literature, emphasising that our method produces good results at reasonable computational cost.

Probabilistic deep learning of turbulent premixed combustion

A probabilistic data-driven approach that models the filtered reaction rate in large-eddy simulation (LES) is investigated. We propose a novel framework that incorporates a conditional generative adversarial network and a Gaussian mixture model to take into account the statistical fluctuations that are present in LES of turbulent reacting flows due to non-resolved subgrid structures, which cannot be predicted by purely deterministic models and machine learning algorithms. The data from a direct numerical simulation of turbulent premixed combustion are spatially filtered using a wide range of filter widths and employed for the training. We extract physically relevant parameters from the database and reduce the input features to the network to the most influential ones based on the result of feature importance analysis. The trained model is then tested on unseen timesteps and untrained LES filter widths, where it is able to accurately predict the distribution of the filtered reaction rate.

Processing Receiver Function Data in Mentawai Region, Sumatra

The Mentawai Islands are located in the western part of Sumatra, in the Indo-Australian and Eurasian plate subduction zone. Because of these tectonic settings, the Mentawai Region has a high level of seismicity, making it particularly appealing for subsurface structure study. We used the receiver function method in this study to see response receiver function on plate subduction zone and compare the two deconvolution. This method uses the conversion of P-to-S waves, which have a larger amplitude in the radial component than vertical waves. The deconvolution technique is used to extract the two components of the receiver function signal. Water level deconvolution and iterative deconvolution were employed in this study. In practice, any deconvolution method cannot produce a suitable receiver function due to various methodologies that cannot be used in all conditions. Therefore, an evaluation is needed of each deconvolution method used to obtain the best results. This study utilizes two teleseismic stations located in the north and south of Mentawai, which were obtained through the IRIS website, and applied the Butterworth filter to obtain organized and noise-free data. The first results show that the water level is stronger than the iterative, which is more sensitive to the Gaussian filter width parameter. However, they produced almost identical results at both stations, precisely the arrival time of Ps, which was in 3-4 seconds and was followed by a low velocity.

A priori test of perfectly stirred reactor approach to evaluating mean fuel consumption and heat release rates in highly turbulent premixed flames

Unsteady three-dimensional Direct Numerical Simulation (DNS) data obtained earlier by Dave et al. ( J Fluid Mech 2020; 884: A46) from a statistically planar and one-dimensional, highly turbulent, moderately lean hydrogen-air flame propagating in a box are processed to perform a priori test of perfectly stirred reactor model. The test aims in particularly at estimating mesh resolution (or filter width within large eddy simulation framework) required to neglect variations in the temperature and mixture composition within a computational cell when evaluating mean (or filtered) fuel consumption and heat release rates. For this purpose, fuel consumption and heat release rates sampled directly from the DNS data and averaged over a cube of width [Formula: see text] are compared with fuel consumption and heat release rates calculated using the temperature and species concentrations averaged over the same cube. Moreover, turbulent burning velocities computed by integrating the former and latter rates are compared with one another. A ratio of [Formula: see text] to a laminar flame thickness [Formula: see text] is varied from 0.44 to 1.8. The obtained results indicate that the tested simple approach performs reasonably well (poor) if [Formula: see text] ([Formula: see text], respectively). This result is further supported by directly filtering fuel consumption rate in a laminar premixed flame. The values of the thickness [Formula: see text], calculated using detail chemical mechanisms for different fuels under elevated temperatures and pressures associated with combustion in piston engines, indicate that it is difficult to satisfy the constraint of [Formula: see text] in contemporary unsteady multidimensional numerical simulations of turbulent burning in such engines.

