基于神经网络的无陀螺捷联惯导系统姿态预测

在无陀螺捷联惯导系统中,以工程问题为研究背景,针对以往解算载体角速度精度不高,导航误差随时间积累较快的问题,提出了基于信息融合理论的BP神经网络模型预测飞行体姿态的系统,并采用LM算法,提高学习速度.在研究了加速度传感器输出信号对飞行体姿态影响的基础上,将加速度计的输出信息作为输入变量,飞行体的实时三轴角速度作为目标信号建立网络模型.选取测试样本进行训练,得到较高精度的角速度输出,再运用四元数法解算姿态角,从一定程度上抑制了误差的积累.仿真结果表明该优化算法收敛速度快,对角速度的预测精度较高,并且合理选择及增加样本信息可以提高网络的泛化能力,为该系统走向工程实践提供了理论依据.     

基于模态数据的长细结构小损伤精确识别研究

本文针对现有的损伤识别方法不能满足部分结构损伤识别精度要求的现状，对结构的小损伤精确识别方法开展研究．以长细结构为研究对象，对具有不同损伤位置和损伤程度的圆柱形的轻阻尼梁结构进行了数值分析和实验研究，应用数值计算方法和实验确定的特征向量和特征频率对长细结构裂缝参数进行识别计算．本文在研究过程中编制了一个创新性的预测程序，通过其一次性生成目标函数图来选择合适的初始参数，从而对识别结果进行分析．研究结果表明，应用本文提出的识别方法，裂缝位置的识别误差可以控制在0.05 %~0.28 %范围内，裂缝深度识别误差低于7 %．     

基于灰色BP神经网络的陀螺电机状态预测

陀螺电机状态直接影响惯导系统的精度和可靠性,对其进行预测是惯导系统性能评估和寿命预测的重要途径。利用灰色理论的建模预测方法对随机性较大的数据预测精度不高;BP神经网络模型的预测方法具有良好的非线性和自学习能力,但训练效率不高且训练效果受样本数影响较大,网络容易限于局部最小值。针对陀螺电机状态特征参数的特点,本文提出一种基于灰色BP神经网络的混合模型。该模型利用BP神经网络对灰色模型误差进行建模,模型输出返回灰色模型进行输入修正。利用灰色理论、BP神经网络以及混合模型对状态特征参数进行建模和预测,结果表明,混合模型的预测误差比灰色模型减小了约2／3,比神经网络减小了约1／3,证明了该模型的有效性。     

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。     

基于转角模态和小波神经网络的连续梁损伤识别研究

基于小波奇异性检测原理和神经网络非线性映射能力,结合结构基本模态参数,提出了一种结合小波神经网络与结构转角模态的损伤识别方法．首先,建立三跨连续梁的有限元模型获取结构模态参数,并对其进行Mexihat小波变换,通过系数图突变点判断结构损伤位置．然后,将小波系数模特征向量作为BP神经网络的输入,分别研究了该方法在单损伤和多损伤工况下的识别能力．最后将不同工况下神经网络预测值与结构实际损伤程度进行对比,得到单处损伤预测误差平均值为0.22%,多处损伤预测误差平均值分别为0.22%和0.18%,结果表明该方法在结构损伤识别方面的有较高有效性及精确度．     

环境温度影响下基于支持向量机与强化飞蛾扑火优化算法的结构稀疏损伤识别

结构处于自然环境中常会受到环境温度变化的影响,引起实测动力响应出现较大误差,进一步影响对结构健康状况的判定.另外,基于优化算法的损伤识别在反演损伤位置及量化损伤程度时,易出现局部最优解,且计算效率低下.针对以上难题,本文提出一种结合支持向量机与强化飞蛾扑火优化算法的损伤识别方法,用于对环境温度影响下的结构稀疏损伤进行识...     

