基于机器学习的岩石节理面力学性能分析及预测
Analysis on mechanical properties of fractured rock mass and intelligent prediction based on machine learning
林永贵 1王海波 2魏立新 1徐江平 1马辉1
作者信息
- 1. 广州市市政工程设计研究总院有限公司,广东 广州 510060
- 2. 中山大学 航空航天学院,广东 深圳 518107
- 折叠
摘要
在岩土及隧道工程中,准确判定破碎岩体的宏观力学性能对工程设计和施工建造至关重要,而不同岩石节理形貌直接影响其宏观力学性能.为了有效界定节理岩体的力学性能,首先将频谱分形维数D和频域幅值积分Rq 作为节理形貌的量化表征参数,进一步基于傅里叶变换技术,设计了可指定形貌特征的节理重构方法,并结合3D打印技术对包含不同形貌节理的破碎岩石进行了直剪试验,验证了所采用的数值模拟方法的准确性.在此基础上,对不同节理形貌的岩石力学性能开展参数分析,研究结果表明分形维数D和频域幅值积分Rq 是能有效量化和评价节理形貌的参数.最后基于遗传算法改进的BP神经网络,构建了分形维数D、频域幅值积分Rq、法向压力、摩擦系数与破碎岩石力学性能之间的映射关系,形成了一种考虑节理形貌特性的破碎岩石力学性能智慧预测方法.
Abstract
Accurate determination of macroscopic mechanical properties of fractured rock mass is critical in geotechnical and tunnel engineering for effective design and construction.The mechanical behavior of these rock masses is directly influenced by the various morphologies of rock joints.In light of this,this study employed the spectrum fractal dimension D and frequency do-main amplitude integral Rq as quantitative parameters to characterize joint morphology.Furthermore,a joint reconstruction method utilizing Fourier transform technology was devised to precisely define the shape characteristics.To validate the proposed approach,direct shear tests were conducted on rocks with different joint morphologies,employing a combination of 3D printing and numerical analysis techniques.The numerical calculation model was subsequently calibrated for accuracy.Building upon these findings,a systematic parameter analysis was performed to evaluate the rock mechanics performance across diverse joint morphologies.The research results indicated that fractal dimension D and frequency domain amplitude integral Rq are effective parameters for quanti-fying and evaluating joint morphology.Finally,based on the genetic algorithm improved BP neural network,a quantitative mapping relationship between fractal dimension D,frequency domain amplitude integral Rq,normal pressure,friction coefficient and the me-chanical properties of fractured rocks was constructed,forming an intelligent prediction method for the mechanical properties of fractured rocks that considers the characteristics of joint morphology.
关键词
节理形貌分析/节理量化重构/数值模拟/直剪试验/机器学习Key words
joint morphology analysis/quantitative reconstruction of joints/numerical simulation/direct shear test/machine learning引用本文复制引用
基金项目
国家自然科学基金青年基金项目(52208381)
出版年
2024
国家科技期刊平台
NSTL
万方数据