Stable Boundary Layers and Subfilter-Scale Motions

Recent high-resolution large-eddy simulations (LES) of a stable atmospheric boundary layer (SBL) with mesh sizes N=(5123,10243,20483) or mesh spacings ▵=(0.78,0.39,0.2) m are analyzed. The LES solutions are judged to be converged based on the good collapse of vertical profiles of mean winds, temperature, and low-order turbulence moments, i.e., fluxes and variances, with increasing N. The largest discrepancy is in the stably stratified region above the low-level jet. Subfilter-scale (SFS) motions are extracted from the LES with N=20483 and are compared to sonic anemometer fields from the horizontal array turbulence study (HATS) and its sequel over the ocean (OHATS). The results from the simulation and observations are compared using the dimensionless resolution ratio Λw/▵f where ▵f is the filter width and Λw is a characteristic scale of the energy-containing eddies in vertical velocity. The SFS motions from the observations and LES span the ranges 0.1<Λw/▵f<20 and are in good agreement. The small, medium, and large range of Λw/▵f correspond to Reynolds-averaged Navier–Stokes (RANS), the gray zone (a.k.a. “Terra Incognita”), and fine-resolution LES. The gray zone cuts across the peak in the energy spectrum and then flux parameterizations need to be adaptive and account for partially resolved flux but also “stochastic” flux fluctuations that represent the turbulent correlation between the fluctuating rate of strain and SFS flux tensors. LES data with mesh 20483 will be made available to the research community through the web and tools provided by the Johns Hopkins University Turbulence Database.

A data-driven approach to model enstrophy transfers in large eddy simulation of forced two-dimensional turbulence

Contemporary research on the application of data-driven techniques to modeling subgrid closure in two-dimensional turbulence has been limited. Furthermore, the incorporation of the enstrophy cascade and other two-dimensional turbulence-specific physics has received insufficient attention. To address these gaps, a novel physics-based shallow feed-forward neural network framework was designed in this study to model subgrid closure in three selected cases of forced two-dimensional turbulence with a forcing that provides energy and enstrophy at a particular wavenumber. As a novel approach, we trained our framework to learn the subgrid vorticity transport vector from a set of appropriate resolved flow variables. Another framework used in recent works which directly learned the subgrid forcing field was also investigated. Both frameworks were assessed using a priori and a posteriori tests for two selected filter widths. Both frameworks performed accurately for the lower filter width but less accurately for the higher filter width. However, we demonstrate that our new framework has wider usefulness for model diagnosis. Ad hoc clipping procedures were used to make the models more generalizable to higher filter widths, and stable and consistent a posteriori tests were observed for all test cases and filter widths when the subgrid forcing field was modified to enhance the model's subgrid dissipative characteristics. In contrast, modifying the enstrophy fluxes did not perform as consistently. These findings demonstrate the potential of the novel physics-based framework for improving subgrid modeling in two-dimensional turbulence.

A posteriori error estimates for the Large Eddy Simulation applied to incompressible fluids

We study the two dimensional time dependent Large Eddy Simulation method applied to the incompressible Navier–Stokes system with Smagorinsky’s eddy viscosity model and a filter width that depends on the local mesh size. The discrete model is based on the implicit Euler scheme and a conforming finite element method for the time and space discretizations, respectively. We establish a reliable and efficient a posteriori error estimation between the numerical LES solution and the exact solution of the original Navier–Stokes system, which involves three types of error indicators respectively related to the filter and to the discretizations in time and space. Numerical results show the effectiveness of adaptive simulations

DEPENDENCE OF TEMPERATURE DETERMINATION ACCURACY ON THE BANDWIDTH OF THE INTERFERENCE FILTER IN OPTICAL PYROMETRY SYSTEMS

