ARIMA方法预测吹填软基沉降的适用性  

Applicability of ARIMA method in settlement prediction of saturated dredger soft foundation

作　　者：孙长帅[1] 王基文 于天文 谭芳 宋志鑫 李敏 魏焕卫[2,3] SUN Changshuai;WANG Jiwen;YU Tianwen;TAN Fang;SONG Zhixin;LI Min;WEI Huanwei(Shandong Electric Power Engineering Consulting Institute Co.,Ltd.,Jinan 250013,China;School of Civil Engineering,Shandong Jianzhu University,Jinan 250013,China;Key Laboratory of Building Structural Retrofitting and Underground Space Engineering(Ministry of Education),Shandong Jianzhu University,Jinan 250013,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：《计算机辅助工程》2023年第1期69-73,共5页Computer Aided Engineering

基　　金：山东省住建厅研究开发项目(2020kykf05-018)。

摘　　要：为解决软基沉降难以预测的问题,基于软基超载预压监测数据,建立差分自回归移动平均(ARIMA)模型、BP神经网络模型以及长短期记忆(LSTM)模型。利用堆载前期的现场监测数据训练模型,根据训练模型对后期软基的沉降进行预测。结果表明:3种模型在软基超载预压处理后的沉降预测中均表现良好,并且ARIMA模型预测能力优于BP神经网络和LSTM模型。将ARIMA模型应用于软基超载预压期间的沉降预测可行。In order to solve the problem that the settlement of soft foundation is difficult to predict,the autoregressive integrated moving average(ARIMA)model,the BP neural network model and the long short-term memory(LSTM)model are established based on the measured settlement data of the soft foundation during the overload preloading process.These three models are firstly trained using in-situ monitoring data obtained in the early period of overloading,and then are used to predict the subsequent settlement of the soft foundation.The comparison results display that all three models perform well in the settlement prediction after the soft foundation overload preloading treatment.The ARIMA model is better prediction ability than the BP neural network model and the LSTM model.The ARIMA model is feasible and worth applying to the settlement prediction during the overload preloading.

关 键 词：差分自回归移动平均模型 BP神经网络 长短期记忆模型 沉降预测 吹填软基 

分 类 号：TU433[建筑科学—岩土工程] TB115[建筑科学—土工工程]