建筑物变形短期预测的神经网络方法  

Neural Network Method of Short-Term Predication of Building Deformation

作　　者：周满贵[1] 潘国荣[2] 

机构地区：[1]同济大学测量系,200092 [2]同济大学测量与国土信息工程系

出　　处：《测绘工程》1999年第4期63-66,共4页Engineering of Surveying and Mapping

摘　　要：从系统论出发，把建筑物变形看成一个黑箱系统，提出了利用神经网络对建筑物变形进行短期预测的一种新方法。在此方法中，利用多层前馈神经网络实现系统的输出与输入映射，采用正的反向传播算法（Bp）作为学习算法。用实例加以证明，其拟合和预测的结果都令人满意。In the paper, building deformation is regarded as a black-box according to the system theory. A method about the short-term predication of bui1ding deforamtion is put forward by analysing artifical neural network. The method adopts the adjusting Bp algrithm as the learning algrithm. At last, a given practical example has proved the efficacy of this method.

关 键 词：神经网络 BP算法 建筑物 变形预测 短期预测 

分 类 号：TU196.1[建筑科学—建筑理论]