Bouc-Wen模型参数在线识别的粒子滤波器算法  

Online identification method of Bouc-Wen model parameters based on particle filter

作　　者：王涛 刘媛[1] 孟丽岩[1] 

机构地区：[1]黑龙江科技大学建筑工程学院,哈尔滨150022 [2]中国地震局工程力学研究所中国地震局地震工程与工程振动重点实验室,哈尔滨150080 [3]同济大学土木工程防灾国家重点实验室,上海200092

出　　处：《黑龙江科技大学学报》2018年第1期102-106,共5页Journal of Heilongjiang University of Science And Technology

基　　金：国家自然科学基金项目(51408157);黑龙江省省属高等学校基本科研业务费基础研究项目(2017)

摘　　要：针对Bouc-Wen模型在结构数值模拟时难以事先确定参数值的问题,采用粒子滤波器在线识别单自由度Bouc-Wen系统模型参数,并对算法识别精度及计算效率进行敏感性分析。结果表明:粒子滤波器具有良好的参数识别效果,随着粒子数量的增多,参数识别精度提高,单步计算耗时近似呈指数增长。当粒子数量大于350时,识别精度提高幅度明显变缓。该研究为粒子滤波器在土木工程中应用提供了参考。This paper is a response to the difficulty facing the Bouc-Wen model used to determine in advance model parameter values in structure numerical simulation. The research involves online identifying parameters of a single degree of freedom Bouc-Wen system using the particle filter（ PF）; and analyzing sensitivities of parameter identification precision and calculation efficiency of the PF algorithm with respect to particle numbers. Results show that the PF algorithm has a better recognition effect; a greater parameter identification accuracy,depending on an increasing particle number; a single calculation step with time-consumption approximate to an exponential growth and above all,an obviously slow growth rate of the identification accuracy when the particle number is greater than 350. The study may provide references for the application of the PF algorithm in civil engineering.

关 键 词：粒子滤波器 参数识别 BOUC-WEN模型 敏感性分析 粒子数量 

分 类 号：TU352[建筑科学—结构工程] TN713[电子电信—电路与系统]