基于物理信息的深度学习求解矩形薄板力学正反问题

建立基于物理信息的神经网络框架,利用深度学习求解矩形薄板力学正反问题.力学正问题为已知矩形薄板的基本参数、边界条件和受力情况,求薄板各点挠度;反问题为已知薄板部分点的挠度、基本参数和受力情况等,识别边界条件.基于物理信息的神经网络模型中,损失函数除基于数据驱动模型的挠度数据拟合部分以外,还引入薄板弯曲基本方程和应力应变...     

Modelling unilateral damage effect in strongly anisotropic materials by the introduction of the loading mode in damage mechanics

Damage weakens the mechanical characteristics of materials. But this weakening can disappear if the cracks close again: this is called the unilateral effect of damage. We propose a model of this phenomenon using damage mechanics in the case of a diffuse network of identical microcracks. The microcracks state is defined with two internal state variables. These two variables are control parameters of the geometry of the microcracks. They also define the loading mode in damage mechanics as in fracture mechanics. In order to limit the anisotropy induced by the microcracks, hence by the loading, we suppose that the geometry of damage spreads into preferential directions. Therefore, this model is essentially applied to materials with a strong anisotropy where the defects follow the constituents of the material. An application is given for composite laminates.     

统计细观损伤力学和损伤演化诱致突变（Ⅱ）

材料的损伤和破坏是固体力学中一个十分复杂的基本问题，近年来在这方面的研究从不同的角度取得一系列新的进展。本文较系统地介绍了统计细观损伤力学和损伤演化诱致突变理论，这是把力学与统计物理学及非线性科学结合起来研究材料损伤和破坏问题的一种探索。     

统计细观损伤力学和损伤演化诱致突变（Ⅱ）

材料的损伤和破坏是固体力学中一个十分复杂的基本问题，近年来在这方面的研究从不同的角度取得一系列新的进展。本文较系统地介绍了统计细观损伤力学和损伤演化诱致突变理论，这是把力学与统计物理学及非线性科学结合起来研究材料损伤和破坏问题的一种探索。     

统计细观损伤力学和损伤演化诱致突变

材料的损伤和破坏是固体力学中一个十分复杂的基本问题，近年来在这方面的研究从不同的角度取得一系列新的进展。本文较系统地介绍了统计细观损伤力学和损伤演化诱致突变理论，这是把力学与统计物理学及非线性科学结合起来研究材料损伤和破坏问题的一种探索。     

统计细观损伤力学和损伤演化诱致突变

材料的损伤和破坏是固体力学中一个十分复杂的基本问题,近年来在这方面的研究从不同的角度取得一系列新的进展。本文较系统地介绍了统计细观损伤力学和损伤演化诱致突变理论,这是把力学与统计物理学及非线性科学结合起来研究材料损伤和破坏问题的一种探索。      

基于爆炸切割试验的有机玻璃本构模型参数反演

为了获取爆炸切割数值模拟中有机玻璃（PMMA）的材料本构模型参数,建立了一种基于神经网络的有机玻璃Johnson Holmquist ceramics (JH-2)本构模型参数反演方法：基于从爆炸切割试验和现有研究得到的JH-2本构模型经验参数,确定本构模型参数的调整区间;使用LS-DYNA数值模拟软件对2.5 mm宽爆炸切割索切割14 mm PMMA平板过程进行数值模拟并收集平板损伤数据集;建立PMMA平板本构模型参数与损伤数据之间的神经网络模型;通过训练完成的神经网络模型对PMMA平板的JH-2本构模型参数进行反演。为验证通过反演参数的可靠性,进行了4.2 mm宽爆炸切割索切割19 mm PMMA平板试验和有限元数值模拟,计算结果中的平板损伤情况与实验结果相差较小,表明通过反演获得的JH-2本构模型参数能较好地应用于PMMA平板爆炸切割数值模拟。传统材料参数获取方法,该参数反演方法相较于可以通过较少的试验及测试,获得比较准确的材料本构模型参数。     

复合材料层合板面内渐进损伤分析的CDM模型

基于连续介质损伤力学,提出了一个预测复合材料层合板面内渐进损伤分析的模型,它包括损伤表征、损伤判定和损伤演化3 部分. 模型能够区分纤维拉伸断裂、纤维压缩断裂、纤维间拉伸损伤和纤维间压缩损伤4 种损伤模式,定义了与4 个损伤模式对应的损伤状态变量,导出了材料主轴系下损伤前后材料本构之间的关系. 损伤起始采用Puck 准则判定,损伤演化由特征长度内应变能释放密度控制. 假定材料服从线性应变软化行为,建立了损伤状态变量关于断裂面上等效应变的渐进损伤演化法则. 模型涵盖了复合材料面内损伤起始、演化直至最终失效的全过程. 完成了含孔[45/0/-45/90]_2S 层合板在拉伸和压缩载荷下失效分析,结果表明该模型能合理进行层合板的强度预测和损伤失效分析.      

