基于SVM的岩体灌浆可灌性分析预测模型

Groutability analysis and prediction model for grouting of rock mass based on SVM

摘要

影响岩体可灌性因素众多，特别是对于隐蔽的岩体裂隙用经验公式判断可灌性准确性较低。针对上述问题，从实现对岩体可灌性的快速准确分析的目的出发，利用支持向量机方法建立岩体可灌性分析预测模型。通过分析可灌性的影响因素，考虑灌浆数据集的小样本特性，分别建立岩体注灰量回归预测SVR模型和岩体可灌性分类预测SVM模型。进一步通过改进灰狼优化算法和增强鲸鱼优化算法，对基于支持向量机的预测模型的进行惩罚因子C、核函数参数g进行寻优。结果表明，通过与其他预测模型相比，所提出的可灌性预测模型分类预测准确率提高约6.5%,并具有收敛速度快的明显优势，验证了基于支持向量机的岩体可灌性分析预测模型的准确性和有效性。

Abstract

There are many factors that affect the groutability of rock mass, especially the accuracy of judging the groutability of rock mass with the empirical formula is low for the hidden rock mass fracture. Aiming at the above problems, the support vector machine method is used to establish the prediction model of rock mass groutability analysis in order to realize the fast and accurate analysis of rock mass groutability. Based on the analysis of the influencing factors and the small sample characteristics of the grouting data set, the regression prediction SVR model and classification prediction SVM model of rock mass groutability were established respectively. Further, by improving the gray Wolf optimization algorithm and enhancing the whale optimization algorithm, the penalty factor C and kernel function parameter g of the prediction model based on support vector machine are optimized. The result show that compared with other prediction models, the classification prediction accuracy of the groutability prediction model proposed in this paper is improved by about 6.5%, and it has the obvious advantage of fast convergence, which verifies the accuracy and effectiveness of the prediction model of rock mass groutability analysis based on support vector machine.

关键词

可灌性预测 / 支持向量机 / 注灰量预测 / 群智能优化 / 岩体灌浆

Key words

groutability prediction / support vector machine / cement take prediction / swarm intelligent algorithm / rock grouting

Author summay

王昶磊(1982—), 男, 高级工程师, 硕士, 从事水利建设与管理。E-mail: [e-mail]

基金资助

国家重点研发计划(2021YFC3090103)

天津大学自主创新基金(2023XJD-0065)