小波变换在结构损伤识别中的应用研究

针对土木工程结构的复杂性和数据的不完备性，以小波分析为理论基础，结合多尺度分析理论和数值模拟，对结构损伤识别进行了研究。主要分析了结构动力参数在损伤后的变化特征，对单自由度框架的加速度信号进行小波分析，得到结构损伤的时刻；研究了利用空间域信息对结构损伤位置的识别问题，对带裂缝悬臂梁的变形曲线进行小波变换，判定了结构损伤位置。     

Possible remedies for intact stability hazards involving contemporary small inland passenger ferries in Bangladesh

On the basis of previous studies, a series of hazards involving the typical small inland passenger ferries operating in inland Bangladeshi waters have been identified. In particular, stormy weather conditions, overloading, and the risk of crowding to one side have been determined as the typical events that are likely to lead to capsizing in the intact condition. In this article, possible hazard mitigation measures are discussed, both from the regulatory and from the design point of view, for a small inland ferry that is very similar to one that actually capsized in the past. The addressed design options involve ballasting and hull modification by means of additional buoyancy above the waterline. From a regulatory point of view, it is proposed that the present weather criterion in force in Bangladesh be modified by increasing the wind speed to be accounted for and by considering the concurrent effects of wind, rolling, and the crowding of people to one side.     

电力电子与绿色能源

介绍了绿色能源的概念,以风力发电为重点介绍了电力电子技术在绿色能源发展中的应用以及分布式发电在绿色能源系统方面的应用.     

双探头直射检测预应力孔道病害技术

超声波双探头对测检测混凝土孔道缺陷,提出了声时系数概念,预应力孔道内不同填充物情况下声时、声程的变化.     

Identification of dynamic parameters in low-mobility mechanical systems: application to short long arm vehicle suspension

The identification of dynamic parameters in low-mobility mechanical systems is addressed and applied to short long arm (SLA) front car suspension. The main goal of the identification technique is to obtain, from experimental measurements, the values of those dynamic parameters (masses, location of the centre of masses, terms of the inertial matrix of the links, constant friction terms and elastic and viscous damping constant terms) that affect the dynamic behaviour of the system. Moreover, additional but important information that could be obtained from the procedure is related to the weight of those terms inside the dynamics of the system, so that simplified dynamic models based on relevant and well-identified parameters can be established. First, a systematic procedure will be presented for obtaining the equations of motion in a linear form with regard to the dynamic parameters to be identified. The main drawbacks related to the identification of parameters in low-mobility mechanisms will be pointed out, mainly the difficulty of determining a natural cutoff point into the singular values of the observation matrix which allows us to determine the true dimension of the set of base parameters and of obtaining an observation matrix well conditioned from the numerical point of view that allows an identification in the presence of measurement error. The procedure proposed for overcoming these problems will be based on the development of symbolic relationships among the physical parameters in order to determine the true rank of the observation matrix and on the consideration of a reduced subset of the base parameters set. These relevant parameters will be selected according to their influence on the dynamic behaviour of the mechanical system. A virtual benchmark will be used for testing purposes. The dynamic models based on relevant parameters show a better adjustment than the complete ones, mainly when the level of noise in the measurements used in the identification process increases.     

车速鉴定中轮胎-路面附着系数估算方法研究

在交通事故鉴定中,车辆行驶速度是事故处理和诉讼的重要依据。其中,路面附着系数是事故车速鉴定的重要参数。本文在大量实验数据基础上,拟合出车辆制动过程的特性曲线,并简化出相应的车速估算模型。利用此模型对不同车型、路面类型、湿滑状况和不同车速情况下的路面附着系数e进行了估算研究。经估算实例验证,文中的估算方法对不同状况下的路面附着系数具有较好的估算能力。     

环境变化下基于核典型相关分析与协整的损伤识别方法

采用传统协整方法进行损伤识别时，需要变量间满足较好的线性关系，而实际工程中监测变量往往存在一定程度的非线性，这使得协整方法的有效性受到影响。为此，提出一种结合核典型相关分析与协整的损伤识别方法。首先利用核典型相关分析能有效处理非线性相关变量的优点，将低维空间存在非线性关系的监测变量映射到高维空间，使其转化为线性相关的核典型变量。然后利用协整方法能够消除变量间共同趋势的特点，对核典型变量间的共同环境因素影响进行分离，并以分离环境因素影响后的协整残差作为损伤指标进行损伤识别。最后通过芬兰Kullaa课题组的木桁架桥试验数据，对协整方法、结合典型相关分析与协整的方法、核典型相关分析和协整相结合方法这3种方法的损伤识别结果进行比较。研究结果表明：在分析非线性数据方面，核典型相关分析要优于典型相关分析；前2种方法受监测变量数目的影响较大，选择不同数目的监测变量将得到不同的识别结果，而该方法对监测变量数目不敏感；且在损伤识别的漏判率方面该方法明显优于其他2种方法。     

重型载货汽车气压制动系统危险状态识别

针对重型载货汽车因气压制动系统发生管路破裂、机械故障或热衰退导致制动效能下降且不易察觉从而引发严重交通事故的问题，提出基于主成分分析降维（PCA降维）和马尔可夫模型的气压制动系统危险状态识别方法。考虑到三轴载货汽车双回路制动系统的结构复杂性以及制动过程制动踏板动作、系统压力建立和实现车辆减速具有明显的时序性特点，首先采用PCA降维的方法对系统状态进行辨识；然后运用驾驶人制动意图与制动系统响应的双层隐形马尔可夫模型对系统状态进行识别。受驾驶人习惯影响制动踏板作用瞬间辨识度低，采用混合高斯聚类法提取不同制动意图时制动保持阶段数据建立制动意图识别模型和系统响应识别模型，通过二者匹配程度判定系统状态。最后，分别依据实车试验数据对模型进行离线训练和在线辨识验证。试验结果表明：系统正常状态下，基于PCA降维和马尔可夫模型相结合的识别方法能够准确、有效地识别制动系统状态；制动管路断开压力降低状态下，PCA降维方法能够及时有效识别其危险状态。     

Identification of nonlinear systems using modified particle swarm optimisation: a hydraulic suspension system

This paper presents a novel modified particle swarm optimisation (MPSO) algorithm to identify nonlinear systems. The case of study is a hydraulic suspension system with a complicated nonlinear model. One of the main goals of system identification is to design a model-based controller such as a nonlinear controller using the feedback linearisation. Once the model is identified, the found parameters may be used to design or tune the controller. We introduce a novel mutation mechanism to enhance the global search ability and increase the convergence speed. The MPSO is used to find the optimum values of parameters by minimising the fitness function. The performance of MPSO is compared with genetic algorithm and alternative particle swarm optimisation algorithms in parameter identification. The presented comparisons confirm the superiority of MPSO algorithm in terms of the convergence speed and the accuracy without the premature convergence problem. Furthermore, MPSO is improved to detect any changes of system parameters, which can be used for designing an adaptive controller. Simulation results show the success of the proposed algorithm in tracking time-varying parameters.     

基于深度卷积神经网络的钢材微观组织分类识别

通过引入多种深度卷积神经网络及分类器构建机器学习模型,对钢材金相图数据集进行学习,研究了一种能够准确、高效识别钢材微观组织的方法。研究结果表明,文章中所涉及的3种深度卷积神经网络在钢材微观组织的分类识别上均表现出优异性能,其中Inception-V3表现尤为突出,其与人工神经网络分类器组合而成的机器学习模型的分类精度可达99.60%。