基于改进的神经网络的大跨度屋盖风压场预测

发布时间：2018-01-27 04:09

本文关键词： 神经网络风压场方差分析大跨度屋盖变异性　出处：《应用力学学报》2014年06期 　论文类型：期刊论文

【摘要】：利用风洞实验测定大跨度屋盖的风压时,由于通常只能通过布置有限测点而获得有限的风压数据,因此采用准确高效的方法预测更多未知点的风压特性具有重要意义。本文在径向基函数神经网络的基础上,结合遗传算法和方差分析对大跨度屋盖的风压场进行了预测。引入整体度量估计作为遗传算法中适应度函数的估计,采用方差分析分解RBF的输出总方差,用求得的均方误差对RBF神经网络模型的权方差进行估计。将本文方法预测得到的大跨度屋盖的风压系数、升力系数等与实验结果和传统RBF方法预测结果进行了对比,并进行了误差统计分析。结果表明:本文方法预测的风压系数与试验结果符合良好,且均方误差小于RBF方法,平均相差约67.21%;在达到同样的收敛精度时,采用本文方法计算所需的时间要比传统RBF模型节省约40%。结合遗传算法和方差分析的RBF神经网络模型是预测大跨度屋盖风压场的有效工具。
[Abstract]:When the wind pressure of long-span roof is measured by wind tunnel experiment, the limited wind pressure data can only be obtained by arranging finite measuring points. Therefore, it is important to use accurate and efficient method to predict the wind pressure characteristics of more unknown points. This paper is based on the radial basis function neural network. Combined with genetic algorithm and ANOVA, the wind pressure field of long-span roof is predicted. The global metric estimation is introduced as the estimation of fitness function in genetic algorithm, and the total variance of RBF output is decomposed by ANOVA. The mean square error is used to estimate the weight variance of the RBF neural network model. The wind pressure coefficient of the long-span roof is predicted by the method in this paper. The lifting coefficient is compared with the experimental results and the traditional RBF method, and the error statistical analysis is carried out. The results show that the wind pressure coefficient predicted by this method is in good agreement with the test results. The mean square error is smaller than that of RBF method, and the average difference is about 67.21. With the same convergence accuracy. The computational time of this method is about 40% less than that of the traditional RBF model. The RBF neural network model combined with genetic algorithm and ANOVA is an effective tool for predicting the wind pressure field of large span roof.
【作者单位】：辽宁工程技术大学建筑工程学院;
【基金】：国家自然科学基金(51108345) 同济大学土木工程防灾国家重点实验室开放基金(SLDRCE-MB-04) 辽宁省教育厅基金(L2013134)
【分类号】：TU312.1
【正文快照】： 1引言近年来大跨度屋盖结构的工程应用越来越广泛,大跨度屋盖在其控制荷载——风荷载作用下风压特性的确定也成为工程研究人员关心的热点问题。现行的结构设计规范[1]对该类结构风压计算的规定并不完善。通常获得大跨度屋盖风压的最可靠方法是通过风洞实验获得风压数据,而对大

【参考文献】