The paper presents a multi-scale modelling approach for simulating macromolecules in fluid flows. Macromolecule transport at low number densities is frequently encountered in biomedical devices, such as separators, de...The paper presents a multi-scale modelling approach for simulating macromolecules in fluid flows. Macromolecule transport at low number densities is frequently encountered in biomedical devices, such as separators, detection and analysis systems. Accurate modelling of this process is challenging due to the wide range of physical scales involved. The continuum approach is not valid for low solute concentrations, but the large timescales of the fluid flow make purely molecular simulations prohibitively expensive. A promising multi-scale modelling strategy is provided by the meta-modelling approach considered in this paper. Meta-models are based on the coupled solution of fluid flow equations and equations of motion for a simplified mechanical model of macromolecules. The approach enables simulation of individual macromolecules at macroscopic time scales. Meta-models often rely on particle-corrector algorithms, which impose length constraints on the mechanical model. Lack of robustness of the particle-corrector algorithm employed can lead to slow convergence and numerical instability. A new FAst Linear COrrector (FALCO) algorithm is introduced in this paper, which significantly improves computational efficiency in comparison with the widely used SHAKE algorithm. Validation of the new particle corrector against a simple analytic solution is performed and improved convergence is demonstrated for ssDNA motion in a lid-driven micro-cavity.展开更多
This paper describes a numerical simulation in the Amazon water system, aiming to develop a quasi-three-dimensional numerical tool for refined modeling of turbulent flow and passive transport of mass in natural waters...This paper describes a numerical simulation in the Amazon water system, aiming to develop a quasi-three-dimensional numerical tool for refined modeling of turbulent flow and passive transport of mass in natural waters. Three depth-averaged two-equation turbulence closure models, k-ε,k-w, and k-w, were used to close the non-simplified quasi-three-dimensional hydrodynamic fundamental governing equations. The discretized equations were solved with the advanced multi-grid iterative method using non-orthogonal body-fitted coarse and fine grids with collocated variable arrangement. Except for steady flow computation, the processes of contaminant inpouring and plume development at the beginning of discharge, caused by a side-discharge of a tributary, have also been numerically investigated. The three depth-averaged two-equation closure models are all suitable for modeling strong mixing turbulence. The newly established turbulence models such as the k-w model, with a higher order of magnitude of the turbulence parameter, provide a possibility for improving computational precision.展开更多
现有的网联自动驾驶车辆(Connected and Automated Vehicles,CAV)换道决策模型鲁棒性较差,存在安全隐患,且单纯依赖自车信息、较小范围内的感知信息,难以在CAV与人工驾驶车辆(Human-Driven Vehicles,HDV)混行的环境中推断出最优动作.综...现有的网联自动驾驶车辆(Connected and Automated Vehicles,CAV)换道决策模型鲁棒性较差,存在安全隐患,且单纯依赖自车信息、较小范围内的感知信息,难以在CAV与人工驾驶车辆(Human-Driven Vehicles,HDV)混行的环境中推断出最优动作.综合考虑感知信息、自车以及车-车通信(Vehicle-to-Vehicle,V2V)范围内上、下游CAV信息,提出一种混合交通流环境下集成多源信息融合的深度强化学习(Multi-Source Information Fusion Deep Reinforcement Learning,MSIF-DRL)端到端网联自动驾驶换道决策模型.首先,构建含有多源信息的状态空间,并为不同信息分配权重;其次,通过编码网络将各种动态信息编码到高维特征空间,进行信息融合得到特征图;然后,将其扁平化送入拥有优先经验回放机制的竞争双深度Q网络中,进行动作的选择和评估;最后,分别设计适用于主线、匝道CAV的奖励函数引导所提MSIF-DRL模型解决高速公路合流区驾驶场景中CAV的自由以及强制换道问题.基于SUMO软件在各种仿真条件下进行实验,将所提出的MSIF-DRL换道决策模型与现有换道模型进行比较,验证其有效性和优越性.研究结果表明:相较于现有模型,所提MSIF-DRL模型在各种仿真条件下均拥有最高的奖励值、换道成功率、合流成功率、平均行车速度、舒适性以及最低的碰撞风险,其中换道成功率、合流成功率、平均行车速度最大分别提升了29.17%、27.71%、17.43%;随着渗透率的提高,该模型在处理混合交通流环境下CAV的换道决策问题时拥有更强的性能和鲁棒性.展开更多
基金supported in part by the European Commission under the 6th Framework Program (Project: DINAMICS, NMP4-CT-2007-026804).
