Dynamics Simulation Data Research Articles

Dual-agent intelligent fire detection method for large commercial spaces based on numerical databases and artificial intelligence

Fires in commercial buildings can lead to significant casualties and property damage. Thus, rapidly and accurately identifying the source of a fire and its intensity is crucial for guiding rescue operations. This study aims to explore the use of distributed fiber optic temperature sensing systems, combined with deep learning algorithms, to predict key information about fires in commercial buildings, including the location of the fire source, the intensity of fire development, and the critical planar distribution of carbon monoxide. Based on the Fire Dynamic Simulation (FDS) data, a dual-agent model was developed, and with a sensor sampling interval of 3 s, 30 s of historical data achieved prediction accuracies of 94.23 % for fire intensity and 99.28 % for fire location. Upon reducing the intervals for fire location, the accuracy reached 93.56 %. Additionally, the Random Forest algorithm for mapping carbon monoxide distribution demonstrated an overall Mean Absolute Percentage Error (MAPE) of 5.87 %. Within the 0–150 ppm range, the MAPE was 2.46 %, far below the error levels of existing CO sensors. The database configurations and the sensors' spatiotemporal arrangement were also optimized to rapid and reliable fire prediction. This research demonstrates the feasibility of using artificial intelligence for critical fire scene information detection.

Comparative analysis of online voltage stability indices based on synchronized PMU measurements

The need for reliable real-time information on voltage stability margins of electrical power systems is an increasingly relevant concern within the current trend of electrification and deployment of power electronics-based devices. This paper conducts the assessment and comparison of four Voltage Stability Indices (VSIs) proposed for this application and based exclusively on synchronized phasor measurements. The robustness and accuracy of each method in identifying the point of maximum power transfer are evaluated as the correlation between load characteristics and consistent estimation of voltage stability margins is explored. In addition, the likelihood inherent to each VSI formulation of triggering false alarms under certain system dynamics is addressed in detail. The comparative analyses are derived from dynamic simulation data of a 3-bus test system, the IEEE 9-bus network and the IEEE 39-bus network, all modelled in the open-source Python-based power system simulator DynPSSimPy. Case studies cover placement of monitoring device, different load types, line disconnection events and presence of measurement noise. The results presented serve as a reference point for the development and/or enhancement of VSIs suitable for real-time applications, highlighting their most significant advantages and drawbacks and providing insights on potential trade-offs that need to be considered when employing such approaches within control center settings.

Simulation data-driven adaptive frequency filtering focal network for rolling bearing fault diagnosis

The scarcity of well-labeled samples severely limits the application of deep learning-based fault diagnosis methods. To address this issue, this paper proposes a novel domain-adaptive intelligent diagnostic method, termed a simulation data-driven adaptive frequency filtering focal network, which transfers knowledge from dynamic simulation data to unlabeled measured data. First, a dynamic simulation model is established to simulate the coupled behavior of rolling bearings and generate substantial labeled simulation data. Subsequently, an adaptive frequency filter is introduced to suppress random interference components in the frequency domain, complementing conventional time-domain feature extraction methods, thereby reducing the distribution discrepancy between simulation and measured data. Finally, a domain-aware weighted focusing strategy is proposed to adjust sample weights based on predicted domain labels, helping the model focus on difficult samples with significant distribution discrepancies and reduce misclassification. Experimental results demonstrate that the proposed method effectively integrates simulation and measured data, achieving high-accuracy fault diagnosis of rolling bearings in unlabeled measured data. The proposed approach offers a promising fault diagnosis solution in scenarios where obtaining labeled samples is challenging or impractical.

