摘要
针对非线性动力时程分析法求解大规模索膜结构风振响应时动力时程分析的次数受到限制而导致一些参数组合下的响应统计值难以预测的问题,引入神经网络,通过少量样本的训练,建立了参数与结构响应间的映射关系。结果表明:提出的神经网络辅助参数分析方法计算效率高、预测精度令人满意,是一种获取足够数据的有效途径;通过该方法可以得到响应统计量及风振系数随平均风速和索、膜预应力变化的规律,为设计风荷载和结构构件极端响应的计算提供了科学依据。
The nonlinear dynamic time history analysis method was an effective approach to solve the wind-induced response for cable-membrane structures. If the scale of the problem was large, the number of times of the nonlinear dynamic time history analysis would be confined, and the statistics of structural response obtained for parametric analysis were too limited. As a result, the statistics for some parametric combinations were difficult to be predicted. To solve this problem, neural network was used. The idea was that the relation between the parameters and structural responses could be established after training with the samples. The results show that the proposed neural network aided parametric analysis method is efficient and accurate. Therefore, the proposed method is an effective access to sufficient data. The rules of variation of response statistics and dynamic coefficients with mean wind velocity and with the structural prestresses are obtained, which provide a scientific basis for calculations of design wind loads and extreme response of structural members.
出处
《建筑科学与工程学报》
CAS
2008年第4期98-104,共7页
Journal of Architecture and Civil Engineering
基金
广州市科技攻关引导项目(2004Z3-E0351)
关键词
索膜结构
风振响应
参数分析
神经网络
动力时程分析
cable-membrane structure
wind-induced response
parametric analysis
neural network
dynamic time history analysis
