Prediction Method and Application of Rock Burst Intensity Grade in Railway Tunnel Based on IDBO-XGBoost
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
维普
万方数据
为减少铁路隧道施工过程中岩爆事故的发生,在施工前做好岩爆烈度等级预测,提出了改进蜣螂优化算法(Improved Dung Beetle Optimizer,IDBO)与极限梯度提升树(Extreme Gradient Boosting,XGBoost)相结合的铁路隧道岩爆烈度等级预测模型.首先,依据岩爆成因及特点,综合选取围岩切向应力(σθ)等四个特征因素作为预测岩爆烈度等级的主控因素,建立岩爆烈度等级预测数据集;其次,引入Sine混沌映射、黄金正弦策略(Golden Sine Strategy,SA)、自适应高斯-柯西变异扰动策略以及贪婪选择策略并进行改进,以提高其全局搜索能力和稳定性;而后利用IDBO优化XGBoost中的超参数提升其预测精度,同时避免XGBoost出现"过拟合"现象;最后,将其结果与DBO-XGBoost、粒子群算法优化反向传播神经网络模型(Particle Swarm-Optimization Back Propagation Neural Network,PSO-BPNN)和遗传算法优化支持向量机模型(Genetic Algorithm Support-Vector Machine,GA-SVM)的结果进行对比.结果表明:IDBO-XGBoost模型准确率最高,相较于其他三种模型在测试样本中的准确率分别提高了8.69%、17.39%、8.69%;IDBO-XGBoost模型在处理岩爆问题上能更好地捕捉岩爆等级与指标之间的联系,可为实际工程的岩爆预测提供科学依据.
In order to reduce the occurrence of rock burst accidents during railway tunnel construction and predict the rock burst intensity level well before construction,a prediction model of rock burst intensity level of railway tunnel was proposed,which combines Improved Dung Beetle Optimization(IDBO)algorithm and Extreme Gradient Boosting(XGBoost).Firstly,based on the genetic characteristics of rock burst,four characteristic factors such as the tangential stress σθ of surrounding rock were selected as the primary control variables for predicting the intensity grade of tunnel rock burst.Consequently,a prediction dataset for rock burst intensity grades was established.Then,the Sine chaotic mapping,Golden Sine Strategy(SA),Gaussian-Kerchy variation perturbation strategy and greedy selection strategy were incorporated to enhance the global search capability and stability.Subsequently,IDBO was employed to optimize the hyperparameters in XGBoost for improved prediction accuracy and prevention of overfitting phenomena.Finally,the results were subsequently compared with DBO-XGBoost,Particle Swarm-Optimization Back Propagation Neural Network(PSO-BPNN)and Genetic Algorithm Support-Vector Machine(GA-SVM)models.The results show that,the IDBO-XGBoost model exhibits the highest accuracy among all models tested,demonstrating an improvement of 8.69%,17.39%and 8.69%over the other models in terms of test sample accuracy.Meanwhile,IDBO-XGBoost model can better capture the relationship between rock burst grade and index in dealing with rock burst problems.The aforementioned statement can furnish a scientific foundation for the anticipation of rock burst incidents in practical engineering.
维普
万方数据