基于Elman神经网络和Kalman滤波的捷联惯导精对准

在采用Kalman滤波进行捷联惯导精对准时,当模型存在误差或系统噪声不能反映实际噪声时,会降低滤波精度甚至导致滤波发散.针对这个问题,提出基于Elman神经网络和Kalman滤波的捷联惯导精对准方法.首先对已知噪声统计特性的系统进行Kalman滤波,将稳定可靠的状态估值作为网络期望输出用来训练Elman网络,然后再用训练好的网络对未知噪声统计特性系统进行状态估计.利用仿真数据对该算法进行验证,结果表明该算法能够克服Kalman滤波精对准的缺陷,提高了对准精度,尤其是航向角的精度.     

基于EEMD和SVR的多自由度结构状态趋势预测

由于工程结构的复杂性和引起结构损伤原因的不确定性,结构早期微弱和潜在的损伤难以识别和预测。为此提出了基于聚类经验模式分解(EEMD)和支持向量机回归(SVR)的结构健康状态趋势预测方法。首先对多自由度结构渐进损伤的加速度振动信号进行聚类经验模式分解(EEMD);再进行希尔伯特变换(HT)计算瞬时频率;然后用回归支持向量机对反映结构健康状态的瞬时频率进行趋势预测。详细分析了各种参数对回归和预测精度的影响,提出了这些参数的选用方法和一般原则。研究表明:该方法具有训练样本少的特点;在采用二阶多项式核函数、回归步长m=3~5、误差惩罚因子C=100、敏感因子ε=0.01时,可以准确地和高精度地预测结构状态趋势,预测精度达到0.24781%。     

基于风洞试验大型冷却塔风致损伤识别研究

针对大型双曲冷却塔结构的损伤识别提出了一种基于经验模式分解(简称EMD)和神经网络技术的冷却塔气弹模型风致损伤识别方法。首先对风洞试验中采集的位移信号进行经验模式分解以获取多个固有模态函数(简称IMF);同时提取若干个包含主要损伤信息的IMF分量中的能量特征参数;然后以这些不同频段的能量特征参数作为神经网络的输入参数来识别冷却塔气弹模型的损伤程度和位置。对气弹模型预先设定的损伤位置和程度的分析结果表明:以EMD为预处理器提取各频带能量作为特征参数的神经网络诊断方法其平均识别误差为6%,可初步识别冷却塔结构中的风致损伤位置和程度。这为真实结构的损伤识别研究提供了新的思路。     

基于BP神经网络的风电机组钢混组合式塔架结构频率预测

风电机组塔架结构固有频率设计是风力发电结构体系设计的基础。针对风电机组新型钢混组合式塔架（“混塔”）结构固有频率传统理论计算和有限元法计算的不足,提出了基于BP神经网络算法进行频率预测的新方法。首先,利用有限元计算和分析,确定了训练模型的特征量和标签;然后,利用32个有限元计算样本,基于BP神经网络算法训练了可用于混塔结构频率分析的模型。经验证,该方法对混塔的一阶频率预测误差仅约为0.1%,具有很高的准确性;利用不同的样本集训练的模型也能快速准确预测混塔一阶频率,说明算法具有高度的稳定性;该方法还可用于预测混塔的多阶频率,结果仍显示出高度的准确性。此外,与基于有限元的频率计算相比,该方法具有突出的计算效率。整体上,本文提出的基于BP神经网络的混塔结构固有频率预测新方法,具有高度的可行性、精准性和高效性,可为风力发电机组塔架结构体系设计提供重要的指导。     

基于Mindlin板理论的偏移损伤成像数值仿真研究

提出了一种应用散射Lamb波的偏移技术对板结构中多部位损伤进行实时识别. 基于Mindlin板理论,推导了板结构中弥散性弯曲波频率-波数域的快速偏移方法. 首先对由线性传感器阵列激励和接收到的入射和散射波场在波数-频率域分别进行延拓,然后根据Huygens原理,结合波场延拓的时间一致性原理施加成像条件,对损伤进行成像识别. 数值仿真研究采用基于Mindlin板理论的有限差分法模拟结构中含不同形状及尺寸损伤时的散射波场. 对模拟散射波场进行偏移成像的结果表明该方法不仅能够识别多部位损伤的位置,还具有识别损伤程度的能力,其快速计算的优点满足在线结构健康监测系统对实时性的要求.      