Assessment of ethanolic extract of Dittrichia viscosa from Kardoussa Douar region of Taza in Morocco as antioxidant and green inhibitor for carbon steel corrosion in acidic medium

This study targeted evaluating the anticorrosive and antioxidative characteristics of Dittrichia viscosa (ZS-OH) extract. ZS-OH was first extracted by deploying a soxhlet and characterized by chromatographic analysis (GC/MS). In addition, a quantitative assessment of the total phenolic (TPC) and flavonoids (FC) content of the ZS-OH extract, as well as phytochemical analyses, were carried out. Antioxidant testing was conducted using the bioassay: 2,2-diphenyl-1-picrylhydrazyl (DPPH). The ability of ZS-OH extract to stymie the corrosion of carbon steel in 1.0 M HCl was considered through potentiodynamic polarization and electrochemical impedance spectroscopy (EIS). ZS-OH extract yield was 4.73 %, and GC-MS revealed 32 distinct chemicals. Neryl isovalerate (29.8 %), 1,10-di-epi-Cubenol (25.4 %), and 2,5-dimethoxy-p-cymene (11.6 %) made up the majority of the ZS-OH extract. The electrochemical results reveal that the ZS-OH extract acts as a mixed-type with an inhibition efficiency of up to 96.8 % at 500 ppm under 303 K. Surface analyses using SEM/EDS confirmed the formation of the protective layer of ZS-OH extract molecules on the C-S surface. The three major components are fully optimized in the DFT through the B3LYP functional and 6–311++G(d,p) basis sets. Molecular dynamics simulation data show the molecules of ZS-OH extract oriented beneficially on metal surfaces, providing a high surface coverage, and thus, a high inhibition performance.

Monitoring and control of air filtration systems: Digital twin based on 1D computational fluid dynamics simulation and experimental data

This study presents the development of a digital model based on one-dimensional computational fluid dynamics for the monitoring and control of filtering systems used for removing flour, dust, and other particulates from the airflow arriving from various sections of industrial production plants.Focusing on a pilot plant equipped with a cyclone bag filter, historical experimental data was integrated with the results of a one-dimensional fluid dynamics simulation model to create a digital twin capable of real-time control and regulation of industrial plants. In particular, measured pressure drop data under different clogging conditions were interpolated to generate the characteristic curves of the filter under various clogging conditions, to be implemented within the digital model of the plant. The generated model, validated through a dedicated experimental campaign, accurately predicted the airflow rate and pressure distribution across the plant. The system’s capability to adapt to changing operational conditions, such as clogging, was demonstrated through simulation, highlighting the model’s utility in maintaining the desired operation levels while minimizing the need for extensive sensor networks.The analyzed case study in the field of air filtration systems aims to fill the gap in the scientific literature related to the application of Digital Twin technology to the control of industrial manufacturing plants. The findings highlight the potential of digital twins in monitoring and control, as well as predictive maintenance, of industrial systems. The findings highlight the potential of Digital Twins in monitoring and control, as well as predictive maintenance, of industrial systems. Future research activities will explore the model’s applicability in failure and anomaly detection, to further enhance predictive maintenance of air filtering systems.

Synthesis and Molecular Modeling Studies and In silico Anthelmintic Activity of N-(benzo[d]thiazol-2-yl)-3- Chloro-6-Nitrobenzo[b]thiophene- 2-Carboxamide Derivatives

