Layer Model Research Articles

Energy-efficient neural network training through runtime layer freezing, model quantization, and early stopping

Background:In the last years, neural networks have been massively adopted by industry and research in a wide variety of contexts. Neural network milestones are generally reached by scaling up computation, completely disregarding the carbon footprint required for the associated computations. This trend has become unsustainable given the ever-growing use of deep learning, and could cause irreversible damage to the environment of our planet if it is not addressed soon. Objective:In this study, we aim to analyze not only the effects of different energy saving methods for neural networks but also the effects of the moment of intervention, and what makes certain moments optimal. Method:We developed a novel dataset by training convolutional neural networks in 12 different computer vision datasets and applying runtime decisions regarding layer freezing, model quantization and early stopping at different epochs in each run. We then fit an auto-regressive prediction model on the data collected capable to predict the accuracy and energy consumption achieved on future epochs for different methods. The predictions on accuracy and energy are used to estimate the optimal training path. Results:Following the predictions of the model can save 56.5% of energy consumed while also increasing validation accuracy by 2.38% by avoiding overfitting.The prediction model developed can predict the validation accuracy with a 8.4% of error, the energy consumed with a 14.3% of error and the trade-off between both with a 8.9% of error. Conclusions:This prediction model could potentially be used by the training algorithm to decide which methods apply to the model and at what moment in order to maximize the accuracy-energy trade-off.

Modeling of Electric Double Layer to Analyze Reaction Kinetics

Electric double layers (EDLs) play a fundamental role in various electrochemical processes such as colloid dispersion, surface charging, and charge-transfer reactions. Increasingly, the role of EDLs on reaction kinetics is being studied (1). Despite the extensive history of EDL modeling, there remain challenges in accurate prediction of its structure. The characteristic length of EDL (nanometer scale) exceeds the grasp of typical ab-initio molecular-dynamic simulations in order to capture a whole picture of the statistical equilibrium. While continuum models offer a means to estimate the thermodynamic equilibrium of EDLs with substantially lower computational cost than molecular dynamics, state–of-the-art continuum models require parameter fitting(2) due to the lack of appropriate expressions for microscopic interactions.Herein, we propose a predictive multiscale continuum model of EDL that eliminates the empirical fitting and relies more on physical relations. This model computes the distribution of the electrostatic potential, electron density, and species’ concentrations by minimizing the total grand potential of the system. The grand potential is calculated from entropic energy, electrostatic energy, electron energy, solute-solute interactions, and electrode-solute interactions. These energy expressions include not only the terms from previous work(2), but also the microscopic interactions that are newly introduced in this work: polarization of hydration shells, electrostatic interaction in parallel plane toward the electrode, and ion-size-dependent entropy. The parameters that identify the electrode and electrolyte materials are obtained from independent experiments such as vacuum work function, bulk modulus, X-ray diffraction, and concentration-dependent dielectric constant. The developed model reproduces the trends in the experimental differential capacitance with multiple electrode and electrolyte materials (Ag (111)-NaF, Ag (111)-NaClO4, and Hg-NaF), which verifies the accuracy and predictiveness of the model in a system without significant specific electrode-electrolyte interactions. The difference in the capacitance between mercury and silver electrode was attributed to the difference in the electron stability in the metal that is parameterized from the material’s bulk modulus. Finally, extension of the model to systems with specific electrode-electrolyte interactions (e.g., Pt (111)-NaClO4) and the analysis of reaction kinetics will be introduced. Overall, the model framework and findings will provide insights into the EDL structures and guidance on how to tailor electrode-electrolyte interface for improved reaction kinetics. Acknowledgements: This work was partially supported by the Hydrogen and Fuel Cell Technologies Office (HFTO), Office of Energy Efficiency and Renewable Energy, U.S. Department of Energy (DOE) through the Million Mile Fuel Cell Truck (M2FCT) consortia, technology managers G. Kleen and D. Papageorgopoulos, under contract number DE-AC02- 05CH11231, as well as by a CRADA with Toyota Central R&D Labs., Inc. and by the Center for Ionomer-based Water Electrolysis (CIWE), a DOE Energy Earthshot Research Center. Reference: S. J. Shin, D. H. Kim, G. Bae, S. Ringe, H. Choi, H. K. Lim, C. H. Choi and H. Kim, Nat Commun, 13, 174 (2022).J. Huang, J Chem Theory Comput, 19, 1003 (2023).

