基于递归神经网络的基坑工程变形多步预测方法研究  

RESEARCH ON MULTI-STEP PREDICTION OF DEEP EXCAVATION DEFORMATION BASED ON RECURRENT NEURAL NETWORK

作　　者：刘勇健[1] 李彰明[1] 张建龙[1] 

机构地区：[1]广东工业大学岩土工程研究所,广州510640

出　　处：《工业建筑》2004年第9期22-25,共4页Industrial Construction

基　　金：广东省自然科学基金资助项目 (编号 :2 0 0 10 0 5 5 );广东工业大学重点学科基金 (编号 :2 0 2 16)

摘　　要：针对深基坑系统的复杂的非线性及基坑工程变形多步预测的重要性 ,将人工神经网络技术引入其中。分析了用BP网络进行多步预测时存在的不足 ,提出了基于递归神经网络的基坑工程变形多步预测模型。通过一软土深基坑工程变形多步预测实例的分析 ,论证了递归神经网络用于基坑工程变形多步预测的可靠性和实用性。该方法有效可行 ,在其他领域的多步预测中同样具有广阔的应用前景。An artificial neural network is introduced in the light of the complexity, nonlinearity of a deep excavation and the importance of multi-step prediction of its deformation. The defect of multi-step prediction by BP network is analyzed and a multi-step prediction model of an excavation deformation based on recurrent neural networks is also proposed. The reliability and practicability of the multi-step prediction of the excavation deformation by the recurrent neural networks are demonstrated through the multi-step prediction of the deformation of deep excavation in soft soil. It can be widely used for the muti-step prediction in other fields.

关 键 词：基坑工程 变形 软土深基坑 非线性 复杂 BP网络 领域 实例 论证 实用性 

分 类 号：TU753[建筑科学—建筑技术科学]