| 摘要: |
| 依托拉哇水电站地下厂房工程建设,建立地下厂房数值模型,确定敏感性显著的岩层的弹性模量E、粘聚力c和摩擦角φ三个力学参数作为反演目标,引入冠豪猪(CPO)算法对最小二乘支持向量机(LSSVM)算法进行超参数寻优,构建基于CPO-LSSVM算法的围岩力学参数反演模型。结果表明,利用CPO算法优化LSSVM的超参数,提高了算法的鲁棒性、计算效率和精度。相较于LSSVM,CPO-LSSVM算法的弹性模量和粘聚力的预测值绝对误差分别下降了51.88%和36.63%。通过后验差法对CPO-LSSVM反分析模型进行评价,评价结果为优,可在类似工程力学参数反演中推广应用。 |
| 关键词: 超参数 力学参数反演 冠豪猪优化算法 最小二乘支持向量机 后验差法 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Analysis on Inversion of Mechanical Parameters of Surrounding Rock in Underground Powerhouse Based on CPO-LSSVM |
|
Chang Liuhong1, Gao Hongyu1, Zhu Yong2, Wu Chuanfeng1, Du Yongcheng1, Zhang Jiaxin1
|
|
1.School of Hydraulic and Ocean Engineering,Changsha University of Science Technology;2.Huaihua Water Resources and Hydropower Survey,Design and Research Institute Co,Ltd
|
| Abstract: |
| Relying on the construction of the underground powerhouse of the Lawa Hydropower Station, a numerical model of the underground powerhouse was established, and three mechanical parameters of the significantly sensitive rock layer, namely the elastic modulus E, cohesion c, and friction angle, were selected as inversion targets. The Crested Porcupine Optimizer (CPO) algorithm was introduced to optimize the hyperparameters of the Least Squares Support Vector Machine (LSSVM) algorithm, and a surrounding rock mechanical parameter inversion model based on the CPO-LSSVM algorithm was constructed. The results show that optimizing the hyperparameters of LSSVM using the CPO algorithm enhances the robustness, computational efficiency, and accuracy of the algorithm. Compared with LSSVM, the absolute errors of the predicted values of elastic modulus and cohesion obtained by the CPO-LSSVM algorithm are reduced by 51.88% and 36.63%, respectively. The CPO-LSSVM back-analysis model was evaluated using the posterior error analysis method, and the evaluation result was rated as "excellent," indicating its potential for promotion and application in the inversion of mechanical parameters in similar engineering projects. |
| Key words: Hyper parameter Inversion of Mechanical Parameters Crested Porcupine Optimizer Least Squares Support Vector Machine Posterior Difference Method |