Helminthiasis is a noteworthy health problem with increased morbidity. In this manuscript, we developed a series of N-(benzo[d]thiazol-2-yl)-3-chloro-6-nitrobenzo[b]thiophene-2-carboxamide derivatives (3a-f) and evaluated against Pheretima posthuma and Perionyx excavates helminths. The compounds 3a-f were synthesized by reaction of 3-chloro-6-nitrobenzo[b]thiophene-2-carbonyl chloride (1) and appropriate benzo[d]thiazol-2-amines (2a-f) presence of pyridine. A compound 1 was synthesized by reaction of 4-nitro-cinnamic acid reacts with thionyl chloride in the presence of pyridine and chlorobenzene. Results showed that molecule 3c observed with maximum anthelmintic against P. excavates and P. posthuma with mean paralyzing time of 13.15 min and 17.39 min and mean death time of 16.17 min and 20.14 min, respectively. Molecular docking against 3VRA receptor showed that molecule 3c showed maximum dock score of -7.3 kcal/mol and reference standard albendazole with dock score of -5.5 kcal/mol. Molecular dynamic simulation data reflected the stable ligand-receptor interaction. As per Molecular Mechanics Poisson-Boltzmann Surface Area analysis, maximum free binding energy of -31.224 kJ/mol showed by 3c and ARG 76 was the most contributed residue. Density functional theory calculation data confirmed that most of the molecules were soft molecules with electrophilic nature. Molecular electrostatic potential data of 3c showed that terminal nitro group, ketone group, -COO group of ester were responsible for electrophilic attack and benzo[b]thiophne and -NH groups were responsible for nucleophilic attack. The activity profile of the synthesized molecules against helminthes was 3c >3e >3d >3f >3b >3a.. KEYWORDS :Molecular docking, Molecular dynamic simulation, Molecular mechanics poisson-boltzmann surface area, Density functional theory, Anthelmintic activity.

On particle dispersion statistics using unsupervised learning and Gaussian mixture models

Understanding the dispersion of particles in enclosed spaces is crucial for controlling the spread of infectious diseases. This study introduces an innovative approach that combines an unsupervised learning algorithm with a Gaussian mixture model to analyze the behavior of saliva droplets emitted from a coughing individual. The algorithm effectively clusters data, while the Gaussian mixture model captures the distribution of these clusters, revealing underlying sub-populations and variations in particle dispersion. Using computational fluid dynamics simulation data, this integrated method offers a robust, data-driven perspective on particle dynamics, unveiling intricate patterns and probabilistic distributions previously unattainable. The combined approach significantly enhances the accuracy and interpretability of predictions, providing valuable insights for public health strategies to prevent virus transmission in indoor environments. The practical implications of this study are profound, as it demonstrates the potential of advanced unsupervised learning techniques in addressing complex biomedical and engineering challenges and underscores the importance of coupling sophisticated algorithms with statistical models for comprehensive data analysis. The potential impact of these findings on public health strategies is significant, highlighting the relevance of this research to real-world applications.

Towards a momentum potential theory for reacting flows

Mutual coupling of (thermofluiddynamic) modes of perturbations can affect the thermo-acoustic stability of combustors and contribute to combustion noise. For example, vortical or entropic perturbations can be transferred to acoustic perturbations if accelerated by the mean flow. The decomposition of perturbation fields into the respective modes and a linear description of their interactions in terms of fluctuating primitive variables is challenging. In contrast, Doak’s momentum potential theory promises an unambiguous decomposition in terms of momentum fluctuations, which is not limited to the linear regime. Whereas classical momentum potential theory takes into account hydrodynamic, acoustic and entropic modes in unconfined flows, the investigation of noise generation in combustion chambers requires the extension of the momentum potential theory to capture modes linked to the fluctuation of species mass fractions (“species mode”) arising from the change in chemical composition due to the reaction. Furthermore, a rigorous treatment of boundary conditions due to the confinement of the flow inside the combustor is required. The herein presented extension to reactive flows consists of two steps, (i) the formulation of a potential for momentum fluctuations related to species modes and (ii) identification of the total fluctuating enthalpy related to species modes. The extended theory is applied to post-process computational fluid dynamic simulation data of the propagation of entropy and species perturbations through one-dimensional ducts, nozzles and premixed flames. We find that although momentum potential theory offers a complete decomposition of momentum perturbations for reactive flows, the meaningful interpretation of this decomposition is rather challenging, even for non-reactive flows.

