基于熵理论的混凝土坝变形安全监控SVM模型

Concrete Dam Deformation Safety Monitoring Model Based on SVM and Entropy Theory

为解决利用混凝土坝安全监测全序列数据建立的支持向量机(SVM)模型存在结构复杂、计算工作量大等问题,提出利用熵理论选择具有代表性样本代替全序列样本进行建模,即通过建立外部档案,根据外部档案更新算法选择具有代表性的样本,然后将外部档案的样本用作支持向量机的训练样本。将该方法用于某蓄水初期的混凝土坝变形模型的构建中,结果表明,该组合算法在保证模型精度的同时有效降低了模型的复杂程度,减少了模型的训练时间,且使模型的泛化能力得到一定的提升。

In order to solve the problems of complex structure and computing workload for modeling support vector machine(SVM)model with the full sequence data of concrete dam safety monitoring,the representative samples instead of the whole sequence samples was selected to establish the model by the theory of entropy.Firstly,an external archive file was build and the representative samples were selected by update of external archive file.Then,the samples in the external archive file were adopted to train support vector machine.The proposed model was used to establish concrete dam deformation model during initial impoundment.The results show that the combination algorithm not only ensures the accuracy of the model but also reducing the complexity and the training time of the model effectively as well as improving the generalization ability of the model.