【作者】 李磊；

【导师】 张纯；

【作者基本信息】 南昌大学 ， 防灾减灾工程及防护工程， 2018， 硕士

【摘要】 众所周知,结构在其服役期间,长期经受自然环境和人为等因素的影响,容易导致结构产生不同程度的损伤和破坏,从而有可能引发严重的安全性事故。为保障建筑结构的安全,发展结构损伤识别技术,提高结构健康监测和安全评估水平,对于保障工程结构的安全性、完整性和耐久性,预防工程事故发生具有重要的意义。粒子滤波是一种基于Monte Carlo方法和Bayes估计的递归式统计滤波的方法,不仅可以利用实时获取的量测值进行非线性系统的状态估计,而且突破传统Kalman滤波算法对高斯噪声的限制。由于粒子滤波在非线性非高斯系统中的优势,其适用范围正迅速扩大。本文对粒子滤波的结构损伤识别进行了系统的分析研究,针对粒子滤波的粒子退化和多样性匮乏以及结构损伤识别中反问题求解的强不适定性问题,提出了两种改进的粒子滤波损伤识别方法,并分别通过数值算例和框架结构的振动实验进一步验证了本文算法的有效性。本文主要研究内容如下:(1)针对粒子滤波运用于结构损伤识别中出现的如粒子退化、反演计算强不适定性等问题,笔者提出一种改进的粒子群优化粒子滤波损伤识别方法。在粒子滤波算法中,利用粒子群优化来驱动粒子群向后验概率密度取值较大的区域移动,并对粒子滤波的采样过程进行优化;同时,根据结构损伤参数分布的稀疏性特点,引入对粒子群中损伤参数部分的零变异操作。这既增加了粒子的多样性,又有效地缓解了反问题求解的不适定性问题,提高了算法损伤识别的鲁棒性。数值仿真和框架结构振动实验的结果均显示出,该方法能正确估计结构运动状态,准确识别结构中损伤位置与程度。(2)针对传统粒子滤波的重要性采样函数难以选取和粒子退化问题,利用UKF来获得重要性采样函数,从而将最新量测信息引入序贯重要性采样过程;同时,根据结构损伤参数分布的稀疏性特点,将L1范数正则化算法引入算法框架中,提出了一种改进的无迹粒子滤波算法。在确保粒子多样性的同时,改善了反问题求解的不适定性,并能有效地抑制噪声干扰,能准确识别结构的损伤位置与程度,具有良好的鲁棒性。(3)设计了一个五层铝框架结构的实验模型,对本文算法的有效性进行了实验验证分析。构造了多种结构损伤工况,通过布置在各层的传感器获取结构的动力响应信号,再利用本文算法进行损伤识别,通过实验证实了本文算法在不同损伤工况下都能准确估计结构运动状态,准确识别结构的损伤位置与损伤程度。更多还原

【Abstract】 During the period of structural service,long-term effects of natural environments and human factors will lead structures to damage,thus causing serious safety accidents.Structural damage identification is the core technology of structural health monitoring(SHM)and safety assessment.The research of structural damage identification methods can guarantee the safety,integrity and durability of the structure and prevent the occurrence of disasters.Particle filter(PF)is a recursive statistical filter based on the Monte Carlo method and Bayesian estimation.PF can estimate the state of nonlinear systems using measurement signals with gaussian or non-gaussian noise.Therefore,this method has been widely applied in various fields.This paper systematically analyzed particle filter in terms of structural damage identification.In order to solve the problems of particles degradation and the illposedness of inverse problem,two improved particle filter methods are proposed.Numerical simulations and vibration experiments of the frame structure verified their effectiveness.The main content of this article is as follows:(1)Based on particle swarm optimization(PSO)algorithm,an improved particle filter method is proposed to solve the problems of particles degeneracy,ill-posed characteristics and so on.These problems are common when the particle filter is applied to identify structural damages.The process of PSO is used to push particles to move toward the regions with higher posterior probability density,so the importance sampling process of particle filter is optimized.Furthermore,according to the sparseness of structural damage parameters distribution,the zero-mutation operation of damage parameters in particles is introduced to maintain the diversity of particles and improve the ill-posedness of inverse problem.Numerical simulations and shaking table test of frame structure show that the proposed method in this article can accurately identify the position and degree of structural damage.(2)Particle filter has difficulties in selecting the importance density function(IDF)and particles degradation.Thus,this paper uses UKF to obtain the IDF,that can introduce the latest measurement information into the sequential importance sampling(SIS)process.Due to the sparseness of structural damage parameters distribution,an improved unscented particle filter(UPF)method with L1-norm regularization algorithm was proposed.This method improves particles diversity,the ill-posedness of inverse problem and anti-noise performance.(3)An experimental model of frame structure is used to verify the effectiveness of the proposed method in this article.The structural dynamic response signals under different damage conditions are obtained through sensors on each layer.The experimental results show that the proposed algorithm can accurately estimate the structural motion state and accurately identify the structure damage location and damage degree.更多还原