Rundi Qiu, Haosen Dong, Jingzhu Wang, Chun Fan, Yiwei Wang
The physics-informed neural networks (PINNs) have shown great potential in solving a variety of high-dimensional partial differential equations (PDEs), but the complexity of a realistic problem still restricts the practical application of the PINNs for solving most complicated PDEs. In this paper, we propose a parallel framework for PINNs that is capable of modeling two-phase flows with complicated interface evolution. The proposed framework divides the problem into several simplified subproblems and solves them through training several PINNs on corresponding subdomains simultaneously. To enhance the accuracy of the parallel training framework in two-phase flow, the overlapping domain decomposition method is adopted. The optimal subnetwork sizes and partitioned method are systematically discussed, and a series of cases including a bubble rising, droplet splashing, and the Rayleigh–Taylor instability are applied for quantitative validation. The maximum relative error of quantitative values in these cases is 0.1319. Our results show that the proposed framework not only can accelerate the training procedure of PINNs, but also can capture the spatiotemporal evolution of the interface between various phases. This framework overcomes the difficulties of training PINNs to solve a forward problem in two-phase flow, and it is expected to model more realistic dynamic systems in nature.
{"title":"Modeling two-phase flows with complicated interface evolution using parallel physics-informed neural networks","authors":"Rundi Qiu, Haosen Dong, Jingzhu Wang, Chun Fan, Yiwei Wang","doi":"10.1063/5.0216609","DOIUrl":"https://doi.org/10.1063/5.0216609","url":null,"abstract":"The physics-informed neural networks (PINNs) have shown great potential in solving a variety of high-dimensional partial differential equations (PDEs), but the complexity of a realistic problem still restricts the practical application of the PINNs for solving most complicated PDEs. In this paper, we propose a parallel framework for PINNs that is capable of modeling two-phase flows with complicated interface evolution. The proposed framework divides the problem into several simplified subproblems and solves them through training several PINNs on corresponding subdomains simultaneously. To enhance the accuracy of the parallel training framework in two-phase flow, the overlapping domain decomposition method is adopted. The optimal subnetwork sizes and partitioned method are systematically discussed, and a series of cases including a bubble rising, droplet splashing, and the Rayleigh–Taylor instability are applied for quantitative validation. The maximum relative error of quantitative values in these cases is 0.1319. Our results show that the proposed framework not only can accelerate the training procedure of PINNs, but also can capture the spatiotemporal evolution of the interface between various phases. This framework overcomes the difficulties of training PINNs to solve a forward problem in two-phase flow, and it is expected to model more realistic dynamic systems in nature.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"206 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142258836","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Semi-local scales have been widely used in compressible wall-bounded turbulence, but it is still unclear whether they are applicable to the scaling of coherent structures, especially under conditions of high Mach number and cold wall temperature. By scrutinizing the direct numerical simulation dataset at different Mach numbers and wall temperatures, this paper demonstrates that the coherent structures normalized by semi-local scales are universal in size. In addition to this, we find that the ratios of Kolmogorov scales to semi-local scales are independent of Mach number and wall temperature. Thus, Kolmogorov scales can achieve the same scaling effect as the semi-local scales. The velocity spectra are also compared to verify the current scaling method quantitatively. A method to determine the threshold for the vortex identification criterion is proposed, allowing the same threshold for different cases to obtain vortices of similar size. The scaling of other statistics including turbulent kinetic energy, streamwise Reynolds normal stress, and root mean square of fluctuating vorticity is also investigated. A new velocity scale is proposed based on the total-stress-based transformation for mean streamwise velocity, which can collapse the profiles of these statistics more accurately than the semi-local velocity scale. The present paper demonstrates that through appropriate normalization, the structures and statistics of compressible turbulence become universal, reaffirming the validity of Morkovin's hypothesis even for the present high Mach number and cold wall cases.
