基于高斯过程机器学习的冲击地压危险性预测  

Forecast of rock burst intensity based on Gaussian process machine learning

作　　者：苏国韶[1,2] 

机构地区：[1]岩土力学与工程国家重点实验室,武汉430071 [2]广西大学土木建筑工程学院,南宁530004

出　　处：《辽宁工程技术大学学报（自然科学版）》2009年第5期762-765,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金资助项目(40702053);中国科学院岩土力学与工程国家重点实验室开放研究基金资助项目(Z110601)

摘　　要：针对多种复杂影响因素条件下,如何有效预测冲击地压危险性这一类复杂的模式识别问题,提出一种基于高斯过程机器学习的冲击地压危险性预测新模型,通过对少量训练样本的学习,能很好地建立冲击地压危险性与其影响因素的非线性映射关系。算例结果表明,该模型科学可行、容易实现且预测精度高,具有良好的工程应用前景。Rock burst is affected by various complex factors.The effective forecast of rock burst based on these various contributing factors becomes a complicated pattern recognition problem.Gaussian Process（GP） model for binary classification is a kernel leaning machine with excellent capability of classification for solving pattern recognition problem of highly nonlinear and small sample size.A new model based on GP machine learning for forecasting rock burst intensity is proposed.Through learning the small training samples collected from a mine,the complicated nonlinear mapping relationship between intensity of rock burst and its contributing factors is established by the proposed model.The case study shows that the GP model is feasible,easy to be implemented.The proposed model is very attractive for a wide application in forecasting rock burst.

关 键 词：冲击地压 高斯过程 机器学习 预测 

分 类 号：TU45[建筑科学—岩土工程]