Mass Transport Regimes in a CO2R Gas Diffusion Electrode Pore with Coupled Microkinetics Model

Gas Diffusion Electrodes (GDEs) have arisen as a popular CO2 electrolyser configuration due to the improved mass transfer properties and higher current densities possible in comparison to a liquid-fed system. In this work, we aim to understand the mass transport regimes in a single liquid-filled pore of a GDE, i.e., an idealized model of the porous catalyst layer. In previous work, Butler-Volmer (BV) expressions have been used to model the CO evolution reaction and the Hydrogen Evolution Reaction (HER). In this study, we extend that in a few ways: (1) a MicroKinetic Model (MKM) of Au at the surface of the electrode is used instead of the BV expressions, (2) volume/steric effects are incorporated through the Generalized Modified Poisson-Nerst-Planck (GMPNP) equation, and (3) more reactions are considered in the electrolyte. By analyzing pores of varying sizes, following the size distribution measured by Focused Ion Beam Scanning Electron Microscopy (FIB/SEM) in a GDE from the literature, we assess conditions where concentration profiles are no longer effectively one-dimensional in the pore axis. We discuss how these three-dimensional mass transfer effects impact CO2 conversion efficiency. In such regimes, we describe why special treatment is required when calling BV expressions in macroscale pore-averaged models.This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. LLNL release number: LLNL-ABS-858384.

(Invited) Modeling of the Electric Double Layer (EDL) Structures of Localized High-Concentration Electrolytes (LHCE)

Various electrical double layer (EDL) models have been proposed but mainly for homogenous solutions. Recently, the so-called Localized high-concentration electrolytes (LHCE) have shown many benefits to high-capacity electrodes. Recently, we have proposed micelle-like structures to enable a complete understanding of microstructures in LHCE, which contains a salt, solvent, and non-ion-solvating diluent, analogical to the roles of oil, surfactant, and water as in micelles. [1] As the solid electrolyte interphase (SEI) formation is sensitive to the EDL structure, [2] it is critical to understand the EDL structures for LHCE with heterogeneous structures.Molecular dynamics simulations were performed in several electrolytes with heterogeneous structures on negatively charged surfaces. It was found the Stern model with an absorbed cation layer adsorption layer and a more diffused layer with cation, anion, solvent, and diluents, should be used to describe the EDL of this type of electrolyte.For the system consisting of LiFSI salt, DME solvent, and TFEO diluent at a concentration of LiFSI-1.2DME-2TFEO, the heterogeneous micelle-like structures in LHCE are maintained in the EDL, with the Li+-rich salt-solvent clusters and the Li+-poor diluent regions. The diluent region contains Li+ near the negatively charged surface but does not have any ions in the bulk of LHCE. This is because TFEO cannot form a complete solvation shell in the bulk electrolyte due to the steric effect, however, only a partial solvation shell is needed in the Stern layer in the EDL. The appearance of Li+ in the diluent is also necessary for the EDL, as TFEO alone cannot screen the charge.The chemical and concentration design space for the three components is very large. To obtain a more general picture, we also explored different salt-solvent ratios and different diluent-salt interactions. As the salt solvent ratio increases from low (LCE), to medium (MCE) to high concentration electrolytes (HCE), where all solvents are coordinated with ions, the salt-solvent structures evolve from fully solvated ions or ion pairs in LCE, 3D branched networks in MCE, to fully connected salt aggregates in HCE. The salt-diluent interactions and diluent concentration further modify the micelle-like structures. The EDL for both LHCE and LMCE are heterogenous. [1] Nature Materials 22, 1531–1539 (2023)[2] J. Am. Chem. Soc. 145, 4, 2473–2484 (2023)

XPS depth profiling of nano-layers by a novel trial-and-error evaluation procedure