Some general issues common to all reactors for gas-phase epitaxy of A3B5 metal-organic compounds include large temperature gradients in the reactor, leading to the formation of convection loops, high gas flow rates that can result in turbulence instead of the expected laminar flow, and the need for good uniformity and precision in temperature control of the semiconductor substrate surface. Since the deposition process for many complex semiconductor devices (such as infrared laser diodes or heterojunction bipolar transistors) is highly temperature-dependent, a system for precise temperature control of the process at the substrate surface directly in the deposition zone is necessary to obtain high-quality heterostructures with reproducible parameters.
 This article describes the features of the epitaxial growth process in MOCVD reactors, as well as highlights the tasks facing the temperature control system. The feasibility of using a radiation compensation system is explored in comparison to measurement methods using a thermocouple and classical pyrometry. Transformation of the input optical signal into a proportional output current in this system is performed by a sensitive photodetector, namely a silicon photodiode. The selection of the wavelength range of the radiation is determined by the optical parameters of the semiconductor substrates and compounds used in the system. The measurement spectrum is determined by the parameters of the interference filter in the system.
 The research was conducted to determine the effect of the interference filter width on the accuracy of determining the real temperature of a pyrometric control system with radiation compensation. The dependence of the interference filter characteristic width and shape on the correction value of the temperature of pyrometric systems after calibration is analyzed. Based on the results obtained, recommendations are made for selecting the optimal working wavelength and interference filter bandwidth of the optical system as a compromise between the determination error and the magnitude of the optical signal. The results can be used for the design of modern precision pyrometric systems.

Eddy viscosity enhanced temporal direct deconvolution models for temporal large-eddy simulation of turbulence

A dynamic eddy viscosity (DEV) model and a constant eddy viscosity (CEV) model are proposed for stabilizing the temporal direct deconvolution model (TDDM) in temporal large-eddy simulation of turbulence. Compared to the original unresolved subfilter-scale model used in TDDM, the new eddy viscosity models reduce the number of empirical coefficients and make TDDM more convenient to be applied in practice. The DEV model does not have any empirical coefficients, and the CEV model has only one constant model coefficient that is independent of the filter width and insensitive to the grid resolution. To solve the stability issue of TDDM, an algorithm called the variable filter-width method (VFM) is proposed. In VFM, the filter width is initialized by a small value or 0 and then grows linearly in a small number of time steps until it reaches the target filter width. The three dimensional homogeneous isotropic turbulence is applied to investigate the performance of the proposed models. In the a posteriori testing at different grid resolutions, eddy viscosity enhanced temporal direct deconvolution models show a good accuracy in predicting various statistics and instantaneous spatial structures of turbulence, and they perform better than the original model, especially in the prediction of subfilter-scale (SFS) stress and SFS energy flux. Moreover, the energy spectrum and other flow statistics predicted by the CEV model with a fixed model coefficient 0.03 are in a good agreement with the filtered DNS.

The Solar Upper Transition Region Imager (SUTRI) Onboard the SATech-01 Satellite

The Solar Upper Transition Region Imager (SUTRI) onboard the Space Advanced Technology demonstration satellite (SATech-01), which was launched to a Sun-synchronous orbit at a height of ∼500 km in 2022 July, aims to test the on-orbit performance of our newly developed Sc/Si multi-layer reflecting mirror and the 2k×2k EUV CMOS imaging camera and to take full-disk solar images at the Ne vii 46.5 nm spectral line with a filter width of ∼3 nm. SUTRI employs a Ritchey–Chrétien optical system with an aperture of 18 cm. The on-orbit observations show that SUTRI images have a field of view of ∼ 41.′6 × 41.′6 and a moderate spatial resolution of ∼8″ without an image stabilization system. The normal cadence of SUTRI images is 30 s and the solar observation time is about 16 hr each day because the earth eclipse time accounts for about 1/3 of SATech-01's orbit period. Approximately 15 GB data is acquired each day and made available online after processing. SUTRI images are valuable as the Ne vii 46.5 nm line is formed at a temperature regime of ∼0.5 MK in the solar atmosphere, which has rarely been sampled by existing solar imagers. SUTRI observations will establish connections between structures in the lower solar atmosphere and corona, and advance our understanding of various types of solar activity such as flares, filament eruptions, coronal jets and coronal mass ejections.