Predicting protein conformational motions using energetic frustration analysis and AlphaFold2

Proteins perform their biological functions through motion. Although high throughput prediction of the three-dimensional static structures of proteins has proved feasible using deep-learning-based methods, predicting the conformational motions remains a challenge. Purely data-driven machine learning methods encounter difficulty for addressing such motions because available laboratory data on conformational motions are still limited. In this work, we develop a method for generating protein allosteric motions by integrating physical energy landscape information into deep-learning-based methods. We show that local energetic frustration, which represents a quantification of the local features of the energy landscape governing protein allosteric dynamics, can be utilized to empower AlphaFold2 (AF2) to predict protein conformational motions. Starting from ground state static structures, this integrative method generates alternative structures as well as pathways of protein conformational motions, using a progressive enhancement of the energetic frustration features in the input multiple sequence alignment sequences. For a model protein adenylate kinase, we show that the generated conformational motions are consistent with available experimental and molecular dynamics simulation data. Applying the method to another two proteins KaiB and ribose-binding protein, which involve large-amplitude conformational changes, can also successfully generate the alternative conformations. We also show how to extract overall features of the AF2 energy landscape topography, which has been considered by many to be black box. Incorporating physical knowledge into deep-learning-based structure prediction algorithms provides a useful strategy to address the challenges of dynamic structure prediction of allosteric proteins.

A new framework for predicting tensile stress of natural rubber based on data augmentation and molecular dynamics simulation data

A new framework for predicting tensile stress of natural rubber based on data augmentation and molecular dynamics simulation data

Two Methods Based on Integral Equation Approaches in Analyzing Polyelectrolyte Solutions: Macrophase Separation.

To understand the phase behaviors of polyelectrolyte solutions, we provide two analytical methods to formulate a molecular equation of state for a system of fully charged polyanions (PAs) and polycations (PCs) in a monomeric neutral component, based on integral equation theories. The mixture is treated in a primitive and restricted manner. The first method utilizes Blum's approach to charged hard spheres, incorporating the chain connectivity contribution by charged spheres via Stell's cavity function method. The second method employs Wertheim's multi-density Ornstein-Zernike treatment of charged hard spheres with Baxter's adhesive potential. The pressures derived from these methods are compared to available molecular dynamics simulations data for a solution of PAs and monomeric counterions as a limiting case. Two-phase equilibrium for the system is calculated using both methods to evaluate the relative strength of phase segregation that leads to complex coacervation. Additionally, the scaling exponents for a selected solution near its critical point are examined.

Thermal conductivity of double polymorph Ga2O3 structures

Recently discovered double gamma/beta (γ/β) polymorph Ga2O3 structures constitute a class of novel materials providing an option to modulate functional properties across interfaces without changing the chemical compositions of materials, in contrast to that in conventional heterostructures. In this work, for the first time, we investigate thermal transport in such homo-interface structures as an example of their physical properties. In particular, the cross-plane thermal conductivity (k) was measured by femtosecond laser-based time-domain thermoreflectance with MHz modulation rates, effectively obtaining depth profiles of the thermal conductivity across the γ-/β-Ga2O3 structures. In this way, the thermal conductivity of γ-Ga2O3 ranging from 1.84 to 2.11 W m−1 K−1 was found to be independent of the initial β-substrates orientations, in accordance with the cubic spinel structure of the γ-phase and consistently with the molecular dynamics simulation data. In turn, the thermal conductivity of monoclinic β-Ga2O3 showed a distinct anisotropy, with values ranging from 10 W m−1 K−1 for [−201] to 20 Wm−1 K−1 for [010] orientations. Thus, for double γ-/β-Ga2O3 polymorph structures formed on [010] β-substrates, there is an order of magnitude difference in thermal conductivity across the γ/β interface, which can potentially be exploited in thermal energy conversion applications.

