基于人工智能的非高斯风压预测虚拟仿真实验教学  

Virtual Simulation Experiment Teaching of Non-Gaussian Wind Pressure Prediction Based on Artificial Intelligence

作　　者：李锦华[1] 邓羊晨 李涛 黄永虎 李春祥[2] LI Jinhua;DENG Yangchen;LI Tao;HUANG Yonghu;LI Chunxiang(School of Civil Engineering and Architecture,East China Jiaotong University,Nanchang 330013,China;School of Civil Engineering and Architecture,Shanghai University,Shanghai 200444,China)

机构地区：[1]华东交通大学土木建筑学院,南昌330013 [2]上海大学土木工程系,上海200444

出　　处：《实验室研究与探索》2024年第11期78-81,共4页Research and Exploration In Laboratory

基　　金：国家自然科学基金项目(11962006);江西省教育科学规划项目(22YB066);江西省学位与研究生教育教学改革研究项目(JXYJG-2021-113)。

摘　　要：利用有限的随机风场时程数据预测未知点位置的随机风场时程,以实现虚拟仿真实验教学的目的,可在一定程度上节省实验费用和资源,降低实验测试难度。利用Matlab仿真软件建立基于支持向量机(SVM)非高斯风压预测仿真方法。仿真结果表明,SVM核函数的选择对非高斯风压预测仿真影响较大,线性核函数模型对非高斯风压的预测仿真效果优于高斯核函数与指数核函数。基于SVM线性核函数模型能有效预测非高斯风压,为风洞试验或风场实测的虚拟仿真教学提供借鉴。Using limited random wind field time history data to predict the random wind field time history of unknown point positions can achieve the purpose of virtual simulation experiment teaching,and can save experimental costs and resources to a certain extent,reduce the difficulty of experimental testing.A non-Gaussian wind pressure prediction simulation method based on support vector machine(SVM)is established using Matlab in the paper.Simulation results indicate that the choice of kernel function in SVM significantly impacts the simulation performance of non-Gaussian wind pressure prediction.The linear kernel function model demonstrates better simulation effectiveness for non-Gaussian wind pressure prediction compared to Gaussian and exponential kernel functions.Therefore,the SVM linear kernel function model can effectively predict non-Gaussian wind pressures,providing valuable insights for virtual simulation experiments in wind tunnel tests or field measurements.

关 键 词：非高斯风压 虚拟仿真 教学改革 人工智能 支持向量机 预测 

分 类 号：U231.2[交通运输工程—道路与铁道工程] G642[文化科学—高等教育学]