Load Extraction from Photoelastic Images Using Neural Networks

Photoelastic materials develop colored fringes under white light when subjected to mechanical stresses, which can be viewed through a polariscope. This technique has traditionally been used for stress analysis of loaded components, however, this can also be potentially used in sensing applications where the requirement may be measurement of the stimulating forces causing the generation of fringes. This leads to inverse photoelastic problem where the developed image can be analyzed for the input forces. However, there could be infinite number of possible solutions which cannot be determined by conventional techniques. This paper presents neural networks based approach to solve this problem. Experiments conducted to prove the principle have been verified with theoretical results and finite element analysis of loaded specimens. The developed technique, if generalized, can be implemented for whole-field analysis of the stress patterns involving complex fringes under different loading conditions. This can also provide direct visualization of the stress field, which may find application in a variety of specialized areas including biomedical engineering and robotics.

结构紧固连接件损伤的神经网络诊断研究

在结构紧固连接件损伤诊断中采用了函数连接型神经网络。与常用的ＢＰ网络不同，函数连接网络通过对输入模式的增强来实现复杂模式的分类。它借助某种函数变换，将原始的模式特征向量变换成另一个更高维的形式，使原始空间里不能区分的类别，在这个高维空间中得以区分，减少了神经网络的层数，提高了网络的学习速度。诊断时选用正弦激振下结构中多个位置上传感器稳态响应的幅值构成表征损伤的特征向量，简化了特征提取的过程。数值模拟和实验结果表明，这一方法能对多个位置的损伤进行准确的识别     

A Turbulent-Wake Estimation Using Radial Basis Function Neural Networks

An estimation of the fluctuations in the passive-tracer concentration for the turbulent wake behind an airfoil is presented. The estimation is based on experimental modelling using Radial Basis Function Neural Networks. For the experiment the fluctuations of the concentration in the turbulent wake were recorded with a visualization method. The records of the concentration in the selected regions of the turbulent wake were used as the input and output regions for the training and estimation with neural networks. The uncertainty of the estimation increased with increasing distance between the input and the output regions. The power spectra, the spatial correlation functions and the profiles of the concentration were calculated from the measured and estimated fluctuations of the concentration. The measured and estimated concentration power spectra were in reasonable agreement. The measured and estimated spatial correlation functions and the profiles of the concentration showed a similar agreement.     

一种基于神经网络的传感器故障诊断方法

提出了基于两级神经网络结构的多传感器斜置组件的故障诊断方法,以消除传感器安装误差、刻度系数误差以及常值偏差对故障检测与隔离准确性的影响。与基于参数估计的奇偶向量补偿方法相比,这种方法不需要各项误差的动态模型和噪声的统计特性。     

Existence and Uniqueness for Stochastic 2D Euler Flows with Bounded Vorticity

The strong existence and the pathwise uniqueness of solutions with \({L^{\infty}}\)-vorticity of the 2D stochastic Euler equations are proved. The noise is multiplicative and it involves the first derivatives. A Lagrangian approach is implemented, where a stochastic flow solving a nonlinear flow equation is constructed. The stability under regularizations is also proved.     

Earthquake Forecasting Using Neural Networks: Results and Future Work

In 1994, a new earthquake forecasting method was developed, that integrated in a neural network several forecasting tools that had been originally developed for financial analysis. This method was tested with the seismicity of the Azores, predicting the July, 1998, and the January, 2004, earthquakes, albeit within very wide time and location windows. Work is beginning to integrate physical precursors in the neural network, in order to narrow the forecasting windows     

结构试验机的神经网络内模自适应跟踪控制系统

针对结构试验系统的非线性和不确定性特性，研究了一种基于神经网络的非线性内模自适应加载控制方法。引入的神经网络内模可跟踪学习对象的时变动力学，控制器的设计较少依赖于对象的先验知识，控制参数的调整是基于被控过程的测量信息，利用导出的神经网络算法来实现的。实验结果证明该系统具有良好的控制效果。     

Assessment of Delayed DES and Improved Delayed DES Combined with a Shear-Layer-Adapted Subgrid Length-Scale in Separated Flows

A comparative study is conducted between the original versions of Delayed Detached-Eddy Simulation (DDES) and Improved DDES (IDDES) and these approaches combined with a new (shear layer adapted) definition of the subgrid length-scale recently proposed in Shur et al. (Flow Turbul. Combust. 95(4), 709–737, 2015). This definition is aimed at accelerating the transition to resolved turbulence in separated shear-layers, which significant delay is typical of the non-zonal hybrid RANS-LES models, in general, and DES-like approaches, in particular. An objective of the study is widening the validation database of the new solutions-dependent definition of the length-scale compared to that employed in the original work of Shur et al. In order to reach this, three different complex separated flows with well-understood flow physics were considered, which all are widely used for the validation of different CFD approaches. These flows are: a flow with non-fixed pressure-induced separation and reattachment (wall-mounted hump), a massively separated flow (NACA 0021 airfoil beyond stall), and a supersonic separated flow (wake behind a cylindrical forebody). The results of simulations suggest that the DDES and IDDES models combined with the shear-layer adapted subgrid length-scale perform according to their design (no unforeseen interactions of the shear-layer adapted length-scale with the empirical functions involved in the DDES and IDDES formulations are observed) and considerably mitigate the delay of transition from fully modeled to partially resolved turbulence in the separated shear layers compared to the standard DES definition of the length-scale (maximum local grid-spacing).     