Structural basis for TNIP1 binding to FIP200 during mitophagy

TNIP1 has been increasingly recognized as a security check to finely adjust the rate of mitophagy by disrupting the recycling of the Unc-51-like kinase (ULK) complex during autophagosome formation. Through tank-binding kinase 1 (TBK1)-mediated phosphorylation of the TNIP1 FIR motif, the binding affinity of TNIP1 for FIP200, a component of the ULK complex, is enhanced, allowing TNIP1 to outcompete autophagy receptors. Consequently, FIP200 is released from the autophagosome, facilitating further autophagosome expansion. However, the molecular basis by which FIP200 utilizes its claw domain to distinguish the phosphorylation status of residues in the TNIP1 FIP200 interacting region (FIR) motif for recognition is not well understood. Here, we elucidated multiple crystal structures of the complex formed by the FIP200 claw domain and various phosphorylated TNIP1 FIR peptides. Structural and isothermal titration calorimetry (ITC) analyses identified the crucial residues in the FIP200 claw domain responsible for the specific recognition of phosphorylated TNIP1 FIR peptides. Additionally, utilizing structural comparison and molecular dynamics (MD) simulation data, we demonstrated that the C-terminal tail of TNIP1 peptide affected its binding to the FIP200 claw domain. Moreover, the phosphorylation of TNIP1 Ser123 enabled the peptide to effectively compete with the peptide p-CCPG1 (the FIR motif of the autophagy receptor CCPG1) for binding with the FIP200 claw domain. Overall, our work provides a comprehensive understanding of the specific recognition of phosphorylated TNIP1 by the FIP200 claw domain, marking an initial step toward fully understanding the molecular mechanism underlying the TNIP1-dependent inhibition of mitophagy.

Prediction of Steam Turbine Blade Erosion Using Computational Fluid Dynamics Simulation Data and Hierarchical Machine Learning

Abstract The information of the degree of blade erosion is vital for the efficient operation of steam turbines. However, it is nearly impossible to directly measure the degree of blade erosion during operation. Moreover, collecting sufficient data of eroded cases for predictive analysis is challenging. Therefore, this paper proposes a blade erosion prediction method using numerical simulation and machine learning. Pressure data of several blade erosion cases are collected from the numerical turbine simulation. The machine learning approach involves training on collected simulation data to predict the degree of erosion for the first-stage stator (1S) and the first-stage rotor blade (1R) from internal pressure data. The proposed erosion prediction model employs a two-step hierarchical approach. First, the proposed model predicts the 1S erosion degree using the k-nearest neighbor (k-NN) regression. Second, the proposed model estimates the 1R erosion degree with linear regression models. These models are tailored for each of the 1S erosion degrees, utilizing pressure data processed through fast Fourier transform (FFT). The evaluation shows that the proposed method achieves the prediction of the 1S erosion with a mean absolute error (MAE) of 0.000693 mm and the 1R erosion with an MAE of 0.458 mm. The evaluation results indicate that the proposed method can accurately capture the degree of turbine blade erosion from internal pressure data. As a result, the proposed method suggests that the erosion prediction method can be effectively used to determine the optimal timing for maintenance, repair, and overhaul (MRO).

Sampling Conformational Ensembles of Highly Dynamic Proteins via Generative Deep Learning.

Proteins are inherently dynamic, and their conformational ensembles are functionally important in biology. Large-scale motions may govern protein structure-function relationship, and numerous transient but stable conformations of intrinsically disordered proteins (IDPs) can play a crucial role in biological function. Investigating conformational ensembles to understand regulations and disease-related aggregations of IDPs is challenging both experimentally and computationally. In this paper we first introduced an unsupervised deep learning-based model, termed Internal Coordinate Net (ICoN), which learns the physical principles of conformational changes from molecular dynamics (MD) simulation data. Second, we selected interpolating data points in the learned latent space that rapidly identify novel synthetic conformations with sophisticated and large-scale sidechains and backbone arrangements. Third, with the highly dynamic amyloid-β 1-42 (Aβ42) monomer, our deep learning model provided a comprehensive sampling of Aβ42's conformational landscape. Analysis of these synthetic conformations revealed conformational clusters that can be used to rationalize experimental findings. Additionally, the method can identify novel conformations with important interactions in atomistic details that are not included in the training data. New synthetic conformations showed distinct sidechain rearrangements that are probed by our EPR and amino acid substitution studies. This approach is highly transferable and can be used for any available data for training. The work also demonstrated the ability for deep learning to utilize learned natural atomistic motions in protein conformation sampling.