In spite of its superior chemical sensitivity, XPS depth profiling is rarely used because of the alteration introduced by the sputter removal process and the resulting inhomogeneous in-depth concentration distribution. Moreover, the application of XPS becomes increasingly challenging in the case of the analysis of thin layers, if the thickness is in the range of 2–3 inelastic mean free paths (IMFP) of the photoelectrons. In this paper we will show that even in these unfavorable cases the XPS depth profiling is applicable. Herein the XPS depth profiling of a model system tungsten-carbide-rich nano-layer of high hardness and corrosion resistance is presented. We will show that the problems arising because of the relatively high IMFP can be corrected by introducing a layer model for the calculation of the observed XPS intensities, while the alteration, e.g. ion mixing, compound formation and similar artefact, introduced by the sputter removal process can be handled by TRIDYN simulation. The method presented here overcomes the limitation of XPS depth profiling.

Extending an Empirical Degradation Model for Li-Ion Batteries into Second-Life Applications

Lithium-ion batteries have been used increasingly in recent years for a wide range of applications, e.g., electric vehicles and large-scale power grids. Lithium-ion batteries suffer from degradation caused by their operation and their exposure to environmental conditions and thus have a limited lifetime within their first application. These batteries might be available for re-use in a different application, commonly under more moderate usage conditions. The influence of such a change in operating conditions on battery degradation within a second life application is not very well understood. Here we present the extension of an empirical degradation model that formulates the degradation process as a function of battery operations for both, first- and second life applications.Our extended semi-empirical model for battery degradation is based on work of Xu et al. [1] and combines the degradation mechanism due to the growth of the SEI layer and the degradation depth obtained in dedicated ageing tests. Within this work we focus on a commercial 64 Ah NMC-graphite pouch cell. The model parametrization is based on a large dataset containing both calendar and cycle life data of more than 100 cells, covering a large window of temperatures (5-45 °C), C-rates (0.2-1.5 C), SoC windows, and average SoC.Figure 1(a) shows the variations of state of health (SoH) with the number of the full equivalent cycle (FEC) for four cells cycling at the same constant ageing conditions (0.75C, 25°C, and 0-100% SoC). Figure 1(b) shows the fitting results of the empirical degradation model using the data from the ageing tests. It is assumed that the SEI layer model parameters are constant for this specific cell. The degradation depth increases linearly with cycle number, where the degradation rate varies with the battery operation condition. The degradation of the battery cell depends not only on external stress factors such as temperature, SoC, DoD, C-rate, but also its previous ageing history. The formation of the SEI film contributes non-linearly to the battery degradation, while the contribution of the degradation rate of different stress factors can be accounted for linearly. The degradation rate depends on the ageing modes: calendar or cycle ageing. Calendar ageing reflects the inherent degradation over time and the rate is affected by temperature and SoC. The calendar degradation rate per time unit can be obtained as a function of temperature and SoC [2, 3]. The cycle SoC window can be interpreted using DoD and mean SoC. The cycle degradation rate per FEC can be estimated as a function of temperature, mean SoC, and DoD [3]. Data-driven models and empirical correlations are applied to obtain battery degradation rate functions. The total degradation depth can be estimated by summarizing the degradation caused by the calendar ageing and the degradation resulting from all cycles.To verify the performance of the empirical degradation model, a dynamic ageing test was designed, including both calendar and cycle ageing. The historical data of SoC and temperature are shown in Figures 2 (a) and 2(b), respectively. Through analyzing the time series data of cell SoC, all cycle information can be obtained to calculate the cycle ageing. The calendar ageing was calculated based on time duration for each cycle, and the average temperature and SoC. Using the correlations for the degradation rates, the degradation depth during for each cycle can be evaluated. The accumulated degradation depth can be used to estimate the change of remaining capacity with time. Figure 2(c) shows the comparison of the prediction of SoH with the experimental data.When moving into the 2nd life regime, degradation mechanisms and their interplay become more complex, leading to a larger statistical variation in ageing rates between individual cells (i.e., see time offset in Fig 2c). This can be accounted for by readjustment of the model’s initial 2nd life SoH to the last SoH from the first life application data, as well as including statistical variation into the model’s parameters.

