基于大数据理论的大跨屋盖风荷载模型及数据仓库研究

项目名称： 基于大数据理论的大跨屋盖风荷载模型及数据仓库研究

项目编号： No.51478155

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 建筑环境与结构工程学科

项目作者： 孙瑛

作者单位： 哈尔滨工业大学

项目金额： 89万元

中文摘要： 大跨屋盖形体多样，绕流特性复杂，风荷载描述难以统一，这在一定程度上制约了其抗风设计理论的发展。通过对大量已有风荷载数据的统计挖掘，提取共性规律，建立普适风荷载模型是解决该问题的有效途径。而风荷载数据具有量大、维数高、格式复杂等特点，常规的数据统计方法难以整理。为此本项目拟引入大数据理论，力求在风荷载数据的挖掘、参数化建模和风荷载预测等方面有所突破。具体内容包括：研究风荷载数据有效信息筛选及信号降噪、高维数据挖掘算法等关键应用技术；对风荷载统计量与频谱的参数化建模、基于聚类算法的风荷载参数分区和基于多元统计学的风荷载参数敏感性分析等问题进行深入研究，提出风荷载参数化模型的统一表达形式；结合CFD数值模拟、神经网络预测和数据仓库技术，探讨数据集成标准，实现风荷载数据库拓展升级，建立大跨屋盖风荷载智能预测系统。希望通过本研究，形成较为完整的大跨屋盖风荷载知识体系，并为工程抗风设计提供有力支撑。

中文关键词： 大数据；大跨屋盖；风荷载模型；数据仓库；神经网络

英文摘要： Due to the diverse shapes of large-span roofs, the flow around the roofs is complicated. It is difficult to describe the wind load on large-span roofs uniformly. The development of wind resistance design theories on large span roofs have been faced with a great challenge. It is a effective way to solve the problem by collecting large amount of wind load data, discovering the common rules and establishing a universal wind load model. However, the wind load data obtained through various means are characterized as huge volume, high dimension and with complex formats, which cannot be easily summarized through regular statistical methodologies. Thus, the theory of big data is introduced, aiming at the breakthrough in studying the wind load characteristics, modeling and predicting the wind load data, etc. The content of this research project includes 1) Research on some key technologies for information screening, signal denoising and data mining algorithms will be carried out first; 2) The parameterization of wind load statistics and spectra, partition of wind load parameters based on clustering algorithm and parametric analyses based on multivariate statistics, finally a uniformed wind load model can be proposed; 3) Set up a wind load intelligent prediction system by normalizing the data integration standards, update the wind load database system, and combine Computational Fluid Dynamics (CFD) simulation, Artificial Neural Network (ANN) and Data Warehouse (DW) technologies to form a intelligence flat. The purpose of this research project is to form a better knowledge system on wind load for large-span roofs and provide a robust technological support for wind resistance design in actual projects.

英文关键词： big data;large-span roofs;wind loading model;data warehouse;Artifical Neural Network