The volume-filtering immersed boundary method

We present a novel framework to deal with static and moving immersed boundaries (IB) based on volume-filtering. In this strategy, called Volume-Filtering Immersed Boundary (VFIB) method, transport equations are derived by filtering the Navier-Stokes equations and accounting for stresses at the solid-fluid interface. The result is that boundary conditions that normally apply on the solid-fluid interface are transformed into bodyforces that apply on the right-hand side of the filtered transport equations. In this method, the filter width acts as a parameter that controls the level of resolution. The IB is considered well-resolved if the filter width is much smaller than the characteristic corrugation scale of the interface. There are several innovations in this IB method. First, it sheds light on the role of the internal flow which arises when the transport equations are solved inside the IB. We show that, it is essential to separate stresses due to the external and internal fluids in order to get accurate forces, and provide a method to do so. Second, we show that the volumes associated with Lagrangian forcing points on the boundary depend on the local topology of the surface. We provide a straightforward way to compute these volumes using a triangle tessellation of the interface and the surface density function. Third, we provide an efficient procedure to compute the solid volume fraction, thus, enabling tagging interior/exterior cells. This volume fraction is also involved in the procedure to separate stresses due to the external fluid from the total stresses. Fourth, we show a path forward to extend the VFIB method to Large Eddy Simulations involving IBs. Lastly, we apply the VFIB in several numerical tests involving two- and three-dimensional static and moving IBs. We show greatly improved results compared to prior IB methods. Further, we test several filter kernels and show that, for well-resolved IBs, the choice of the kernel plays little role.

The metrological structural resolution for dimensional x-ray CT: analysis of the comparability between the CEB and the PBS method

The metrological structural resolution (MSR) describes the size of the smallest surface feature that can be measured dimensionally with a given accuracy. Several methods to determine the MSR for dimensional x-ray CT (dXCT) have been proposed in the past, two of which are compared and related in this publication, i.e. the curved-edge based (CEB) and the profile-based spectral (PBS) method. Both methods consider the surface structure as being described by a single surface on the relevant local scale and are also suitable for the application to optical or tactile coordinate measurement systems (CMSs). The CEB method evaluates the radii of circular shapes to determine the width of the Gaussian filter that describes the filtering of the surface by the CMS. The PBS method evaluates the instrument transfer function (ITF) determined by means of the measurement of a surface profile with a broad-band spatial frequency spectrum with a finite cut-off frequency. The PBS method yields the threshold wavelength for which the amplitude of the ITF drops below a certain level. While the resulting quantities of the two methods are very different they evaluate the same characteristic of the CMS. In this publication an analytical relation between those results is derived and shown to exist which is used to define the MSR. Simulated CT scans as well as dXCT measurements are performed to verify this relation. The results for the MSR obtained from both methods are consistent and deviations to the expectations based on theory are within a reasonable range.

L-kurtosis-based optimal wavelet filtering and its application to fault diagnosis of rolling element bearings

Repetitive transients are a key symptom for the occurrence of incipient fault of rolling element bearings. Therefore, an optimal wavelet filtering method is developed by maximizing the L-kurtosis through the genetic algorithm to extract the weak repetitive transients buried in the heavy noise and disturbed by the outliers. First, the capability of L-kurtosis for characterizing the impulsiveness and cyclostationary of repetitive transients is numerically studied at different degrees of noise. Then, the center frequency and band width of morlet wave filter are adaptively determined by the genetic algorithm and the maximization of L-kurtosis. Finally, both simulation and experiments are performed to validate the efficacy of the proposed method. Results show that the proposed method is more powerful and reliable than the other commonly used indexes-based optimal wavelet filtering methods.

Dissipation and amplification management in an electrical model of microtubules: Hybrid behavior network

The control of dissipation and amplification of solitary waves in an electrical model of a microtubule is demonstrated. This model consists of a shunt nonlinear resistance–capacitance (J(V)–C(V)) circuit and a series resistance–inductance (R–L) circuit. Through linear dispersion analysis, two features of the network are found, that is, low bandpass and bandpass filter characteristics. The effects of the conductance’s parameter λ on the linear dispersion curve are also analyzed. It appears that an increase of λ induces a decrease (an increase) of the width of the bandpass filter for positive (negative) values of λ. By applying the reductive perturbation method, we derive the equation governing the dynamics of the modulated waves in the system. This equation is the well-known nonlinear Schrödinger equation extended by a linear term proportional to a hybrid parameter σ, i.e., a dissipation or amplification coefficient. Based on this parameter, we successfully demonstrate the hybrid behavior (dissipation and amplification) of the system. The exact and approximate solitary wave solutions of the obtained equation are derived, and the effects of the coefficient σ on the characteristic parameters of these waves are investigated. Using the analytical solutions found, we show numerically that the waves that are propagated throughout the system can be dissipated, amplified, or remain stable depending on the network parameters. These results are not only in agreement with the analytical predictions, but also with the existing experimental results in the literature.