Understanding the impact of Li2CO3 distribution within solid electrolyte interphases on lithium metal via thermal conditioning

The formation of a solid electrolyte interphase (SEI) is crucial for the performance and safety of Li metal batteries (LMBs). The SEI consists of various compounds that are assumed to prevent electrolyte decomposition and dendrite growth. Among these constituents, Li2CO3 has a contentious position; some studies stated that it is a beneficial component, whereas others viewed it as an impurity. This study investigated the influence of strategic heat treatments on the distribution of Li2CO3 formed in the SEI of Li metal. These treatments affect the cycling performance of Li/LiCoO2 cells, as evidenced by the differences in cycle life and capacity. In particular, the study showed that samples heat-treated at higher temperatures generally exhibited enhanced performances, illustrating the impact of heat treatment on SEI quality. The quality of the SEI, which depends on the Li2CO3 formation model (solid electrolyte interface layer or a solid polymer electrolyte layer), varies depending on the applied heat treatment. The samples heat-treated at 120 °C followed the solid polymer electrolyte layer model, resulting in the mosaic-style formation of Li2CO3. This intricate pattern provides superior mechanical stability compared with that of a plain layer-type structure and can be used to develop potential strategies for the enhancement of the cycling stability and overall longevity of LMBs.

Ionic liquid–electrode interface: Classification of ions, saturation of layers, and structure-determined potentials

Progress in electrochemical applications of ionic liquids builds on an understanding of electrical double layer. This computational study focuses on structure-determined quantities – maximum packing density, potentials, and capacitances – evaluated using a one-electrode electrical double layer model. Interfaces of the 40 studied ions are grouped into four distinct classes according to their characteristic packing at the model surface. The simulations suggest that the exact screening by a monolayer of counter-ions (preceding the crowding of ions) is unlikely for ions in known air- and water-stable ionic liquids within their electrochemical stability window. This work discusses how the assessed structure-determined quantities can guide the experimental tuning of (electro/mechano)chemical properties and characterize the structure of ionic liquid–electrode interfaces.

Multispecies macrozoobenthic seasonal bioturbation effect on sediment erodibility

Bioturbation in estuarine environments describes all sediment reworking processes implied in sediment transport. However, modelling at large spatial and temporal scales remains a challenge because of the need to consider the fauna at the community level, and because animal behaviour is highly seasonal with non-linear effects of macrofauna functional trait interactions. Bioturbation processes can be linked to the activity of organisms, based on the principle of energy ecology, linking the metabolic rate to the erodibility of a sediment colonised by benthic fauna. This study investigates this postulate by evaluating the erodibility parameters of a sediment subjected to: i) the bioturbation under seasonal temperature variations; ii) the synergistic bioturbation of different species. The experimental design consisted of: i) three temperature levels (winter, spring and summer), ii) three types of species duos (Cerastoderma edule and Macoma balthica; Scrobicularia plana and Hediste diversicolor; Corophium volutator and Peringia ulvae) at 4 different relative densities. Two successive experiments were carried out on the same individuals: measurement of oxygen consumption of fauna then measurement of the erodibility of the colonised sediment in a flume. The oxygen consumption confirmed that the metabolic rate is a good model of the fauna respiration, regardless of species. The erosion results indicated that the metabolic rate in the case of the fluff layer resuspension is an interesting descriptor for 1) the assessment of the bioturbation under variable temperatures and 2) the integration of the two different bioturbator species that could co-occur in the same habitat. In contrast, the effect of bioturbation on the mass sediment erosion threshold cannot be easily modelled by using the metabolic rate and the classification in functional groups is required. Bioturbation models of the fluff layer using metabolic rate is a promising tool for modelling the effects of faunal communities on sediment transport at the scale of an estuary and over the long term, even projected in the context of global warming.

Numerical simulation of one-dimensional perovskite solar cell model

Perovskite solar cell (PSC) is one of the advanced third-generation solar cells that have rapid efficiencies but instability issues in terms of air, moisture, and UV light sensitivity becomes a barrier to commercialization. The instability issue is due to the charge accumulation at the interface of the PSC reducing its efficiency. This research focuses on the operation of PSCs through a one-dimensional (1D) drift-diffusion model for planar heterojunction PSCs. The model also accounts for the electric potential by Poisson’s equation, the ion generation, and the recombination rate. The method of lines technique is applied to solve the model for the perovskite layer numerically using the finite difference method which is then solved forward in time using the ‘ode15s’ solver in MATLAB. It is highlighted that the comparison with the experimental data from the reference shows good agreement. The effect of parameter thickness variation of the perovskite layer upon the efficiency of PSCs is analyzed. The result shows that the best efficiency obtained is 19.77% obtained at thickness 0.25 𝜇𝑚. The results may prove useful for a guideline of the cell thickness that predicts the perovskites cell performance.

