基于GA-SVM的碾压混凝土重力坝参数反演

Back Analysis of Mechanical Parameters of RCC Gravity Dam Based on GA-SVM

基于碾压混凝土重力坝的实测位移、应变反演弹性模量和徐变度,综合运用Fortran语言和数值计算软件Flac3D,开发了徐变体数值计算程序,实现了施工期和运行期全过程模拟计算。利用水位骤升期的位移和应变测值变化量反演坝体和坝基的稳定弹性模量,基于正交试验设计和徐变体数值计算程序,获得待反演参数与监测点应变计算值对应关系样本,借助智能算法支持向量机(SVM)和遗传算法(GA),建立应变计算值与待反演参数关系的GA-SVM预测模型。在反演计算过程中,以监测点的实测应变值为真值,应变计算值直接采用GA-SVM预测模型计算得到,省去了数值计算环节。根据反演参数反馈计算的应变值与实测值能够较好地吻合,研究成果表明徐变体反演计算方法合理有效,极大地减少了徐变体全过程数值计算及其参数反演所需的时间。

In this paper, the elastic modulus and creep degree of RCC gravity damwere calculated based on the measured displacement and strain. Using Fortran language and numerical calculation software Flac3 D, a creep numerical calculation program was developed to realize the simulation calculation of the whole process during construction and operation. The displacement and strain measurements during the sudden rise of water level were used to invert the stable elastic modulus of the dam body and dam foundation. Based on the orthogonal experiment design and the creep numerical calculation program, a sample of the correspondence between the parameters to be inverted and the calculated strain values at the monitoring points was obtained. With the help of intelligent algorithm support vector machine(SVM) and genetic algorithm(GA), the calculated strain values and GA-SVM prediction model for inverse parameter relationship was established. During the inversion calculation, the measured strain value at the monitoring point was true, and the calculated strain value was directly calculated by the GA-SVM prediction model, eliminating the need for numerical calculation. The strain values calculated according to the feedback of the inversion parameters were in good agreement with the measured values. The research results show that the inversion calculation method of the creep variant in this paper is reasonable and effective, which greatly reduces the numerical calculation of the entire process of the creep variant and its parameter inversion time required.