基于3DCNN的陆上风机基础竖向位移预测研究  

PREDICTION OF VERTICAL DISPLACEMENT OF ONSHORE WIND TURBINE FOUNDATION BASED ON 3DCNN

作　　者：李仁杰[1] 张伟[1] 卢向星 刘中华 魏焕卫[2,3] 谭芳 LI Renjie;ZHANG Wei;LU Xiangxing;LIU Zhonghua;WEI Huanwei;TAN Fang(Shandong Electric Power Engineering Consulting Institute Corporation Limited,Jinan 250013,China;School of Civil Engineering,Shandong Jianzhu University,Jinan 250101,China;Key Laboratory of Building Structural Retrofitting and Underground Space Engineering(Shandong Jianzhu University),Ministry of Education,Jinan 250101,China)

机构地区：[1]山东电力工程咨询院有限公司,济南250013 [2]山东建筑大学土木工程学院,济南250101 [3]山东建筑大学建筑结构加固改造与地下空间工程教育部重点实验室,济南250101

出　　处：《力学与实践》2025年第1期98-106,共9页Mechanics in Engineering

基　　金：山东省自然科学基金项目(ZR2019BEE076);山东建筑大学博士基金项目(X19024Z)资助。

摘　　要：为准确预测风机基础的沉降,避免风机基础不均匀沉降过大导致风机结构变形,影响安全及寿命,依托某陆上风机基础加固项目,构建了一种基于三维卷积神经网络(three-dimensional convolutional neural network,3DCNN)的风机基础竖向位移预测模型,对不同位置测点的竖向位移监测数据进行时空重构,并通过时空矩阵将监测数据导入三维卷积核学习数据间的时空特征。对比当下热门的神经网络模型,可以发现3DCNN模型在沉降预测方面具有更高的预测精度,其泛化性和稳定性也更优越。In order to accurately predict the settlement of the wind turbine foundation and avoid the deformation of the wind turbine structure due to the excessive uneven settlement of the wind turbine foundation,which affects the safety and life span of the wind turbine,a vertical displacement prediction model of the wind turbine foundation based on a three-dimensional convolutional neural network(three-dimensional convolutional neural network,3DCNN)is constructed,originated from a certain onshore wind turbine foundation reinforcement project.During the construction process of the proposed 3DCNN model,the vertical displacement monitoring data from the measurement points at different locations are spatio-temporally reconstructed firstly.The monitoring data is then imported into a spatio-temporal matrix,where the 3D convolutional kernel learns the spatio-temporal features between the data.Comparing with several prevalent neural network models,the 3DCNN model is found to have higher prediction accuracy in settlement prediction,as well as superior generalization and stability.

关 键 词：卷积神经网络 陆上风机 基础竖向位移 时空特征 

分 类 号：TU433[建筑科学—岩土工程]