Advanced multi-image segmentation-based machine learning modeling strategy for corrosion prediction and rust layer performance evaluation of weathering steel

Corrosion images are the most accessible type of data to collect. To utilize these data to monitor and evaluate the corrosion of weathering steel (WS) efficiently. This study established a dataset of corrosion images of WS by accelerated corrosion experiments with dry and wet cycles. Image segmentation algorithms and machine learning were used for training, modeling, and data mining of the image data. The results showed that 22 features in the corrosion images were highly correlated with the corrosion thickness loss, with the highest correlation coefficient exceeding 0.7. The rust layer texture and color features showed a synergistic evolution during the corrosion process. On this basis, a classification model of the rust layer was developed to quantify the evolution of the protective properties of the rust layer during the corrosion process. In addition, the established integrated learning model achieves accurate prediction (R2 > 0.94) for WS corrosion thickness loss.

The Behavior of the Tunnel Reinforced with Geogrid in Soft Soil Under the Effect of Axial Load

The soft soil's poor tensile strength requires reinforcing to increase bearing capacity, improve stability, and reduce settlements. This study assessed the efficacy of using geogrid layers to enhance and secure the soil surrounding tunnels. enabling the tunnel to endure pressure, particularly during excavation. The utilization of geogrids in soil reinforcement has experienced a substantial rise as a result of their consistent dimensions and exceptional tensile strength. To quantify the exerted force transferred to the tunnel, during this study utilized various testing tools, including a soil container, a steel loading frame, data loggers, a 0.5-ton load cell, and a miniature pressure cell. The vertical loads are applied by utilizing a hydraulic jack. A series of eleven tests were conducted on the tunnel at two depths of 1.5D and 2.5D, where D is the tunnel's diameter. The different models of geogrid layers showed that using two layers of geogrid at the first dimension, 0.5B and 1B from the base, led to a significant increase in tunnel stability. Two layers of reinforcement were used in both directions, giving the soil a high bearing capacity for the loads applied to the tunnel. This resulted in an improvement, a 1.65 in 1.5D and a 1.82 in 2.5D. The pressure above the pipe decreased by approximately 7.1kPa at the first tunnel depth and about 3.5kpa at the second depth. In conclusion, the study found the geogrid improves the stability of the tunnel by equally distributing loads and minimizing stress concentrations, hence decreasing the chances of collapses or deformations. Doi: 10.28991/CEJ-2024-010-08-04 Full Text: PDF

Investigation of the surface charge behaviour of ettringite: Influence of pH, calcium, and sulphate ions

Ettringite is an important mineral that contributes to the overall performance of cementitious materials. Knowledge of the surface charge behaviour of a solid is necessary for a mechanistic description of surface processes such as adsorption or particle-particle interactions. The objective of this study was to develop a model capable of reproducing ettringite surface charge as a function of calcium, sulphate, and pH.Ettringite was synthesised and characterised using different analytical, microscopic, and spectroscopic techniques with the help of density functional theory. Electrophoretic mobility was measured using laser Doppler electrophoresis in alkaline waters representative of the cementitious environment.The behaviour of the ettringite surface charge was shown to be quite complex as sulphate and calcium acted in a competitive manner on the overall charge. The ζ-potential increases when the calcium content increases, whereas it decreases when sulphate increases. This is due to the possible adsorption of these ions at the surface, and the extent of the effect depends on the relative concentrations of Ca and SO42−.An electrostatic double layer model (DLM) was used to calculate the surface potential, considering the adsorption of both calcium and sulphate, as possible ions determining the potential (IDP), and formation of different complexes with ettringite surface functional groups (SOH). The variations of the ζ-potential could be satisfactorily predicted under the different chemical conditions of interest in a cementitious environment.

