融合神经元网络与建模分析的基坑变形预测研究

发布时间：2018-01-13 09:07

本文关键词：融合神经元网络与建模分析的基坑变形预测研究　出处：《北京交通大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着城市化的快速发展,地铁工程及高层建筑等大型建设项目不断增多,与之相应的基坑工程的稳定对于上部结构及周围建筑物的安全具有重要影响,在基坑开挖过程中需要对其变形进行实时的监测和预测。基坑变形预测的理论及技术研究、改进对施工控制、事故防范和安全施工具有非常重要实际意义。本文探讨大型基坑的变形特点、监测技术及变形预测的理论和方法,在深入分析研究人工神经网络模型用于基坑变形预测的基础上,根据基坑工程的特点,总结了将有限元模型所反映的基坑力学特性与人工神经网络模型相融合的基坑变形预测方法,建立了相应的变形预测模型,结合实际基坑工程实例监测方法和实测数据分析,对其进行了近似建模与神经网络融合模型的变形预测,并对结果进行了对比分析。本文针对神经网络模型在力学关系上反映不足的弱点,发挥其数值上贴合度好的优势,综合考虑实际基坑结构力学关系复杂的特点,使用Abaqus有限元软件对基坑结构以及基本开挖工况进行了近似建模,并将模型变形由于力学特性而表现的变化关系引入到神经网络模型中,使用Matlab开发平台实现了BP神经网络融合模型的构建及变形预测处理。通过将融合模型的变形预测结果与基于实测数据的神经网络模型变形预测结果进行对比分析,结果表明融合模型能更好地反映基坑结构内部在施工过程中的力学关系特征,对基坑的变形预测结果优于单一的神经网络预测模型,对工程施工的安全控制及变形预测方法的研究改进具有积极的参考作用。
[Abstract]:With the rapid development of city urbanization, increasing large metro and high-rise building construction project, foundation pit engineering and the corresponding stability has an important influence on the upper structure and the safety of the surrounding buildings, the deformation monitoring and prediction of the real-time need in the excavation process. The theory and technology of prediction of the deformation of foundation pit. Improvement of construction control, has very important practical significance for accident prevention and safety construction. This paper discusses the deformation characteristics of foundation pit, the theory and method of monitoring and deformation prediction, in the research of artificial neural network model for analysis based on deformation prediction of foundation, according to the characteristics of the foundation pit deformation prediction method of foundation pit excavation integration summarizes the characteristics of the finite element model and artificial neural network model, established the corresponding deformation prediction model, combined with the Analysis of the actual excavation monitoring method and the measured data, the approximate deformation prediction model and neural network fusion model, and the results were compared and analyzed. Based on neural network model reflecting the weakness in mechanics, the numerical fit to play good advantage, considering the actual characteristics of structural mechanics the pit complex, using Abaqus finite element software for structure and foundation pit excavation of the basic conditions of approximate modeling, and model deformation relationship and performance due to mechanical properties into the neural network model, using the Matlab development platform for the realization of the construction of BP neural network for deformation prediction and fusion processing. The deformation prediction results the fusion model and neural network model based on the measured data of deformation prediction results are compared and analyzed. The results show that the fusion model can It can better reflect the mechanical characteristics of the foundation pit structure in the construction process, and the prediction result of the deformation of foundation pit is better than the single neural network prediction model. It has a positive reference function for the safety control of engineering construction and the research and improvement of deformation prediction method.

【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TU433

【相似文献】