摘要
利用大渡河上游岩石物理力学试验数据,采用基于贝叶斯机器学习的高斯混合模型构建了大渡河上游岩石物理力学参数概率分布模型。该方法和模型打破了现有方法必须事先假设各参数概率分布类型及相关结构类型的假设,可在考虑统计不确定条件下精准刻画联合分布特征及各参数间的相关性,构建了区域性多元岩石物理力学参数概率分布,给出了所选参数联合概率分布的GMM模型,明确了模型参数,可为后续工程和设计工作提供参考。
Based on the physical and mechanical experimental data of rock in the upper reaches of Dadu River,the probability distribution model of rock physical and mechanical parameters in the upper reaches of Dadu River was constructed by using the Gaussian mixture model(GMM)based on Bayesian machine learning.This method and model break the assumption that the probability distribution type of each parameter and the type of related structure must be assigned in advance in the traditional methods.This method can accurately describe the joint distribution characteristics and the correlation between parameters considering statistical uncertainty.The probability distribution of regional multivariate rock physical and mechanical parameters is constructed.The GMM model of the joint probability distribution of the selected parameters is given,and the model parameters are clarified,which provides a reference for subsequent projects and design work.
作者
邓钦宣
张韶鹏
邵伟
杨洋
DENG Qinxuan;ZHANG Shaopeng;SHAO Wei;YANG Yang(Sichuan Zumuzu River Basin Hydropower Development Co.,Ltd.,Chengdu Sichuan 610041;College of Water Resources and hydropower,Sichuan University,Chengdu Sichuan 610200)
出处
《四川水力发电》
2023年第5期83-89,95,共8页
Sichuan Hydropower
关键词
贝叶斯机器学习
高斯混合模型
岩石物理力学参数
概率分布
Bayesian Machine Learning
Gaussian Mixture Model
Physical and mechanical parameters of rock
Probability distribution
