基于深度学习的超高速碰撞碎片云生成模型

数据驱动的模型已经被广泛研究,并成功应用到了计算力学。基于深度学习技术,提出一种新的采用数据驱动的碎片云生成模型。此模型可以学习SPH数值模拟结果,然后在多种控制条件下快速生成碎片云。在模型训练前的数据预处理阶段,对SPH模拟结果进行空间网格划分和质量聚合,实现了改善数据分布规律、加速模型训练和提升模型泛化性的目的。以高速靶球撞击薄壁圆筒后的碎片云质量分布为例,模拟并测试了多种控制条件下深度学习模型计算结果的正确性和稳定性,以及计算速度的高效性。实验证明,深度学习模型可以从训练集学习碎片云的物理规律,然后在训练集控制参数范围内进行良好的推理及插值;并且可以在训练数据集控制参数范围外,进行小范围推理预测;同时深度学习模型的计算速度远快于SPH方法。通过深度学习方法建立碎片云模型,可能是一种在空间飞行器防护结构原型设计阶段,实现碎片云实时生成的潜在方案。     

基于深度学习的超高速碰撞碎片云生成模型

数据驱动的模型已经被广泛研究,并成功应用到了计算力学。基于深度学习技术,提出一种新的采用数据驱动的碎片云生成模型。此模型可以学习SPH数值模拟结果,然后在多种控制条件下快速生成碎片云。在模型训练前的数据预处理阶段,对SPH模拟结果进行空间网格划分和质量聚合,实现了改善数据分布规律、加速模型训练和提升模型泛化性的目的。以高速靶球撞击薄壁圆筒后的碎片云质量分布为例,模拟并测试了多种控制条件下深度学习模型计算结果的正确性和稳定性,以及计算速度的高效性。实验证明,深度学习模型可以从训练集学习碎片云的物理规律,然后在训练集控制参数范围内进行良好的推理及插值;并且可以在训练数据集控制参数范围外,进行小范围推理预测;同时深度学习模型的计算速度远快于SPH方法。通过深度学习方法建立碎片云模型,可能是一种在空间飞行器防护结构原型设计阶段,实现碎片云实时生成的潜在方案。     

物理增强的流场深度学习建模与模拟方法

流体运动理论上可用Navier?Stokes方程描述, 但由于对流项带来的非线性, 仅在少数情况可求得方程解析解. 对于复杂工程流动问题, 数值模拟难以高效精准计算高雷诺数流场, 实验或现场测量难以获得流场丰富细节. 近年来, 人工智能技术快速发展, 深度学习等数据驱动技术可利用灵活网络结构, 借助高效优化算法, 获得对高维、非线性问题的强大逼近能力, 为研究流体力学计算方法带来新机遇. 有别于传统图像识别、自然语言处理等典型人工智能任务, 深度学习模型预测的流场需满足流体物理规律, 如Navier?Stokes方程、典型能谱等. 近期, 物理增强的流场深度学习建模与模拟方法快速发展, 正逐渐成为流体力学全新研究范式: 根据流体物理规律选取网络输入特征或设计网络架构的方法称为物理启发的深度学习方法, 直接将流体物理规律显式融入网络损失函数或网络架构的方法称为物理融合的深度学习方法. 研究内容涵盖流体力学降阶模型、流动控制方程求解领域.      

水下发射水动力的多尺度预测网络研究

空泡的演化和水动力特征的预测在航行体发射的设计中有非常重要的意义.人工智能技术已经成为了参数预测的重要手段.为了能够快速预测航行体水下发射过程的尾部压力的复杂变化,提出了一种多尺度深度学习网络.该网络模型以一维卷积网络(1DCNN)为基础,构建了一种编码--解码型网络结构,通过不同的采样频率将原始数据分解为光滑部分和脉...     

Data-driven feature identification and sparse representation of turbulent flows

Identifying coherent structures in fluid flows is of great importance for reduced order modelling and flow control. However, extracting such structures from experimental or numerical data obtained from a turbulent flow can be challenging. A number of modal decomposition algorithms have been proposed in recent years which decompose time-resolved snapshots of data into spatial modes, each associated with a single frequency and growth-rate. Most prominently among them is dynamic mode decomposition (DMD). However, DMD-like algorithms create an arbitrary number of modes. It is common practice to then choose a smaller subset of these modes, for the purpose of model reduction and analysis, based on some measure of significance. In this work, we present a method of post-processing DMD modes for extracting a small number of dynamically relevant modes. We achieve this through an iterative approach based on the graph-theoretic notion of maximal cliques to identify clusters of modes and representing each cluster with a single representative mode.     

Efficiency of randomised dynamic mode decomposition for reduced order modelling

ABSTRACT

The purpose of this paper is the identification of a reduced order model (ROM) from numerical code output by non-intrusive techniques (i.e. not requiring projecting of the governing equations onto the reduced basis modes). In this paper, we perform a comparison between two methods of model order reduction based on dynamic mode decomposition (DMD). The first method is a deterministic (classic) DMD technique endowed with a dynamic filtering criterion of selection of modes used in the ROM model. The second method is an adaptive randomised DMD algorithm (ARDMD) based on a randomised singular value decomposition. This produced an accelerating algorithm, which is endowed with a few additional advantages. In addition, the reduced order model is guaranteed to satisfy the boundary conditions of the full model, which is crucial for surrogate modelling. For numerical illustration, we use the shallow water equations model.     