{"title":"Scaling of coherent structures in compressible wall-bounded turbulence","authors":"Fuzhou Lyu, Chunxiao Xu","doi":"10.1063/5.0231296","DOIUrl":"https://doi.org/10.1063/5.0231296","url":null,"abstract":"Semi-local scales have been widely used in compressible wall-bounded turbulence, but it is still unclear whether they are applicable to the scaling of coherent structures, especially under conditions of high Mach number and cold wall temperature. By scrutinizing the direct numerical simulation dataset at different Mach numbers and wall temperatures, this paper demonstrates that the coherent structures normalized by semi-local scales are universal in size. In addition to this, we find that the ratios of Kolmogorov scales to semi-local scales are independent of Mach number and wall temperature. Thus, Kolmogorov scales can achieve the same scaling effect as the semi-local scales. The velocity spectra are also compared to verify the current scaling method quantitatively. A method to determine the threshold for the vortex identification criterion is proposed, allowing the same threshold for different cases to obtain vortices of similar size. The scaling of other statistics including turbulent kinetic energy, streamwise Reynolds normal stress, and root mean square of fluctuating vorticity is also investigated. A new velocity scale is proposed based on the total-stress-based transformation for mean streamwise velocity, which can collapse the profiles of these statistics more accurately than the semi-local velocity scale. The present paper demonstrates that through appropriate normalization, the structures and statistics of compressible turbulence become universal, reaffirming the validity of Morkovin's hypothesis even for the present high Mach number and cold wall cases.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"43 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142258848","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A theoretical and numerical investigation of electrokinetic flow is performed in a nanochannel with the charged symmetric corrugated surfaces. The perturbation and numerical solutions of electrokinetic flow variables are given, and the effects of corrugation geometry, such as wave amplitude and wave number, on the electrokinetic flow characteristics are systematically examined. The results show that the electrokinetic flow recirculation may occur easily at wave crest due to the electroviscous effect. The velocity profile is strongly dependent on wave number, but the maximum or minimum velocity may be insusceptible to wave number. Furthermore, the distributions of streaming potential and energy conversion efficiency are also investigated. We find that, for some special geometry of corrugations, the streaming current and conversion efficiency obtained from the present corrugated nanochannel are higher than that from the smooth nanochannel. Specially, when the dimensionless wave number is 0.5/π, the magnitude of streaming potential is enhanced about 29% at δ = 0.5 and the peak value of conversion efficiency is enhanced about 2% at δ = 0.1. We believe that the optimal corrugation geometry parameters can be of benefit in designing a microfluidic device with higher streaming current and conversion efficiency.
{"title":"Electrokinetic flow and energy conversion induced by streaming potential in nanochannels with symmetric corrugated walls","authors":"Zhiyong Xie, Xingyu Chen, Fang Tan","doi":"10.1063/5.0226494","DOIUrl":"https://doi.org/10.1063/5.0226494","url":null,"abstract":"A theoretical and numerical investigation of electrokinetic flow is performed in a nanochannel with the charged symmetric corrugated surfaces. The perturbation and numerical solutions of electrokinetic flow variables are given, and the effects of corrugation geometry, such as wave amplitude and wave number, on the electrokinetic flow characteristics are systematically examined. The results show that the electrokinetic flow recirculation may occur easily at wave crest due to the electroviscous effect. The velocity profile is strongly dependent on wave number, but the maximum or minimum velocity may be insusceptible to wave number. Furthermore, the distributions of streaming potential and energy conversion efficiency are also investigated. We find that, for some special geometry of corrugations, the streaming current and conversion efficiency obtained from the present corrugated nanochannel are higher than that from the smooth nanochannel. Specially, when the dimensionless wave number is 0.5/π, the magnitude of streaming potential is enhanced about 29% at δ = 0.5 and the peak value of conversion efficiency is enhanced about 2% at δ = 0.1. We believe that the optimal corrugation geometry parameters can be of benefit in designing a microfluidic device with higher streaming current and conversion efficiency.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"188 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227360","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The wettability and imbibition dynamics of water within 2-dimensional hexagonal boron nitride (h-BN) nanochannels were investigated through nanoscale molecular dynamics simulations. Results from the sessile drop and liquid plug methods indicate that the contact angle on h-BN is notably lower than that on graphene, with single-layer h-BN exhibiting greater hydrophobicity compared to multilayer h-BN. The disjoining pressure in liquid nanoplug was calculated to validate the Young–Laplace equation. During the imbibition process, the penetration length follows l2 = Slt. Simultaneously, the decrease in internal energy (ΔE) follows ΔE = −SEt1/2. While the Lucas–Washburn expression (l2 ∼ wt) can capture such behavior, it does not account for the dependence on channel width (w), where w = Nb, with N denoting the number of h-BN sheets and b the thickness. In wide nanoslits (N > 4), the penetration velocity decreases as the channel width increases. The final ΔE converge to the same value, and SE2/Sl remains constant. In narrow nanoslits (N ≤ 4), the penetration velocity does not decrease consistently with channel width. The final ΔE does not converge to a consistent value for N = 1, 1.5, and 2, and SE exhibits distinct trends with Sl. Comparisons reveal that water in h-BN nanochannels exhibits a notably higher imbibition velocity than in graphene due to differences in the driving force.