Sampling Conformational Ensembles of Highly Dynamic Proteins via Generative Deep Learning.

Proteins are inherently dynamic, and their conformational ensembles are functionally important in biology. Large-scale motions may govern protein structure-function relationship, and numerous transient but stable conformations of intrinsically disordered proteins (IDPs) can play a crucial role in biological function. Investigating conformational ensembles to understand regulations and disease-related aggregations of IDPs is challenging both experimentally and computationally. In this paper first an unsupervised deep learning-based model, termed Internal Coordinate Net (ICoN), is developed that learns the physical principles of conformational changes from molecular dynamics (MD) simulation data. Second, interpolating data points in the learned latent space are selected that rapidly identify novel synthetic conformations with sophisticated and large-scale sidechains and backbone arrangements. Third, with the highly dynamic amyloid-β1-42 (Aβ42) monomer, our deep learning model provided a comprehensive sampling of Aβ42's conformational landscape. Analysis of these synthetic conformations revealed conformational clusters that can be used to rationalize experimental findings. Additionally, the method can identify novel conformations with important interactions in atomistic details that are not included in the training data. New synthetic conformations showed distinct sidechain rearrangements that are probed by our EPR and amino acid substitution studies. The proposed approach is highly transferable and can be used for any available data for training. The work also demonstrated the ability for deep learning to utilize learned natural atomistic motions in protein conformation sampling.

Performance prediction and geometry optimization of ejector in PEMFC system using coupled CFD-BPNN and genetic algorithm

The ejector is a promising hydrogen recirculation device in proton exchange membrane fuel cell (PEMFC) systems. However, the limited entrainment performance of the ejector at wide operating conditions hinders its development and widespread application in PEMFC systems. To address this challenge and design a high-performance ejector adapted to the wide power range of PEMFC systems, a backpropagation neural network (BPNN) model with computational fluid dynamics (CFD) simulation data is developed. The sensitivity analysis of geometric parameters based on the CFD model reveals that the nozzle throat diameter and the mixing chamber diameter exert the most pronounced impact on the entrainment performance, with average influence rates of 24% and 57%, respectively. Furthermore, two advanced multi-objective genetic algorithms are applied to improve ejector performance. The linear weighted genetic algorithm (LWGA) method proves effective in elevating the overall ejector performance, achieving a remarkable enhancement in entrainment performance of up to 7.9%. On the other hand, the non-dominated sorting genetic algorithm (NSGA- II) method is favorable for expanding the operational power range of the ejector by 30%, as well as a 4.5% increase in overall ejector performance. This work provides a robust framework for designing and optimizing high-performance ejectors in high-power PEMFC systems.

Isoflavones and lysozyme interplay: Molecular insights into binding mechanisms and inhibitory efficacies of isoflavones against protein modification

