基于改进Db1GEGA-SVM的大坝变形预警模型

Dam Deformation Monitoring Model Based on Improved Db1GEGA-SVM

支持向量机(SVM)能有效解决高维数非线性问题,且具有很好的泛化能力,其关键在于惩罚因子及核参数的选取;遗传算法具有良好的全局搜索能力与潜在的并行性,但局部搜索能力差,且易陷入早熟收敛。为提高大坝变形预警模型精度和泛化能力,提出利用改进的双切点交叉遗传算法(Db1GEGA)对SVM模型进行参数寻优,构建了基于改进Db1GEGA-SVM的大坝变形预警模型,并通过实例应用做了比较。结果表明,基于改进Db1GEGA-SVM的大坝变形预警模型具有更强的泛化能力和更高的预测精度。

The high-dimensional nonlinear problem can be solved by support vector machine（SVM）effectively, and it has good generalization ability. The key of SVM is to choose the penalty factor and kernel parameter. Genetic algorithm has good global search capability and potential parallelism, but it has poor local search ability and is easy to fall into ＂pre- mature convergence＂. In order to raise the generalization ability and prediction accuracy of dam deformation monitoring model, improved DblGEGA algorithm was used to optimize the parameters of SVM. And then the dam deformation mo- nitoring model was established based on improved DblGEGA-SVM. Compare with example＇s application, the results show that the proposed model has stronger generalization ability and higher prediction accuracy.