基于RBF网络的有源噪声控制

提出一种神经自适应噪声有源控制（ANC）的方法。应用RBF（Radial Basis Function)网络对噪声进行有源控制。针对RBF的网络特点,使用递阶遗传算法确定网络参数（连接权、隐节点中心和宽度），同时解决了网络拓扑结构的优化训练。利用RBF网络的有源噪声控制系统用于三维空间传播的宽频带空调噪声的降噪获得了良好的效果。     

Reduced-Order Modeling of Subsurface Multi-phase Flow Models Using Deep Residual Recurrent Neural Networks

Transport in Porous Media - We present a reduced-order modeling technique for subsurface multi-phase flow problems building on the recently introduced deep residual recurrent neural network...     

基于动态神经元网络的激光陀螺输出误差模型

惯性敏感器误差补偿技术对提高武器装备的性能具有重要的意义,而误差补偿的关键在于误差模型的辨识。由于动态神经元网络是在前馈网络的节点引入前馈和反馈环节,理论上已证明其具有很强的动态逼近能力,可用来描述任意的非线性动态系统。根据惯性敏感器误差的动态特性,本探讨将动态神经元网络引入到激光陀螺误差建模中去,详细介绍了网络结构和对应的动态梯度算法。通过仿真算例说明,动态神经元网络在激光陀螺输出误差建模时具有一定的优点：网络收敛速度快、较好的跟踪性能、稳定性好。     

Compound Wall Treatment for RANS Computation of Complex Turbulent Flows and Heat Transfer

We present a generalised treatment of the wall boundary conditions for RANS computation of turbulent flows and heat transfer. The method blends the integration up to the wall (ItW) with the generalised wall functions (GWF) that include non-equilibrium effects. Wall boundary condition can thus be defined irrespective of whether the wall-nearest grid point lies within the viscous sublayer, in the buffer zone, or in the fully turbulent region. The computations with fine and coarse meshes of a steady and pulsating flow in a plane channel, in flow behind a backward-facing step and in a round impinging jet using the proposed compound wall treatment (CWT) are all in satisfactory agreement with the available experiments and DNS data. The method is recommended for computations of industrial flows in complex domains where it is difficult to generate a computational grid that will satisfy a priori either the ItW or WF prerequisites.     

Subgrid Scale Modeling in Large-Eddy Simulation of Turbulent Combustion Using Premixed Flamelet Chemistry

Large-eddy simulation (LES) of turbulent combustion with premixed flamelets is investigated in this paper. The approach solves the filtered Navier–Stokes equations supplemented with two transport equations, one for the mixture fraction and another for a progress variable. The LES premixed flamelet approach is tested for two flows: a premixed preheated Bunsen flame and a partially premixed diffusion flame (Sandia Flame D). In the first case, we compare the LES with a direct numerical simulation (DNS). Four non-trivial models for the chemical source term are considered for the Bunsen flame: the standard presumed beta-pdf model, and three new propositions (simpler than the beta-pdf model): the filtered flamelet model, the shift-filter model and the shift-inversion model. A priori and a posteriori tests are performed for these subgrid reaction models. In the present preheated Bunsen flame, the filtered flamelet model gives the best results in a priori tests. The LES tests for the Bunsen flame are limited to a case in which the filter width is only slightly larger than the flame thickness. According to the a posteriori tests the three models (beta-pdf, filtered flamelet and shift-inversion) show more or less the same results as the trivial model, in which subgrid reaction effects are ignored, while the shift-filter model leads to worse results. Since LES needs to resolve the large turbulent eddies, the LES filter width is bounded by a maximum. For the present Bunsen flame this means that the filter width should be of the order of the flame thickness or smaller. In this regime, the effects of subgrid reaction and subgrid flame wrinkling turn out to be quite modest. The LES-results of the second case (Sandia Flame D) are compared to experimental data. Satisfactory agreement is obtained for the main species. Comparison is made between different eddy-viscosity models for the subgrid turbulence, and the Smagorinsky eddy-viscosity is found to give worse results than eddy-viscosities that are not dominated by the mean shear. Paper presented on the Eccomas Thematic Conference Computational Combustion 2007, submitted for a special issue of Flow, Turbulence and Combustion.     

Subgrid length-scales for large-eddy simulation of stratified turbulence

The influence of buoyancy on the length-scales for the dissipation rate of kinetic energy, and for momentum, heat, and other scalar transport has to be known for subgrid-scale (SGS) models in a large-eddy simulation (LES). For the inertial subrange, Lilly (1967) has shown that grid spacing is the relevant length-scale for SGS effects. Deardorff (1980) proposed to reduce all the length-scales for stable stratification. Numerical and experimental data show, however, that the dissipation length-scale may strongly increase in stable layers with little shear. Lumley's (1964) theory for the energy spectrum in a stratified fluid also suggests such an increase. In this paper we apply the analysis of previous algebraic second-order closure SGS models, parameter studies with different length-scale models in LES, and the analysis of direct simulations of sheared and unsheared stably stratified homogeneous turbulence. These analyses show advantages of first-order closures for LES and suggest that the limiting effect of stratification should only be applied to the length-scales of vertical eddy-diffusivities of heat and scalars but not to those of momentum and dissipation.Dedicated to Professor J.L. Lumley on the occasion of his 60th birthday.This work was supported by the Deutsche Forschungsgemeinschaft.     