{"title":"Surface wettability and capillary flow of water in nanoslits of two-dimensional hexagonal-boron nitride","authors":"Ya-Wun Lu, Hsin-Yu Chang, Heng-Kwong Tsao, Yu-Jane Sheng","doi":"10.1063/5.0224117","DOIUrl":"https://doi.org/10.1063/5.0224117","url":null,"abstract":"The wettability and imbibition dynamics of water within 2-dimensional hexagonal boron nitride (h-BN) nanochannels were investigated through nanoscale molecular dynamics simulations. Results from the sessile drop and liquid plug methods indicate that the contact angle on h-BN is notably lower than that on graphene, with single-layer h-BN exhibiting greater hydrophobicity compared to multilayer h-BN. The disjoining pressure in liquid nanoplug was calculated to validate the Young–Laplace equation. During the imbibition process, the penetration length follows l2 = Slt. Simultaneously, the decrease in internal energy (ΔE) follows ΔE = −SEt1/2. While the Lucas–Washburn expression (l2 ∼ wt) can capture such behavior, it does not account for the dependence on channel width (w), where w = Nb, with N denoting the number of h-BN sheets and b the thickness. In wide nanoslits (N > 4), the penetration velocity decreases as the channel width increases. The final ΔE converge to the same value, and SE2/Sl remains constant. In narrow nanoslits (N ≤ 4), the penetration velocity does not decrease consistently with channel width. The final ΔE does not converge to a consistent value for N = 1, 1.5, and 2, and SE exhibits distinct trends with Sl. Comparisons reveal that water in h-BN nanochannels exhibits a notably higher imbibition velocity than in graphene due to differences in the driving force.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"12 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227239","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Non-equilibrium condensation (NQC) induced heat transfer in the supersonic nozzle (SSN) results in entropy production and alters the flow structure. The analysis of entropy production offers valuable insights for enhancing the design of industrial equipment by pinpointing the origins of energy losses. The interplay between frictional entropy, thermal entropy, and NQC is a fascinating but relatively unexplored subject in the field. This study aims to examine the impact of the superheat degree on NQC, frictional entropy, and thermal entropy. The goal is to enhance our understanding of the interconnectedness among these three parameters and their relationship. The findings revealed that within the SSN, the generation of frictional entropy surpasses that of thermal entropy generation. Upon analyzing the variations in entropy production with an increase in the degree of superheat, a general trend of ascending–descending can be observed for thermal, frictional, and total entropy productions. Furthermore, as the degree of superheat increases, both the droplet diameter and liquid mass fraction within the nozzle decrease. Optimization techniques were employed to determine the optimal degree of superheat for the given scenario. After the optimization process, the range of 70–90 was identified as the optimal degree of superheat. At a superheat degree of 70, the parameters of production entropy, input flow rate, condensation loss, and energy kinetics undergo changes of 19.3%, 9.8%, 99.9%, and 14.3%, respectively.
{"title":"A comprehensive investigation and optimization of superheat degree on performance of supersonic nozzle by considering non-equilibrium condensation and entropy generation analysis","authors":"Rujie Xia, Delu Li, Mohammad Ali Faghih Aliabadi","doi":"10.1063/5.0224884","DOIUrl":"https://doi.org/10.1063/5.0224884","url":null,"abstract":"Non-equilibrium condensation (NQC) induced heat transfer in the supersonic nozzle (SSN) results in entropy production and alters the flow structure. The analysis of entropy production offers valuable insights for enhancing the design of industrial equipment by pinpointing the origins of energy losses. The interplay between frictional entropy, thermal entropy, and NQC is a fascinating but relatively unexplored subject in the field. This study aims to examine the impact of the superheat degree on NQC, frictional entropy, and thermal entropy. The goal is to enhance our understanding of the interconnectedness among these three parameters and their relationship. The findings revealed that within the SSN, the generation of frictional entropy surpasses that of thermal entropy generation. Upon analyzing the variations in entropy production with an increase in the degree of superheat, a general trend of ascending–descending can be observed for thermal, frictional, and total entropy productions. Furthermore, as the degree of superheat increases, both the droplet diameter and liquid mass fraction within the nozzle decrease. Optimization techniques were employed to determine the optimal degree of superheat for the given scenario. After the optimization process, the range of 70–90 was identified as the optimal degree of superheat. At a superheat degree of 70, the parameters of production entropy, input flow rate, condensation loss, and energy kinetics undergo changes of 19.3%, 9.8%, 99.9%, and 14.3%, respectively.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"62 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227195","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Accurate prediction of scour depth is essential for the safety of the bridge. The downflow upstream of the pier plays a significant role in scour hole formation. The present study used Lagrangian coherent structures (LCSs) to derive an expression to estimate the force exerted by the downflow on the sediment bed. The LCSs extracted a trapping region upstream of the pier, which trapped the fluid, which was then converted into downflow. The expressions derived in this Letter can be used to improve the efficiency of scour depth prediction equations.