Equilibrium atmospheric boundary layer model for numerical simulation of urban wind environment

The numerical simulation of urban wind environments faces difficulties in capturing the turbulent characteristics due to the large computational domain. Traditional Reynolds-averaged methods (RANS) can effectively capture the average wind characteristics of urban areas. However, due to the significant dissipation and attenuation of turbulent energy in the downstream direction, this method fails to provide accurate turbulent characteristics after time-averaging processing. Therefore, in order to obtain a higher-precision turbulent wind field distribution within urban areas, this paper proposed a new numerical method named an equilibrium atmospheric boundary layer model (EABL) by modifying the control equation of the shear stress transport k–ω model. During the process, the equilibrium atmospheric boundary layer was achieved successfully, and the attenuation problem of the turbulent kinetic energy and dissipation rate during the computational fluid dynamics numerical simulation was resolved. Simultaneously, a wind tunnel experiment and six turbulence models [standard k–ε, realizable k–ε, renormalization group k–ε, large eddy simulation—narrowband synthesis random flow generator (LES-NSRFG) and LES vortex method and EABL] were employed to simulate the wind field characteristics in an actual residential area. The simulation results demonstrate that, relative to traditional RANS models, the EABL model enhances the simulation accuracy of turbulence characteristics by over two times. Furthermore, compared to LES models, the EABL model can reduce computational time by threefold.

Turbulent Ice–Ocean Boundary Layers in the Well-Mixed Regime: Insights from Direct Numerical Simulations

Abstract The meltwater mixing line (MML) model provides a theoretical prediction of near-ice water mass properties that is useful to compare with observations. If oceanographic measurements reported in a temperature–salinity diagram overlap with the MML prediction, then it is usually concluded that the local dynamics are dominated by the turbulent mixing of an ambient water mass with near-ice meltwater. While the MML model is consistent with numerous observations, it is built on an assumption that is difficult to test with field measurements, especially near the ice boundary, namely, that the effective (turbulent and molecular) salt and temperature diffusivities are equal. In this paper, this assumption is tested via direct numerical simulations of a canonical model for externally forced ice–ocean boundary layers in a uniform ambient. We focus on the well-mixed regime by considering an ambient temperature close to freezing and run the simulations until a statistical steady state is reached. The results validate the assumption of equal effective diffusivities across most of the boundary layer. Importantly, the validity of the MML model implies a linear correlation between the mean salinity and temperature profiles normal to the interface that can be leveraged to construct a reduced ice–ocean boundary layer model based on a single scalar variable called thermal driving. We demonstrate that the bulk dynamics predicted by the reduced thermal driving model are in good agreement with the bulk dynamics predicted by the full temperature–salinity model. Then, we show how the results from the thermal driving model can be used to estimate the interfacial heat and salt fluxes and the melt rate. Significance Statement We investigate the turbulent dynamics and thermodynamical properties of water masses below ice shelves using new data from high-resolution simulations. This is important because there are currently too few observations of ice–ocean boundary layer dynamics to construct reliable models of ice-shelf melting as a function of ocean conditions. Our results demonstrate that the turbulent diffusivities of salt and temperature are approximately equal in the well-mixed regime. This implies a linear correlation between the mean temperature and salinity profiles, consistent with the meltwater mixing line prediction that is often used to interpret polar observations. We take advantage of this correlation to propose a reduced model of ice–ocean boundary layers that can predict ice-shelf melt rates at relatively low computational cost.

Plasticization-assisted slow crack growth modeling of high-density polyethylene

The existing crack layer model can theoretically predict the slow crack growth behavior of high-density polyethylene. However, it can only be applied to oxidative environments causing chemical degradation. Most oil and gas field components, such as pressurized pipes, are subjected to hydrocarbon exposures, leading to a plasticized material response, i.e., shear yielding instead of crazing. Therefore, the effect of such sorptive media diffusion on the slow crack growth behavior and lifespans tf should be understood. In this work and for the first time, a novel crack layer model is developed that can simulate the diffusion-assisted slow crack growth behavior of plasticized high-density polyethylene. The proposed model was validated by comparing its prediction with experimental results and by conducting a sensitivity study on several input parameters. Using the proposed model, the reported plasticization results were reconstructed successfully including the SCG rate with R2 = 0.95. This study expands the applicability of the crack layer model for the reliability assessment of polyethylene pipes under various environmental conditions, including plasticizers.