2021-03期目录

针对毁伤试验数据少、不均匀、不连续、范围窄等带来的计算精度不高的问题。研究通过数据挖掘技术进行毁伤效应计算。利用数据库管理毁伤数据,通过数据清洗技术识别并清除数据异常点,以保证数据库中数据的质量。建立了算法评价方法以选择最优经验算法。通过特征选择对高维毁伤数据进行降维,确定毁伤效应的主要控制参数进行神经网络学习和k-近邻检索。在此基础上建立基于数据融合的“三阶段”毁伤效应计算模型,可依据试验数据、经验算法和神经网络模型进行毁伤效应计算。实际应用表明,所提出的计算方法,能够满足实际应用需求。     

Statistical aspects of the identification of material parameters for elasto-plastic models

Summary The presented method to identify material parameters for inelastic deformation laws is based on the numerical analysis of inhomogeneous stress and strain fields received from suitable experiments. Tensile and bending tests were carried out to obtain elastic and hardening parameters. The deformation law for small elasto-plastic strains is presented as a system of nonlinear differential and algebraic equations (DAE) consisting of the stress–strain relation, evolution equations for the internal variables and the yield condition. Different rules for the evolution equations of isotropic, kinematic and distorsional hardening are proposed. The DAE are discretized using an implicit Euler method, and the resulting system of nonlinear algebraic equations is solved using the Newton method. Deterministic optimization procedures are preferred to identify material parameters from a least-squares functional of numerical and measured comparative quantities. The gradient of the objective function was calculated using a semianalytical sensitivity analysis. Due to measurement errors, the optimal sets of material parameters are non unique. The approximate estimation of confidence regions and the calculation of correlation coefficients is presented. The results of several optimization processes for material parameters of elasto-plastic deformation laws show a good agreement between measured and calculated values, but they show also problems which may occur if systematic errors will not be recognized and deleted. Received 30 September 1999; accepted for publication 8 March 2000     

延性金属层裂模型比较

在平面一维弹塑性流动有限差分计算程序中加入4种延性金属层裂模型，对平板撞击层裂实验进行数值模拟。结果表明:简单最大拉伸应力模型和简单损伤累积模型能定性反映层裂的物理现象，由于忽略损伤对本构的影响，计算结果和实验有偏差，但模型要求参数较少，对于一些精度要求不是很高的工程问题，可以采用；从材料损伤断裂物理本质出发，采用微损伤统计方法得到的NAG模型和封加波损伤度函数模型，能很好地再现实测的自由面速度剖面,数值计算结果与实验吻合很好。     

复合材料层合板面内渐进损伤分析的CDM模型简

基于连续介质损伤力学,提出了一个预测复合材料层合板面内渐进损伤分析的模型,它包括损伤表征、损伤判定和损伤演化3 部分. 模型能够区分纤维拉伸断裂、纤维压缩断裂、纤维间拉伸损伤和纤维间压缩损伤4 种损伤模式,定义了与4 个损伤模式对应的损伤状态变量,导出了材料主轴系下损伤前后材料本构之间的关系. 损伤起始采用Puck 准则判定,损伤演化由特征长度内应变能释放密度控制. 假定材料服从线性应变软化行为,建立了损伤状态变量关于断裂面上等效应变的渐进损伤演化法则. 模型涵盖了复合材料面内损伤起始、演化直至最终失效的全过程. 完成了含孔[45/0/-45/90]_2S 层合板在拉伸和压缩载荷下失效分析,结果表明该模型能合理进行层合板的强度预测和损伤失效分析.     

Identifying Multidimensional Damage in a Hierarchical Dynamical System

In this paper, we present a novel method for multidimensional damage identification based on a dynamical systems approach to damage evolution. This approach does not depend on the knowledge of particular damage physics, and is appropriate for systems where damage evolves on much slower time scale than the directly observable dynamics. In an experimental context, the phase space reconstruction and locally linear models are used to quantify small distortions occurring in a dynamical system's phase space due to damage accumulation. These measurements are then related to the drifts in damage variables. A mathematical model of a harmonically driven cantilever beam in a force field of two battery-powered electromagnets is used to demonstrate validity of the method. It is explicitly demonstrated that an affine projection of the described feature vector accurately tracks the two competing damage processes. For practical damage identification purposes, the tracking data is analyzed using the proper orthogonal decomposition (POD) and smooth orthogonal decomposition (SOD) methods. Both methods correctly identify the two dominant damage modes. However, the SOD is more impervious to changes in fast-time dynamics and provides a significantly better signal-to-noise ratio. The damage modes identified using SOD are demonstrated to be within a linear transformation from the actual damage states and can be used to reconstruct the corresponding phase space trajectory.     

