基于BP神经网络的船闸基坑变形预测方法

Deformation prediction method of lock foundation pit based on BP neural network

船闸基坑开挖改变了地层中的原始应力状态,造成临近地面及临近建筑物产生变形。对基坑临近地面及建筑物变形进行预测,以选择经济合理的支护措施尤为重要。依托实际船闸基坑工程,考虑粘聚力、内摩擦角、弹性模量、基坑深度、放坡坡率以及地下水因素共同作用,采用有限元软件Midas GTS/NX对各因素通过正交设计后的基坑施工工艺组合开展了数值模拟,得到了不同组合下基坑临近地面不同位置处的变形特征。基于BP神经网络理论,建立了滨海地区船闸基坑无支护情况下开挖时临近地面变形的三层结构(7-7-1)神经网络预测模型。利用得到的179组基坑沉降数据对模型进行训练,21组数据对模型进行验证。结果表明,模型能够很好地对滨海地区基坑临近地面沉降进行预测。该方法可为类似工程支护措施设计提供参考。

The initial stress state in the ground has changed due to the excavation of the foundation pit,and resulted in the deformation of the ground and adjacent buildings.It is very important to predict the deformation of adjacent ground and buildings,so as to select the economic and reasonable supporting measures.Relied on a foundation pit engineering,the coupling effects of cohesion,internal friction angle and elastic modulus of soil,the excavation depths,slope rate and groundwater were considered,and the finite element software Midas GTS/NX was adopted to simulation the construction scheme which was orthogonal designed of those factors,and the deformation characteristics of the adjacent the ground at different positions are obtained.A three-layer structure(7-7-1)neural network model was established for the near ground of pit engineering that without support in coastal area based on BP neural network theory.The prediction model was trained by using 179 sets of settlement data,and validated by using another set of data.The results show that the model can well predict the settlement of adjacent ground of the foundation pit in coastal area.To some extent,this method provide reference for similar foundation pit engineering.