文摘The paper presents a multi-scale modelling approach for simulating macromolecules in fluid flows. Macromolecule transport at low number densities is frequently encountered in biomedical devices, such as separators, detection and analysis systems. Accurate modelling of this process is challenging due to the wide range of physical scales involved. The continuum approach is not valid for low solute concentrations, but the large timescales of the fluid flow make purely molecular simulations prohibitively expensive. A promising multi-scale modelling strategy is provided by the meta-modelling approach considered in this paper. Meta-models are based on the coupled solution of fluid flow equations and equations of motion for a simplified mechanical model of macromolecules. The approach enables simulation of individual macromolecules at macroscopic time scales. Meta-models often rely on particle-corrector algorithms, which impose length constraints on the mechanical model. Lack of robustness of the particle-corrector algorithm employed can lead to slow convergence and numerical instability. A new FAst Linear COrrector (FALCO) algorithm is introduced in this paper, which significantly improves computational efficiency in comparison with the widely used SHAKE algorithm. Validation of the new particle corrector against a simple analytic solution is performed and improved convergence is demonstrated for ssDNA motion in a lid-driven micro-cavity.
基金supported by FAPESP (Foundation for Supporting Research in So Paulo State), Brazil, of the PIPE Project (Grant No. 2006/56475-3)
文摘This paper describes a numerical simulation in the Amazon water system, aiming to develop a quasi-three-dimensional numerical tool for refined modeling of turbulent flow and passive transport of mass in natural waters. Three depth-averaged two-equation turbulence closure models, k-ε,k-w, and k-w, were used to close the non-simplified quasi-three-dimensional hydrodynamic fundamental governing equations. The discretized equations were solved with the advanced multi-grid iterative method using non-orthogonal body-fitted coarse and fine grids with collocated variable arrangement. Except for steady flow computation, the processes of contaminant inpouring and plume development at the beginning of discharge, caused by a side-discharge of a tributary, have also been numerically investigated. The three depth-averaged two-equation closure models are all suitable for modeling strong mixing turbulence. The newly established turbulence models such as the k-w model, with a higher order of magnitude of the turbulence parameter, provide a possibility for improving computational precision.
文摘现有的网联自动驾驶车辆(Connected and Automated Vehicles,CAV)换道决策模型鲁棒性较差,存在安全隐患,且单纯依赖自车信息、较小范围内的感知信息,难以在CAV与人工驾驶车辆(Human-Driven Vehicles,HDV)混行的环境中推断出最优动作.综合考虑感知信息、自车以及车-车通信(Vehicle-to-Vehicle,V2V)范围内上、下游CAV信息,提出一种混合交通流环境下集成多源信息融合的深度强化学习(Multi-Source Information Fusion Deep Reinforcement Learning,MSIF-DRL)端到端网联自动驾驶换道决策模型.首先,构建含有多源信息的状态空间,并为不同信息分配权重;其次,通过编码网络将各种动态信息编码到高维特征空间,进行信息融合得到特征图;然后,将其扁平化送入拥有优先经验回放机制的竞争双深度Q网络中,进行动作的选择和评估;最后,分别设计适用于主线、匝道CAV的奖励函数引导所提MSIF-DRL模型解决高速公路合流区驾驶场景中CAV的自由以及强制换道问题.基于SUMO软件在各种仿真条件下进行实验,将所提出的MSIF-DRL换道决策模型与现有换道模型进行比较,验证其有效性和优越性.研究结果表明:相较于现有模型,所提MSIF-DRL模型在各种仿真条件下均拥有最高的奖励值、换道成功率、合流成功率、平均行车速度、舒适性以及最低的碰撞风险,其中换道成功率、合流成功率、平均行车速度最大分别提升了29.17%、27.71%、17.43%;随着渗透率的提高,该模型在处理混合交通流环境下CAV的换道决策问题时拥有更强的性能和鲁棒性.