{"title":"Lagrangian coherent structures around a bridge pier with scour hole","authors":"Murali Krishnamraju Kalidindi, Rakesh Khosa","doi":"10.1063/5.0229952","DOIUrl":"https://doi.org/10.1063/5.0229952","url":null,"abstract":"Accurate prediction of scour depth is essential for the safety of the bridge. The downflow upstream of the pier plays a significant role in scour hole formation. The present study used Lagrangian coherent structures (LCSs) to derive an expression to estimate the force exerted by the downflow on the sediment bed. The LCSs extracted a trapping region upstream of the pier, which trapped the fluid, which was then converted into downflow. The expressions derived in this Letter can be used to improve the efficiency of scour depth prediction equations.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"319 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227198","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper, we construct first- and second-order implicit–explicit schemes for the closed-loop geothermal system, which includes the heat transfer between the porous media flow with Darcy equation in the geothermal reservoir and the free flow with Navier–Stokes equation in the pipe. The constructed fully discrete schemes are based on the exponential auxiliary variable method in time, which we have proposed in Li et al. [“New SAV-pressure correction methods for the Navier-Stokes equations: Stability and error analysis,” Math. Comput. 91, 141–167 (2022)] and the finite element method in space. These schemes are linear and uniquely solvable, decoupling not only the two flow regions but also the temperature field, and only require solving a sequence of linear differential equations with constant coefficients at each time step. In addition, we rigorously prove that the constructed first- and second-order schemes are unconditionally stable without any time step and stability parameter restrictions. Finally, some numerical simulations, including convergence tests, the benchmark problem for thermal convection in a square cavity, and the heat transfer in simplified closed-loop geothermal systems, are demonstrated to present the reliability and efficiency of the constructed schemes.
本文构建了闭环地热系统的一阶和二阶隐式-显式方案，其中包括地热储层中含达西方程的多孔介质流与管道中含纳维-斯托克斯方程的自由流之间的传热。所构建的全离散方案基于指数辅助变量时间法，我们在 Li 等人的论文["Navier-Stokes 方程的新 SAV 压力修正方法：稳定性和误差分析"，Math.Comput.91，141-167 (2022)]和空间有限元法。这些方案是线性和唯一可解的，不仅解耦了两个流动区域，还解耦了温度场，并且只需要在每个时间步求解一连串具有常数系数的线性微分方程。此外，我们还严格证明了所构建的一阶和二阶方案是无条件稳定的，不受任何时间步长和稳定参数的限制。最后，我们演示了一些数值模拟，包括收敛性测试、方形空腔中热对流的基准问题以及简化闭环地热系统中的传热问题，以展示所构建方案的可靠性和效率。
{"title":"First- and second-order unconditionally stable and decoupled schemes for the closed-loop geothermal system based on the coupled multiphysics model","authors":"Xinhui Wang, Xiaoli Li","doi":"10.1063/5.0228565","DOIUrl":"https://doi.org/10.1063/5.0228565","url":null,"abstract":"In this paper, we construct first- and second-order implicit–explicit schemes for the closed-loop geothermal system, which includes the heat transfer between the porous media flow with Darcy equation in the geothermal reservoir and the free flow with Navier–Stokes equation in the pipe. The constructed fully discrete schemes are based on the exponential auxiliary variable method in time, which we have proposed in Li et al. [“New SAV-pressure correction methods for the Navier-Stokes equations: Stability and error analysis,” Math. Comput. 91, 141–167 (2022)] and the finite element method in space. These schemes are linear and uniquely solvable, decoupling not only the two flow regions but also the temperature field, and only require solving a sequence of linear differential equations with constant coefficients at each time step. In addition, we rigorously prove that the constructed first- and second-order schemes are unconditionally stable without any time step and stability parameter restrictions. Finally, some numerical simulations, including convergence tests, the benchmark problem for thermal convection in a square cavity, and the heat transfer in simplified closed-loop geothermal systems, are demonstrated to present the reliability and efficiency of the constructed schemes.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"54 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227199","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This study proposes a novel model of a dual-purpose nonlinear wave farm, wherein multiple wave energy converters (WECs) equipped with nonlinear stiffness mechanism (NSM) are deployed for energy production and wave attenuation. A hybrid semi-analytical approach integrating the eigenfunction expansion matching method (EEMM) with the multi-harmonic balance method (MHBM) is developed to address the nonlinear wave-structure interactions among an array of WECs. Each device is modeled as a truncated cylinder, and the effects of the nonlinear interaction on power absorption and wave evolution from the array are studied. The analytical results are validated through published experimental results and computational fluid dynamics (CFD) results. A multi-parameter analysis is conducted to explore the impact of various factors including power takeoff (PTO) damping, NSM configuration, wave direction, and layout geometry on both wave power absorption and wave evolution. The results demonstrate that the nonlinear wave farm exhibits improved power-capture efficiency and enhanced wave attenuation compared to the linear wave farm, attributed to the phase control mechanism of NSM. This work may provide profound guidelines for large-scale wave energy exploitation and coast protection.
