摘要
针对岩爆样本数据噪声高、数量少从而导致岩爆等级预测准确率较低的问题,提出了基于非线性分数阶中值鉴别空间学习(nonlinear fractional-order median discriminative space learning,NFMDSL)的岩爆预测方法。该方法用类中值代替类均值,构建了中值鉴别空间学习方法,更好地保留了样本的有效信息,降低了噪声对预测效果的影响。为了有效捕捉岩爆数据间的非线性鉴别结构,进一步借助核技术将样本数据投影到核空间中。此外,引入分数阶对散度矩阵的特征值和奇异值进行重新估计,可以从少量样本中提取出具有良好区分能力的岩爆特征。结果表明,NFMDSL方法在岩爆等级预测中的平均准确率达到了95.75%,相比其他方法具有更高的准确率和更强的鲁棒性。该方法能够有效应用于矿山和隧道工程领域的岩爆预测。
To address the issues of low accuracy of rockburst grade prediction due to high noise and small number of rockburst sample data,a prediction method based on nonlinear fractional-order median discriminative space learning(NFMDSL)was proposed.The class sample medians were used to replace class sample means,resulting in the construction of a median discriminative space learning method that better preserved effective sample information and reduced the impact of noise on prediction performance.To effectively capture the nonlinear discriminative structures among rockburst data,the sample data were further projected into a kernel space using kernel techniques.Additionally,fractional-order methods were introduced to re-estimate the eigenvalues and singular values of the scatter matrix,enabling the extraction of rock burst features with strong discriminative capabilities from a limited number of samples.The results indicated that the NFMDSL method achieved an average accuracy of 95.75%in rock burst level prediction and it demonstrated higher accruacy and stronger robustness compared to other methods.This method can be applicable to rockburst prediction in the fields of mining and tunnel engineering.
作者
樊腾悦
苏树智
朱彦敏
FAN Tengyue;SU Shuzhi;ZHU Yanmin(School of Computer Science and Engineering,Anhui University of Science&Technology,Huainan 232001,China;School of Mechatronics Engineering,Anhui University of Science&Technology,Huainan 232001,China)
出处
《湖北民族大学学报（自然科学版）》
CAS
2024年第4期480-485,513,共7页
Journal of Hubei Minzu University:Natural Science Edition
基金
国家自然科学基金项目(52374155)
安徽省自然科学基金项目(2308085MF218)
安徽省高等学校自然科学研究项目(2022AH040113)
安徽省高校中青年教师培养行动项目(YQZD2023035)
淮南市指导性科技计划项目(2023142,2023147)
安徽理工大学青年基金项目(QNZD202202)。
关键词
岩爆预测
类中值
核技术
散度矩阵
奇异值
少量样本
矿山和隧道工程
rockburst prediction
class median
kernel techniques
scatter matrix
singular value
small sample size
mining and tunnel engineering
