大跨度煤棚主体结构设计风荷载的神经网络建模研究及应用

风荷载是大跨度煤棚结构设计中的主要控制荷载。随着结构抗风研究尤其是风洞试验数据的积累，结合数据挖掘进行结构智能化抗风设计是一种趋势。基于701组工况4581个柱面及球面屋盖风洞试验样本进行数据挖掘和统计分析，建立了脉动风荷载参数的广义回归神经网络预测模型；通过对12480个工况的单、双层柱面及球面网壳结构进行参数化风振响应分析，总结了等效静风荷载的经验表达式；建立了基于人工神经网络预测气动风荷载的主体结构等效静风荷载的抗风设计基本框架，并通过国内某超大跨度干煤棚张弦结构进行了有效性验证。结果表明：采用本文提出的风荷载数据库预测与等效静风荷载方法效率较高，且能够在一定程度包络风振响应分析结果，可用于结构初步设计阶段对主体结构设计风荷载快速预估。

Research and application on artificial neural network modeling of design wind load on large-span coal storage sheds

Wind load is the main control loads on the design of large-span coal storage sheds. With the researching accumulation of wind resistance of structures,especially with the accumulation of wind tunnel data, intellectual wind-resistant design using the concept of data mining gained increasing popularity. Based on 701 cases, 4581 samples of wind tunnel data of cylindrical and spherical roofs, data mining and statistical analyses were carried out, and the generalized regression neural network for the prediction of wind load parameters was established. With 12480 cases of parametric wind-induced response analyses on single- and double- layer cylindrical and spherical latticed shells, the empirical estimation method of the equivalent static wind load was proposed. Finally, a basic framework of wind-resistant design for the main load-bearing system was built by combining the artificial neural network for aerodynamic wind load and empirical prediction for equivalent static wind load. The effectiveness of the method was proved by the application on the prestressed latticed shell structure of a domestic super-large-span coal storage. It can be concluded that the presented method can provide an efficient estimation of the design wind load which can envelope the wind-induced response analysis results.The method can be used in quick estimating the design wind load in the preliminary design stage.