RFS-RF的局部非线性模型辨识新方法
A New RFS-RF Method for Local Nonlinear Model Identification
姜洋 1马砚秋 2陈榕 2刘景良 2张羲岭3
作者信息
- 1. 福建农业职业技术学院 园艺园林学院,福建 福州 350303
- 2. 福建农林大学 交通与土木工程学院,福建 福州 350108
- 3. 福建省国电调试院有限公司,福建 福州 350025
- 折叠
摘要
鉴于恢复力曲面法(Restoring Force Surface,RFS)和随机森林(Random Forest,RF)模型在参数辨识领域的优越性,结合上述两种方法提出一种新的基于RFS-RF的局部非线性模型辨识方法.首先,针对局部非线性模型求解其动力响应.其次,根据获得的动力响应计算恢复力曲面与边际谱,然后再通过边际谱求解非线性指标.再次,通过多次改变结构的刚度和阻尼参数生成若干组非线性指标并建立随机森林模型.然后,将新的非线性指标作为预测集输入已经建立的随机森林模型并判断系统的非线性类型和非线性函数形式.最后,采用最小二乘法对局部非线性系统的待求参数进行精确识别.通过一个四层剪切型框架结构模型对所提方法进行验证,研究结果表明:基于RFS-RF的多自由度局部非线性模型辨识方法能够准确识别结构系统的非线性类型、函数形式以及未知参数.
Abstract
In view of the superiority of the restoring force surface(RFS)method and random forest(RF)model,a new nonlinear model identification method called RFS-RF is proposed by combing the two methods mentioned above.In this method,the dynamic responses of local nonlinear model are solved at first.Second,the restoring force surfaces and marginal spectrums are calculated on a basis of the solved responses,and then the nonlinear indices are yielded by a use of marginal spectrums.Third,the RF model is established by changing the stiffness and damping parameters of the local nonlinear model gradually and hence several sets of nonlinear indices are gen-erated as well.After that,the generated nonlinear indices are used as inputs of the established RF model to predict nonlinear types and functions.Finally,the least square algorithm is applied to accurately estimate the parameters of the local nonlinear model.A numerical example of a four-story shear frame model is investigated to demonstrate the effectiveness of the proposed method and the result shows that the RFS-RF approach is capable of identifying nonlinear types,functions and unknown parameters very well.
关键词
恢复力曲面/随机森林/边际谱/模型辨识/局部非线性Key words
restoring force surface/random forest/marginal spectrum/model identification/local nonlinearity引用本文复制引用
基金项目
国家自然科学基金青年项目(51608122)
福建省自然科学基金项目(2020J01581)
中国博士后基金面上项目(2018M632561)
出版年
2024
NETL
NSTL
万方数据