In this study, we investigated the complexation of bioactive isoflavones, specifically biochanin A (BCA) and genistein (GEN), with hen egg white lysozyme (HEWL) and explored their inhibitory effects on HEWL modification using a combination of multi-spectroscopic and computational methods. The observed binding affinity was of a moderate nature (on the order of 104 M−1), and a static quenching mechanism was identified in the fluorescence quenching process. Notably, the binding constant (Kb) for GEN (4.449 ± 0.262 × 104 M−1) was found to be higher than that for BCA (3.707 ± 0.108 × 104 M−1) towards HEWL. Our spectroscopic measurements, complemented by molecular docking calculations, suggested the involvement of Trp62 in the binding site of the isoflavones within the geometry of HEWL. The micro-environment surrounding the Trp-residues exhibited an increase in hydrophilicity, as indicated by Synchronous fluorescence (SFS) and three-dimensional fluorescence (3D) studies. Interestingly, circular dichroism (CD) studies revealed no marked alteration in the secondary structure of HEWL upon binding with the isoflavones. Furthermore, our investigation into the interaction patterns, employing FTIR and molecular docking studies, revealed a predominance of hydrogen bonding and hydrophobic interactions. Beyond the binding study, the isoflavones demonstrated a promising inhibitory effect on the d-ribose-mediated glycation of HEWL, as well as on HEWL fibrillation, as evidenced by fluorescence emission studies. Our findings not only exhibited an excellent correlation with experimental observations but also provided precise insights into the location and dynamics of isoflavones within the binding site through detailed analyses of molecular docking and molecular dynamics simulation data.

Synthesis and characterization of bis-isoxazoline-thiosemicarbazone as a corrosion inhibitor for carbon steel: Experimental study, and molecular simulation

In this study, a novel compound, bis-isoxazoline-thiosemicarbazone (BIT), was synthesized and tested for its ability to protect carbon steel from corrosion in a 1 M HCl solution. The compound's structure was confirmed using spectroscopic techniques such as 1H NMR, 13C NMR, and HRMS. Various methods, including mass loss analysis, potentiodynamic polarization (PDP), electrochemical impedance spectroscopy (EIS), SEM, and EDX, were employed to assess BIT's effectiveness in inhibiting corrosion and its adsorption behavior on carbon steel in acidic conditions. The findings demonstrated the high effectiveness of BIT as a corrosion inhibitor, with a 95 % inhibition rate observed at a concentration of 0.77 mM and a temperature of 298°K. Analysis through PDP revealed that BIT operates as a mixed-type inhibitor, conforming to the Langmuir adsorption isotherm during its attachment to the carbon steel surface, indicating a mixed (chemical-physical) adsorption process.SEM-EDX analysis demonstrated even distribution of the inhibitor on the steel surface. Notably, BIT contains several functional groups (methoxy, carbonyl, isoxazolyl,..) and heteroatoms (N, O, and S) that play a crucial role, as supported by Density Functional Theory (DFT) studies. Experimental findings were validated by DFT and molecular dynamics (MD) simulation data.

A hybrid Gaussian mixture/DSMC approach to study the Fourier thermal problem

In rarefied gas dynamics scattering kernels deserve special attention since they contain all the essential information about the effects of physical and chemical properties of the gas–solid surface interface on the gas scattering process. However, to study the impact of the gas–surface interactions on the large-scale behavior of fluid flows, these scattering kernels need to be integrated in larger-scale models like Direct Simulation Monte Carlo (DSMC). In this work, the Gaussian mixture (GM) model, an unsupervised machine learning approach, is utilized to establish a scattering kernel for monoatomic (Ar) and diatomic (H2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\hbox {H}_{2}$$\\end{document}) gases directly from Molecular Dynamics (MD) simulations data. The GM scattering kernel is coupled to a pure DSMC solver to study isothermal and non-isothermal rarefied gas flows in a system with two parallel walls. To fully examine the coupling mechanism between the GM scattering kernel and the DSMC approach, a one-to-one correspondence between MD and DSMC particles is considered here. Benchmarked by MD results, the performance of the GM-DSMC is assessed against the Cercignani–Lampis–Lord (CLL) kernel incorporated into DSMC simulation (CLL-DSMC). The comparison of various physical and stochastic parameters shows the better performance of the GM-DSMC approach. Especially for the diatomic system, the GM-DSMC outperforms the CLL-DSMC approach. The fundamental superiority of the GM-DSMC approach confirms its potential as a multi-scale simulation approach for accurately measuring flow field properties in systems with highly nonequilibrium conditions.