基于尾流时程目标识别的流场参数选择研究

浸入流场中的固体壁面会形成高度复杂且具有一定特征的尾流流场, 利用尾流所包含的信息对物体的外形特征进行识别具有重要的应用价值. 然而, 在较高雷诺数情况下尾流流场形态及其时序特征复杂, 难以通过传统的数学物理方法对流场信号进行特征的识别与提取. 本文提出了基于尾流时程数据深度学习的流场特征提取与分析方法, 实现了基于一点的物理量时程进行流场中物体外形的识别; 同时, 对流场中不同物理参数时程的识别精度与识别结果进行分析与研究, 得到适用于目标识别的最优物理量参数. 通过对圆柱和方柱的尾流数据研究结果表明, 本文提出的基于卷积神经网络的模型具有好的训练收敛性和高的预测精度, 能够识别并提取得到时程数据中包含的流场特征, 采用流场横向速度时程作为物体外形识别信号的模型准确率高. 证明了本方法用于浸入流场中物体外形识别的可行性, 是一种目标识别的高精度方法.      

Reduced finite difference scheme and error estimates based on POD method for non-stationary Stokes equation

The proper orthogonal decomposition (POD) is a model reduction technique for the simulation of physical processes governed by partial differential equations (e.g.,fluid flows). It has been successfully used in the reduced-order modeling of complex systems. In this paper, the applications of the POD method are extended, i.e., the POD method is applied to a classical finite difference (FD) scheme for the non-stationary Stokes equation with a real practical applied background. A reduced FD scheme is established with lower dimensions and sufficiently high accuracy, and the error estimates are provided between the reduced and the classical FD solutions. Some numerical examples illustrate that the numerical results are consistent with theoretical conclusions. Moreover, it is shown that the reduced FD scheme based on the POD method is feasible and efficient in solving the FD scheme for the non-stationary Stokes equation.     

页岩凝析气藏相平衡的快速准确计算方法

对页岩油气藏中复杂流体的相平衡计算需要建立考虑毛细作用效应的先进的数值模型, 并设计出快速可靠的算法以应对实际工况中储层流体包含多达数十种组分的复杂情况. 本文将基于适合页岩油气藏常见组分的真实流体状态方程, 即Peng?Robinson状态方程构建具有热力学一致性的VT型孔观相平衡计算体系. 通过引入描述毛细压力做功的数学模型实现对页岩流体热力学性质更准确的刻画. 结合扩散界面模型建立动力学演化格式, 采用成熟的凸分裂方法求解摩尔数和体积分数的演变, 从而描述相平衡的动态过程. 在此基础上, 本文开发了一套具有自适应性的深度学习算法, 设计了独特的双网络结构以实现对不同流体中不同组分的广泛适用性. 该神经网络的输入和输出参数均在热力学分析的基础上选取关键的热力学性质参数, 并进行了全面的超参调试以确定最合适的网络架构和最后形成的预测模型的基本结构, 且通过多种深度学习技术解决了过拟合问题, 在显著加速了传统的基于迭代方法的闪蒸计算的同时保证了相平衡状态预测的准确性, 得到了较好的预测效果. 相分离判定自动整合在预测结果中, 且从最终预测结果可以显著地捕捉到毛细作用的影响. 这一套快速、准确、可靠地基于深度学习算法的页岩油气孔观相平衡计算体系可以为后续的多相流动模拟提供具有物理意义的相分布初场, 确定系统内各个阶段的相数, 并可以作为构建具有物理守恒性的多相数值模型的热力学基础.      

Strongly coupled simulation of fluid–structure interaction in a Francis hydroturbine

This work simulates a complex fluid flow in fluid–structure interaction (FSI). The flow under consideration is governed by Navier–Stokes equations for incompressible viscous fluids and modeled with the finite volume method. Large eddy simulation is used to simulate the unsteady turbulent flow. The structure is represented by a finite element formulation. The present work introduces a strongly coupled partitioned approach that is applied to complex flow in fluid machinery. In this approach, the fluid and structure equations are solved separately using different solvers, but are implicitly coupled into one single module based on sensitivity analysis of the important displacement and stress modes. The applied modes and their responses are used to build up a reduced‐order model. The proposed model is used to predict the unsteady flow fields of a 3D complete passage, involving in stay, guide vanes, and runner blades, for a Francis hydro turbine and FSI is considered. The computational results show that a fairly good convergence solution is achieved by using the reduced‐order model that is based on only a few displacement and stress modes, which largely reduces the computational cost, compared with traditional approaches. At the same time, a comparison of the numerical results of the model with available experimental data validates the methodology and assesses its accuracy. Copyright © 2008 John Wiley & Sons, Ltd.