摘要
对于薄壁弯箱结构,推导了材料常数的动态Bayes误差函数,提出步长的一维Fibonacci序列自动寻优方案后,利用Powell优化理论研究了薄壁弯箱材料常数的动态识别方法,同时给出了具体的计算步骤,并研制了相应的计算程序.算例分析表明,Powell理论用于弯箱材料常数识别时表现出良好的数值稳定性和收敛性,在迭代过程中,Powell理论不涉及有限元偏导数处理,与以往材料常数的梯度优化方法相比,计算效率较高;建立的动态Bayes误差函数能同时计入系统参数的随机性和系统响应的随机性;提出的Fibonacci序列寻优方案无需通过试算确定最优步长所在区间,有效地解决最优步长的一维自动寻优问题.
For thin-walled curve box girders,dynamic Bayesian error function of material constants of the structure was founded. Combined with one-dimensional Fibonacci series automatic search scheme of optimal step length,the Powell’s optimization theory was utilized to perform the stochastic identification of material constants of thin-walled curve box. Then the steps of parameters’identification were presented in detail and the Powell’s identification procedure of material constants of thin-walled curve box was compiled,in which the mechanical analysis of thin-walled curve box was completed based on finite curve strip element（ FCSE） method. Through some classic examples,it is obtained that the Powell’s identification of material constants of thin-walled curve box has numerical stability and convergence,which demonstrates that the present method and the compiled procedure are correct and reliable. And during parameters’iterative processes,the Powell’s theory is irrelevant with the calculation of FCSE’s partial differentiation, which proves high computation efficiency of the studied methods. The stochastic performances of systematic parameters and systematic responses are simultaneously deliberated in dynamic Bayesian error function. The one-dimensional optimization problem of the optimal step length is solved by adopting Fibonacci series search method and there is no need to determine the region in which the optimized step length lies.
出处
《应用数学和力学》
CSCD
北大核心
2011年第1期93-102,共10页
Applied Mathematics and Mechanics
基金
国家自然科学基金资助项目(10472045
10772078
11072108)
国家高技术研究发展计划(863)资助项目(2007AA11Z106)
作者简介
张剑（1978-），男，安徽青阳人，博士（联系人．Tel：＋86-25-83713137；E—mail：zjmeeh@163．com）．
