一重集团大连设计研究院有限公司研发大楼基坑变形研究

发布时间：2019-01-02 18:41

【摘要】：深基坑自身条件复杂,其稳定性易受多方面因素影响,且易与周边环境发生关联,在开挖施工过程,不仅要保证基坑本身的稳定性,还要确保基坑周边环境的安全。基坑变形监测是确保基坑及周围环境安全运行的基本手段,通过分析基坑变形监测资料,掌握基坑变形规律,并对基坑变形进行预测,可及时发现异常变化,防范于未然。本文主要研究BP神经网络理论,通过L-M算法优化传统BP网络中SDBP算法、小波变换理论与BP神经网络理论结合,引入两种BP神经网络优化模型。结合一重集团大连设计研究院有限公司研发大楼基坑变形实测监测数据,对其进行变形研究。本文主要内容及成果如下:(1)对基坑工程特点、影响因素、变形机理进行分析,对基坑工程监测技术进行研究,阐明基坑监测内容、监测方法,积累基坑工程监测经验。(2)通过输入层和输出层设计、隐含层节点数选定方法、网络函数的选定,确定基坑变形预测模型结构。(3)结合一重集团大连设计研究院有限公司研发大楼基坑工程监测实例,选择桩顶水平位移、锚索内力及地表沉降为研究对象。选取监测数据、建立数据样本、处理样本数据、设置网络参数,设计BP神经网络基坑变形预测算法。(4)应用MATLAB R2012a软件编写BP神经网络优化算法程序,通过对模型节点的调试训练,引入基于L-M算法的BP神经网络、小波神经网络两种基坑变形预测模型,将小波神经网络的拟合和预测结果及BP神经网络的拟合和预测结果进行对比分析,并分析基坑变形规律,评价基坑稳定性。结果表明两种方法都能很好反应基坑短期变形规律,但小波神经网络模型比BP神经网络模型在收敛性、计算速度和预测精度上更具有优越性,是一种较好的基坑位移预测方法,其在基坑工程领域有较好的发展和应用前景。
[Abstract]:The condition of deep foundation pit is complex, its stability is easily affected by many factors, and it is easy to be related to the surrounding environment. In the process of excavation construction, not only the stability of foundation pit itself should be guaranteed, but also the safety of the surrounding environment of foundation pit should be ensured. Foundation pit deformation monitoring is a basic means to ensure the safe operation of foundation pit and its surrounding environment. By analyzing the monitoring data of foundation pit deformation, mastering the deformation law of foundation pit, and predicting the foundation pit deformation, the abnormal changes can be found in time and the precautions can be taken. In this paper, the theory of BP neural network is studied, and two kinds of BP neural network optimization models are introduced by using L-M algorithm to optimize SDBP algorithm in traditional BP network, wavelet transform theory and BP neural network theory. Based on the measured monitoring data of foundation pit deformation of Dalian Design and Research Institute Co., Ltd, the deformation of Yizhong Group is studied. The main contents and achievements of this paper are as follows: (1) the characteristics, influencing factors and deformation mechanism of foundation pit engineering are analyzed, the monitoring technology of foundation pit engineering is studied, and the monitoring contents and methods of foundation pit monitoring are expounded. Accumulation of foundation pit engineering monitoring experience. (2) through the input layer and output layer design, hidden layer node number selection method, network function selection, The structure of foundation pit deformation prediction model is determined. (3) the horizontal displacement of pile top, internal force of anchor cable and surface settlement are selected as the research object combined with the monitoring example of foundation pit engineering of Dalian Design and Research Institute Co., Ltd of Yizhong Group. Select the monitoring data, set up the data sample, process the sample data, set up the network parameter, design the BP neural network foundation pit deformation prediction algorithm. (4) use MATLAB R2012a software to compile the BP neural network optimization algorithm program. Through debugging training of model nodes, two kinds of foundation pit deformation prediction models based on L-M algorithm, BP neural network and wavelet neural network, are introduced. The fitting and prediction results of wavelet neural network and BP neural network are compared and analyzed, and the deformation law of foundation pit is analyzed to evaluate the stability of foundation pit. The results show that the two methods can well reflect the short-term deformation of foundation pit, but the wavelet neural network model has more advantages than BP neural network model in convergence, calculation speed and prediction accuracy, so it is a better foundation pit displacement prediction method. It has a good prospect of development and application in the field of foundation pit engineering.
【学位授予单位】：辽宁师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TU753

【相似文献】