{"title":"Dual-purpose wave farm with nonlinear stiffness mechanism for energy extraction and wave attenuation","authors":"Huaqing Jin, Haicheng Zhang, Siming Zheng, Ye Lu, Daolin Xu, Deborah Greaves","doi":"10.1063/5.0227165","DOIUrl":"https://doi.org/10.1063/5.0227165","url":null,"abstract":"This study proposes a novel model of a dual-purpose nonlinear wave farm, wherein multiple wave energy converters (WECs) equipped with nonlinear stiffness mechanism (NSM) are deployed for energy production and wave attenuation. A hybrid semi-analytical approach integrating the eigenfunction expansion matching method (EEMM) with the multi-harmonic balance method (MHBM) is developed to address the nonlinear wave-structure interactions among an array of WECs. Each device is modeled as a truncated cylinder, and the effects of the nonlinear interaction on power absorption and wave evolution from the array are studied. The analytical results are validated through published experimental results and computational fluid dynamics (CFD) results. A multi-parameter analysis is conducted to explore the impact of various factors including power takeoff (PTO) damping, NSM configuration, wave direction, and layout geometry on both wave power absorption and wave evolution. The results demonstrate that the nonlinear wave farm exhibits improved power-capture efficiency and enhanced wave attenuation compared to the linear wave farm, attributed to the phase control mechanism of NSM. This work may provide profound guidelines for large-scale wave energy exploitation and coast protection.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"1 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227200","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yan Liu, Qingyang Zhang, Xinhai Chen, Chuanfu Xu, Qinglin Wang, Jie Liu
The rapid development of artificial intelligence has promoted the emergence of new flow field prediction methods. These methods address challenges posed by nonlinear problems and significantly reduce computational time and cost compared to traditional numerical simulations. However, they often struggle to capture the dynamic sparse characteristics of the flow field effectively. To bridge this gap, we introduce LKFlowNet, a new large kernel convolutional neural network specifically designed for complex flow fields in nonlinear fluid dynamics systems. LKFlowNet adopts a multi-branch large kernel convolution computing architecture, which can skillfully handle the complex nonlinear dynamic characteristics of flow changes. Drawing inspiration from the dilated convolution mechanism, we developed the RepDWConv block, a re-parameterized depthwise convolution that extends the convolutional kernel's coverage. This enhancement improves the model's ability to capture long-range dependencies and sparse structural features in fluid dynamics. Additionally, a customized physical loss function ensures accuracy and physical consistency in flow field reconstruction. Comparative studies reveal that LKFlowNet significantly outperforms existing neural network architectures, providing more accurate and physically consistent predictions in complex nonlinear variations such as velocity and pressure fields. The model demonstrates strong versatility and scalability, accurately predicting the flow field of various geometric configurations without modifying the architecture. This capability positions LKFlowNet as a promising new direction in fluid dynamics research, potentially revolutionizing flow field prediction by combining high efficiency and accuracy. Our results suggest that LKFlowNet could become an indispensable tool in intelligent flow field prediction, reshaping the analysis and processing of fluid dynamics.
