基于SVM理论的大坝变形监测模型改进方法研究

Improved Method of Dam Deformation Monitoring Model Based on SVM

分析与处理大坝变形监测资料对于大坝安全运行意义重大.支持向量机(SVM)能够有效解决高维数的非线性问题,并且具有良好的泛化能力.以SVM理论为基础,建立大坝变形监测模型,并在此基础上研究其改进方法.改进思路为充分利用马尔科夫链适用于原始监测数据波动大的优势,对其残差进行处理;同时考虑到核参数和惩罚因子的选择对SVM模型有很大影响,采用改进粒子群算法对其参数进行寻优.通过实例分析比较各种改进方法,结果表明,提出的对SVM模型的改进方法可以提高预测的泛化能力及精度.

Analyzing and processing the dam deformation monitoring data are significant.The support vector machine（SVM）can solve high-dimensional nonlinear problem effectively,and it has good generalization ability.This paper built dam deformation monitoring model based on SVM,and came up with improved methods.Taking advantage of Markov chain model＇s advantage which adapts to data fluctuating greatly to process the residual.And considering that choice of kernel parameter and penalty factor influences SVM model greatly,the parameter optimization is based on improved method of particle swarm optimzation（PSO）.Comparing with different improved methods through case studies,the results indicate that the improved method can increase the generalization ability of the model and precision.