Gravity Anomalies and Implications for Shallow Mantle Processes of the Western Cocos‐Nazca Spreading Center

AbstractThis study analyzes up‐to‐date gravity data in the Galapagos triple junction region to understand crustal structure and melt distribution beneath the propagating Cocos‐Nazca spreading center (CNSC). Application of a standard thermal model to the mantle Bouguer gravity anomaly (MBA) does not appear to result in a realistic crustal thickness in this region. The cross‐CNSC MBA profiles flatten and axial values increase from east toward the western end of the CNSC. A simple smoothing filter applied to the standard thermal model with different filter widths can explain the progressive flattening of the MBA and is interpreted as different distribution widths (concentrations) of partial melt in the mantle. The east‐west residual MBA gradient along the CNSC is similar to the east flank of the East Pacific Rise (EPR), suggesting that the along‐CNSC gradient could partly reflect the shallow mantle properties associated with the EPR.

Optical modeling and analysis of pixel organic light-emitting diode using a mixed-level algorithm considering light leakage effects

Fast optical simulation is the basis of optical optimization for cross-scale 3D pixel organic light-emitting diodes (OLEDs), and it is among the most critical challenges in the industry. We constructed an optical simulation model by using a mixed-level algorithm to fasten the analysis of pixel OLEDs, considering the light leakage effect, which was described in terms of the ratio of light extraction (RLE). Model correctness and efficiency were verified using a commercial software. Pixel OLEDs with different sizes were used, and the influence of the height and angle of pixel definition layers (PDLs) and width of color filters on the performance was analyzed. These geometric features change the path of light propagation, which changes performance. Device B was more sensitive to the PDL angle, whereas devices A and C were more sensitive to the PDL height. Device B had the least RLE (RLEB>RLEA>RLEC). An optimal structure can reduce light leakage and improve the performance of pixel OLEDs. The proposed method provides guidelines for designing optimal pixel OLEDs and for the high-throughput simulation of pixel OLEDs.

Detecting agglomeration patterns on solid propellant surface via a new curvature-based multiscale method

In this article, a new algorithm detecting particles and aggregates on the surface of a burning solid propellant containing inert particles is presented. Shadowgraphy images are captured using a set-up at ONERA. Four propellants containing particles of different sizes are used. Detecting the surface being well studied, the method rather focuses on detecting protrusions of a continuous 1D curve. Curvature is the basic tool used, it can be calculated using Gaussian filters of different widths. A normalization is proposed, ensuring curvatures of different filter widths to be compared, meaning the algorithm can detect protrusions of different sizes. The result is the Extreme of the Normalized Curvature (ENC), detecting concave parts of a curve, independently of the size of the protrusion. While the ENC detects the main concavities, the limits of the targeted patterns (convexities) are found by a multi-scale approach following the detection.Evaluating the performances of the algorithm is possible thanks to the annotations of some images. Both detection performances and limit research performances are investigated. The influence of the initial particle size on the performances is studied.

Generating Melodies from Melodic Outlines Towards an Improvisation Support System for Non-musicians

One promising approach for supporting non-musicians’ music improvisation is to enable them to input the coarse shape of melodies they want to play. Therefore, drawing a melodic outline is adopted as an input method of existing improvisation support systems. However, generating melodies from melodic outlines with deep neural networks has not been fully exploited. In this paper, we report an attempt to generate melodies from melodic outlines using a convolutional neural network. Given a melodic outline, it is reduced with two convolution layers and then coverted to a sequence of notes with two deconvolution layers. Objective and subjective evaluations imply that our model with sufficient filter width and channels could generate melodies of moderate, close quality to human melodies.