{"title":"LKFlowNet: A deep neural network based on large kernel convolution for fast and accurate nonlinear fluid-changing prediction","authors":"Yan Liu, Qingyang Zhang, Xinhai Chen, Chuanfu Xu, Qinglin Wang, Jie Liu","doi":"10.1063/5.0221881","DOIUrl":"https://doi.org/10.1063/5.0221881","url":null,"abstract":"The rapid development of artificial intelligence has promoted the emergence of new flow field prediction methods. These methods address challenges posed by nonlinear problems and significantly reduce computational time and cost compared to traditional numerical simulations. However, they often struggle to capture the dynamic sparse characteristics of the flow field effectively. To bridge this gap, we introduce LKFlowNet, a new large kernel convolutional neural network specifically designed for complex flow fields in nonlinear fluid dynamics systems. LKFlowNet adopts a multi-branch large kernel convolution computing architecture, which can skillfully handle the complex nonlinear dynamic characteristics of flow changes. Drawing inspiration from the dilated convolution mechanism, we developed the RepDWConv block, a re-parameterized depthwise convolution that extends the convolutional kernel's coverage. This enhancement improves the model's ability to capture long-range dependencies and sparse structural features in fluid dynamics. Additionally, a customized physical loss function ensures accuracy and physical consistency in flow field reconstruction. Comparative studies reveal that LKFlowNet significantly outperforms existing neural network architectures, providing more accurate and physically consistent predictions in complex nonlinear variations such as velocity and pressure fields. The model demonstrates strong versatility and scalability, accurately predicting the flow field of various geometric configurations without modifying the architecture. This capability positions LKFlowNet as a promising new direction in fluid dynamics research, potentially revolutionizing flow field prediction by combining high efficiency and accuracy. Our results suggest that LKFlowNet could become an indispensable tool in intelligent flow field prediction, reshaping the analysis and processing of fluid dynamics.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"62 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Sajjad Pashazadeh, Arvindh Seshadri Suresh, Viney Ghai, Tobias Moberg, Anders Brolin, Roland Kádár
A general procedure for combining material functions and numerical modeling to predict the orientation of highly filled wood polymer composites (WPCs) in a single screw extrusion and validation thereof is elaborated in this study. Capillary rheometry was used to determine the shear viscosity and wall slip functions as well as the melt density of the biocomposites. The numerical model consisted of a model film die where the melt flow was simulated using a finite element method in the generalized Newtonian constitute equation framework. Fiber orientation was modeled using the Folgar–Tucker approach and included fiber–fiber interaction during the process. Reference extrusion tests were performed on a single screw extruder on the biocomposites. The extrusion setup included two melt pressure transducers that were used to determine the die inlet initial conditions (end of the extruder/die inlet) and provide feedback on the wall slip boundary conditions (pressure discharge along the die). Overall, the pressure error between experiments and simulations was less than 6.5% for all screw speeds investigated in 20 wt. % WPCs. Extrudates were produced, and the wood fiber orientation was estimated based on scanning electron microscopy micrographs and image analysis and compared with the simulations of fiber orientation. We show that the general procedure outlined can be calibrated to predict the overall orientation distribution of wood fiber biocomposites during single screw extrusion.
{"title":"Predicting orientation in extruded wood polymer composites","authors":"Sajjad Pashazadeh, Arvindh Seshadri Suresh, Viney Ghai, Tobias Moberg, Anders Brolin, Roland Kádár","doi":"10.1063/5.0224011","DOIUrl":"https://doi.org/10.1063/5.0224011","url":null,"abstract":"A general procedure for combining material functions and numerical modeling to predict the orientation of highly filled wood polymer composites (WPCs) in a single screw extrusion and validation thereof is elaborated in this study. Capillary rheometry was used to determine the shear viscosity and wall slip functions as well as the melt density of the biocomposites. The numerical model consisted of a model film die where the melt flow was simulated using a finite element method in the generalized Newtonian constitute equation framework. Fiber orientation was modeled using the Folgar–Tucker approach and included fiber–fiber interaction during the process. Reference extrusion tests were performed on a single screw extruder on the biocomposites. The extrusion setup included two melt pressure transducers that were used to determine the die inlet initial conditions (end of the extruder/die inlet) and provide feedback on the wall slip boundary conditions (pressure discharge along the die). Overall, the pressure error between experiments and simulations was less than 6.5% for all screw speeds investigated in 20 wt. % WPCs. Extrudates were produced, and the wood fiber orientation was estimated based on scanning electron microscopy micrographs and image analysis and compared with the simulations of fiber orientation. We show that the general procedure outlined can be calibrated to predict the overall orientation distribution of wood fiber biocomposites during single screw extrusion.","PeriodicalId":20066,"journal":{"name":"Physics of Fluids","volume":"74 1","pages":""},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142227197","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：