Damage detection and denoising in two-dimensional structures using curvelet transform by wrapping method

The main objectives of this study are to present a vibration-based damage identification method and also a denoising mode shape approach applicable to two-dimensional structures using curvelet transform. For this purpose, the curvelet transform via wrapping method is employed. The reliability of the proposed technique is demonstrated through a verification study by comparing the results of numerical and those of the experimental data in plate structures. Two case studies, one-story and three-story shear walls assuming damages at arbitrary locations, are examined in which different noise levels are included. Good agreement between the simulated and assumed damage in both example is demonstrated. The results confirm the robustness and high performance of the proposed method in detecting the damage in plate structure and eliminating the noises.     

Identification of orthotropic damages within a thin uniform plate

The line crack-like damage generated within a small material volume may change the material behavior of the material volume from initially isotropic to effectively orthotropic, depending on damage orientation. Thus, the change in material behavior can be used to identify the orientation of line crack-like damage with respect to the reference coordinates. Motivated from this observation, first the equation of motion is derived for the thin uniform plate with line crack-like local damages. The locations and severities of damages are characterized by using a damage distribution function, and a damaged small material volume is represented by the effective orthotropic elastic stiffnesses, which are derived in terms of damage orientation and size. Next, a new damage identification theory is developed to identify the orientations of local damages, in addition to their locations and severities, by using the frequency response functions measured from the damaged plate. Finally, the effects of damage orientation on the vibration responses of a plate are numerically investigated, and the numerically simulated damage identification tests are conducted to verify the present damage identification theory.     

Structural damage localization and evaluation based on modal data via a new evolutionary algorithm

This paper is intended to present a method for the localization and evaluation of damage in plates based on the changes in natural frequencies and mode shapes of the damaged plate using an optimization approach. The colonial competitive algorithm is employed to detect damage (or damages) in plates by optimizing a damage function. The performance of the proposed method is demonstrated by implementing the technique to two examples; a shear wall and a four-fixed supported plate with and without modal data noise including one or a large number of damages. The results confirm the applicability and efficiency of the presented method in detecting damage localization and quantification in the shear walls. Furthermore, the proposed method is implemented to the four-fixed supported plate aimed at demonstrating the high sensitivity of the proposed method in quantitative estimation of damaged plate structures. Finally, the reliability of the presented method is explored through the comparison of the obtained results and those of the other methods. It is concluded that the proposed method can be viewed as a powerful and robust method for structural damage detection in plate structures.     

Sequential non-linear least-square estimation for damage identification of structures with unknown inputs and unknown outputs

The detection of structural damages real-time on-line, based on vibration data measured from sensors, is an important but challenging research topic, and it has received considerable attentions recently. Due to practical limitations, it is highly desirable to install as few sensors as possible in the structural health monitoring system, leading to incomplete measurements of structural responses and excitations. The traditional time-domain analysis techniques, such as the least-square estimation (LSE) method and the extended Kalman filter (EKF) approach, require that all the external excitations (inputs) be available, which may not be the case for most structural health monitoring systems. Recently, the adaptive sequential non-linear least-square estimate (SNLSE) method has been proposed for the on-line identification of structural damages. In this paper, we extend the SNLSE method to cover the general case with unknown (unmeasured) excitations (inputs) and unknown (unmeasured) acceleration responses (outputs) in order to reduce the number of sensors required in the structural health monitoring system, referred to as the SNLSE-UI-UO. Analytic recursive solutions for the new approach are derived and presented. The accuracy and effectiveness of the proposed approach have been demonstrated using the Phase I ASCE structural health monitoring benchmark building, a 5-degree-of-freedom non-linear hysteretic building model, and a 3-story steel frame finite-element model. Simulation results indicate that the proposed approach is capable of tracking the changes of structural parameters leading to the identification of damages.     