Mesh Node Distribution in Terms of Wall Distance for Large-eddy Simulation of Wall-bounded Flows

In this note, basic turbulent statistics in a pipe flow are computed accurately by large-eddy simulation using a mesh resolution coarser than the viscous sublayer. These results are obtained when a regular Cartesian mesh is used for the spatial discretization of the circular pipe thanks to an immersed boundary method combined with high-order schemes. In this particular computational configuration, the near-wall features of mean velocity and Reynolds stress profiles are found to be correctly captured at a scale significantly smaller than the mesh size. Comparisons between channel and pipe flow configurations suggest that an irregular mesh distribution in terms of wall distance may be a favourable condition to explicitly compute by large-eddy simulation reliable wall turbulence without any extra-modelling in the near-wall region.     

Nondestructive Evaluation of Crack Depth in Concrete Using PCA-compressed Wave Transmission Function and Neural Networks

Cracks in concrete are common defects that may enable rapid deterioration and failure of structures. Determination of a crack’s depth using surface wave transmission measurement and the cut-off frequency in the transmission function (TRF) is difficult, in part due to variability of the measurement data. In this study, use of complete TRF data as features for crack depth assessment is proposed. A principal component analysis (PCA) is employed to generate a basis for the measured TRFs for various simulated crack (notch) cases in concrete. The measured TRFs are represented by their projections onto the most significant PCs. Then neural networks (NN), using the PCA-compressed TRFs, are applied to estimate the crack depth. An experimental study is carried out for five different artificial crack (notch) cases to investigate the effectiveness of the proposed method. Results reveal that the proposed method can effectively estimate the artificial crack depth in concrete structures, even with incomplete NN training.     

Single-Pulse Solutions for Oscillatory Coupling Functions in Neural Networks

We study the existence and linear stability of stationary pulse solutions of an integro-differential equation modeling the coarse-grained averaged activity of a single layer of interconnected neurons. The neuronal connections considered feature lateral oscillations with an exponential rate of decay and variable period. We identify regions in the parameter space where solutions exhibit areas of excitation with single- and dimpled-pulses. When the gain function reduces to the Heaviside function, we establish existence of single-pulse solutions analytically. For a more general gain function, we include numerical support of the existence of pulse-like solutions. We then study the linear stability behavior of these solutions.     

Shape Optimization for 3D Turbulent Flows Using Automatic Differentiation

Abstract

We present through this paper some new results with our approach for optimal shape design based on a CAD-free framework for shape and unstructured mesh deformations, automatic differentiation by program for the gradient computation and mesh adaption by metric control. Automatic differentiation allows for an easy, reliable and fast discrete adjoint computation. We managed to get the Jacobian of our Navier-Stokcs solver including a k-e turbulence model and wall-laws. The CAD-free framework is shown to be particularly convenient for optimization when the mesh connectivities and the shape discretization arc variable during optimization. Using these ingredients constrained optimization for turbulent transonic flows has been investigated in both 2 and 3D.     

On Flows of Viscoelastic Fluids of Oldroyd Type with Wall Slip

We consider the boundary-value problem for the steady isothermal flow of an incompressible viscoelastic liquid of Oldroyd type in a bounded domain with a Navier type slip boundary condition. We prove that under some restrictions on the material constants and the data, there exists a strong solution which is locally unique. The proof is based on a fixed point argument in which the boundary-value problem is decomposed into a transport equation and a Stokes system.     

Sensitivity Analysis Using Adjoint Parabolized Stability Equations for Compressible Flows

An input/output framework is used to analyze the sensitivity of two- and three-dimensional disturbances in a compressible boundary layer for changes in wall and momentum forcing. The sensitivity is defined as the gradient of the kinetic disturbance energy at a given downstream position with respect to the forcing. The gradients are derived using the parabolized stability equations (PSE) and their adjoint (APSE). The adjoint equations are derived in a consistent way for a quasi-two-dimensional compressible flow in an orthogonal curvilinear coordinate system. The input/output framework provides a basis for optimal control studies. Analysis of two-dimensional boundary layers for Mach numbers between 0 and 1.2 show that wall and momentum forcing close to branch I of the neutral stability curve give the maximum magnitude of the gradient. Forcing at the wall gives the largest magnitude using the wall normal velocity component. In case of incompressible flow, the two-dimensional disturbances are the most sensitive ones to wall inhomogeneity. For compressible flow, the three-dimensional disturbances are the most sensitive ones. Further, it is shown that momentum forcing is most effectively done in the vicinity of the critical layer. This revised version was published online in July 2006 with corrections to the Cover Date.