Energy equality of weak solutions of the 2D reduced‐gravity two‐and‐a‐half layer system allowing vacuum

The energy conservation problem for the reduced‐gravity two‐and‐a‐half layer model remains a challenging open problem, with several recent contributions addressing various aspects of it. In this paper, we establish sufficient conditions on the regularity of weak solutions that guarantee energy conservation even in the presence of a vacuum. Our theorem corresponds to an improvement of some recent results on this problem and contains some well‐known results as a particular case.

Prediction of IGBT Gate Oxide Layer's Performance Degradation Based on MultiScaleFormer Network.

Insulated gate bipolar transistors (IGBTs) are widely used in power electronic devices, and their health prediction problems have attracted much attention in the field of power electronic equipment health management. The performance degradation of IGBT gate oxide is one of the most important failure modes. In order to analyze this failure mechanism and the ease of implementation of a monitoring circuit, the gate leakage current of IGBTs was selected as the fault precursor parameter for the degradation of their gate oxide performance, and feature selection and fusion were carried out by using time domain characteristic analysis, grayscale correlation, Mahalanobis distance, Kalman filter, and other methods. Thus, a health indicator was obtained to characterize the degradation of IGBT performance, which was used to indicate the degree of aging of the IGBT gate oxide layer. In this paper, we propose an improved degradation prediction model called MultiScaleFormer, inspired by advanced design ideas of the iTransformer network architecture, combined with the health parameters of IGBTs to construct a degradation prediction model for the IGBT gate oxide layer. MultiScaleFormer showed the highest fitting accuracy compared with the Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Support Vector Regression (SVR), Gaussian Process Regression (GPR), CNN-LSTM, and Transformer models in our experiment. The mean absolute error (MAE) of the MultiScaleFormer prediction was as low as 0.0087. Extraction of the health indicator and the construction and verification of the degradation prediction model were carried out on the dataset released by the NASA-Ames Laboratory. These results demonstrate the feasibility of the gate leakage current as a fault precursor parameter for IGBT gate oxide failure, and the feasibility and accuracy of the MultiScaleFormer prediction model for IGBT performance degradation.

Evaluation of the eddy diffusivity in a pollutant dispersion model in the planetary boundary layer

Evaluation of the eddy diffusivity in a pollutant dispersion model in the planetary boundary layer

Deciphering BSA adsorption onto COL-BC: Interpretations from statistical physics modeling

In this current study, a single-energy double layer model was employed to inspect the adhesion mechanism of bovine serum albumin (BSA) on porous microsphere of cellulose (BC) collagen (COL) bacterium (COL-BC). Six physicochemical variables defining the retention mechanism, :the number of anchored BSA protein per site (n), the amount of (COL-BC) available pores (NM), the capacity of adsorbed quantity (Qa), the specific surface area (As), the concentrations at half saturation (C1/2) and the molar binding energy (ΔE1) are determined. Calculations revealed that n ≥ 1 indicating that an average of one BSA per site is linked to one (COL-BC) receptor site with forming two layers. The BSA/(COL-BC) docking system exhibits a weak binding affinity as evidenced by the computed molar adsorption energy (ΔE1) which is below 40 kJ/mol. This energetic observation aligns with the characteristics of a physical and exothermic binding mechanism. Analysis of the (COL-BC) surface using the Kelvin equation revealed a pore size distribution (PSDs) extending across 0.05 μm–2 μm, providing insights into the stereochemistry of its receptor sites. Discussion of the BSA/COL-BC adhesion reaction through Polanyi equation unveils an adsorption energy distribution (AED) localized between 0 and 15 kJ/mol, consistent with a weak interaction profile characteristic of a physisorption mechanism. Finally, three thermodynamic functions were estimated in order to macroscopically study the BSA/(COL-BC) adsorption systems.This work is involved to provide more information about the adsorption of BSA on the COL-BC microsphere and investigate a theorical approach for the application of the controlled release technology.