Sequential non-linear least-square estimation for damage identification of structures

An early detection of structural damage is an important goal of any structural health monitoring system. In particular, the ability to detect damages on-line, based on vibration data measured from sensors, will ensure the reliability and safety of the structures. In this connection, innovative data analysis techniques for the on-line damage detection of structures have received considerable attentions recently, although the problem is quite challenging. In this paper, we proposed a new data analysis method, referred to as the sequential non-linear least-square (SNLSE) approach, for the on-line identification of structural parameters. This new approach has significant advantages over the extended Kalman filter (EKF) approach in terms of the stability and convergence of the solution as well as the computational efforts involved. Further, an adaptive tracking technique recently proposed has been implemented in the proposed SNLSE to identify the time-varying system parameters of the structure. The accuracy and effectiveness of the proposed approach have been demonstrated using the Phase I ASCE structural health monitoring benchmark building, a non-linear elastic structure and non-linear hysteretic structures. Simulation results indicate that the proposed approach is capable of tracking on-line the changes of structural parameters leading to the identification of structural damages.     

固有模态函数振动传递率损伤识别实验研究

针对风荷载、地震荷载等存在但难以精确量测的问题,提出一种无需量测外荷载的新的损伤识别方法。将经验模态分解(Empirical Mode Decomposition,EMD)应用于结构损伤识别,通过求振动响应信号固有模态函数(Intrinsic Mode Function,IMF)的振动传递率,构建了一种新的结构损伤识别参数。对预置不同程度开胶损伤的玻璃幕墙试件进行动态测试,得到不同损伤程度下玻璃幕墙的固有模态函数振动传递率,与无损伤条件下的固有模态函数振动传递率进行比较来识别和评估玻璃幕墙开胶损伤程度。研究表明:此方法无需量测外荷载也能识别结构损伤,损伤参数值能反映损伤大小。     

Solution of the moiré hole drilling method using a finite-element-method-based approach

The moiré hole drilling method in a biaxially loaded infinite plate in plane stress is an inverse problem that exhibits a dual nature: the first problem results from first drilling the circular hole and then applying the biaxial loads, while the other problem arises from doing the opposite, i.e., first applying the biaxial load and then drilling the circular hole. The first problem is hardly ever addressed in the literature but implies that either separation of stresses or material property identification may be achieved from interpreting the moiré signature around the hole. The second is the well-known problem of determination of residual stresses from interpreting the moiré fringe orders around the hole. This paper addresses these inverse problem solutions using the finite element method as the means to model the plate with a hole, rather than the typical approach using the Kirsch solution, and a least-squares optimization approach to resolve for the quantities of interest. To test the viability of the proposed method three numerical simulations and one experimental result in a finite width plate are used to illustrate the techniques. The results are found to be in excellent agreement. The simulations employ noisy data to test the robustness of this approach. The finite-element-method-based inverse problem approach employed in this paper has the potential for use in applications where the specimen shape and boundary conditions do not conform to symmetric or well-used shapes. Also, it is a first step in testing similar procedures in three-dimensional samples to assess the residual stresses in materials.     

复合材料构件损伤类型识别的一种方法

复合材料构件在使用过程中会产生脱层、内部裂纹、裂缝等损伤，这些损伤会引起构件动态特性的变化。本文提出了通过测取构件的动态特性，结合波形分析和模式识别技术进行复合材料损伤检测和损伤类型识别的方法。在比较了飞机环境噪声信号和周期脉冲信号的优缺点的基础上，提出使用周期脉冲信号作为构件的激振信号，来进行构件动态特性的测量。本文的研究中制作了多种类型的损伤试件，进行了实验，已发现显著特征，能够对构件的多种类型损伤进行识别