考虑晶体滑移面分解正应力的细观损伤模型

材料内部的解理、滑移面剥离等细观损伤是引起宏观失效的根源, 从细观尺度研究损伤的发生和发展有助于深入认识材料的变形和失效过程. 本文基于晶体塑性理论, 从滑移系的受力和变形出发研究材料的细观损伤, 建立了考虑滑移面分解正应力的细观损伤模型, 为晶体材料解理断裂的分析提供了新方法. 首先, 在晶体弹塑性变形构型的基础上引入损伤变形梯度张量的概念, 从变形运动学着手建立了考虑损伤能量耗散的本构方程, 并推导了塑性流动方程与损伤演化方程; 然后, 建立了相应的数值计算方法, 给出了应力与状态变量的更新算法, 推导了Jacobian矩阵的表达式; 接着, 以$[100]$取向的单晶铜材料为例, 通过有限元计算与试验结果的对比, 并采用粒子群优化算法标定了11个材料细观参数; 最后, 将所提细观损伤模型应用于RVE单轴拉伸过程的模拟, 得到了考虑损伤影响的应力应变曲线, 并分析了材料的塑性流动与损伤演化过程. 结果表明, 本文所提模型能够计算材料在受载过程中的损伤累积效应, 合理反映晶体材料的细观损伤机理.      

利用振动模态测量值和神经网络方法的结构损伤识别研究

提出了一种基于模态测量参数和神经网络的结构损伤检测方法,建造了两种输入方式的BP神经网络,即自振频率以及结合自振频率与振型,并讨论了不同数量的输入信息对结构损伤检测精度和计算效率的影响。证明了输入的参数越多,神经网络就越聪明,训练的收敛速度越快;以及在保证一定的测量精度的情况下,基于频率与振型的损伤识别结果要好于基于频率的检测结果。最后,通过对3层框架模型的4种损伤工况下的结构损伤检测结果的分析,认为利用模态测量参数和神经网络方法能够准确地识别结构损伤的位置,而且能较精确地识别结构损伤的大小。     

A constitutive model dealing with damage due to cavity growth and the Mullins effect in rubber-like materials under triaxial loading

In this work, we attempted to describe the evolution of damage in rubber-like materials due to the Mullins effect and the cavity growth process. To this end we introduced two distinct internal variables into the constitutive laws; the first one essentially describes the Mullins damage and the second describes the cavity growth. The Mullins effect was considered as a continuous type of damage that can be modelled within the continuum damage theory. The cavity growth, being discontinuous at the microscopic scale, was also modelled by a continuous variable after a homogenization procedure. These analyses allow the establishment of a compressible constitutive law describing the strain-softening phenomena for rubber-like materials. In order to identify the material parameters and to verify the efficiency of the model, we carried out experimental studies involving uniaxial, biaxial, and hydrostatic tensions under monotonic and cyclic loading. Comparison between the model-predicted results and the experimental data shows that the present model can efficiently describe both the Mullins damage and the porosity evolution of rubber-like materials under triaxial monotonic or cyclic loading with a satisfactory accuracy. The proposed concept is simple and easy to apply to engineering calculations.     

A NEW DAMAGE MECHANICS BASED APPROACH TO FATIGUE LIFE PREDICTION AND ITS ENGINEERING APPLICATION

An approach based on continuum damage mechanics to fatigue life prediction for structures is proposed. A new fatigue damage evolution equation is developed, in which the parameters are obtained in a simple way with reference to the experimental results of fatigue tests on standard specimens. With the utilization of APDL language on the ANSYS platform, a finite element implementation is presented to perform coupling operation on damage evolution of material and stress redistribution. The fatigue lives of some notched specimens and a Pitch-change-link are predicted by using the above approach. The calculated results are validated with experimental data.     

A cohesive zone model for fatigue crack growth allowing for crack retardation

A damage-based cohesive model is developed for simulating crack growth due to fatigue loading. The cohesive model follows a linear damage-dependent traction–separation relation coupled with a damage evolution equation. The rate of damage evolution is characterized by three material parameters corresponding to common features of fatigue behavior captured by the model, namely, damage accumulation, crack retardation and stress threshold. Good agreement is obtained between finite element solutions using the model and fatigue test results for an aluminum alloy under different load ratios and for the overload effect on ductile 316 L steel.     

Analytical and numerical techniques to predict carbon nanotubes properties

In this paper, two different approaches for modeling the behaviour of carbon nanotubes are presented. The first method models carbon nanotubes as an inhomogeneous cylindrical network shell using the asymptotic homogenization method. Explicit formulae are derived representing Young’s and shear moduli of single-walled nanotubes in terms of pertinent material and geometric parameters. As an example, assuming certain values for these parameters, the Young’s modulus was found to be 1.71 TPa, while the shear modulus was 0.32 TPa. The second method is based on finite element models. The inter-atomic interactions due to covalent and non-covalent bonds are replaced by beam and spring elements, respectively, in the structural model. Correlations between classical molecular mechanics and structural mechanics are used to effectively model the physics governing the nanotubes. Finite element models are developed for single-, double- and multi-walled carbon nanotubes. The deformations from the finite element simulations are subsequently used to predict the elastic and shear moduli of the nanotubes. The variation of mechanical properties with tube diameter is investigated for both zig-zag and armchair configurations. Furthermore, the dependence of mechanical properties on the number of nanotubules in multi-walled structures is also examined. Based on the finite element model, the value for the elastic modulus varied from 0.9 to 1.05 TPa for single and 1.32 to 1.58 TPa for double/multi-walled nanotubes. The shear modulus was found to vary from 0.14 to 0.47 TPa for single-walled nanotubes and 0.37 to 0.62 for double/multi-walled nanotubes.