仙女岩隧道下穿油气管道段爆破振动效应预测研究

发布时间：2018-01-12 04:22

本文关键词：仙女岩隧道下穿油气管道段爆破振动效应预测研究　出处：《西南交通大学》2015年硕士论文　论文类型：学位论文

【摘要】：本文以西成客专仙女岩隧道出口段掘进爆破为工程背景。由于隧道下穿油气管道,并且隧道走向与天然气管道走向相交,垂直距离约55m。隧道采用钻爆法施工,必然会对周围岩层及油气管道造成一定影响。我们需要控制对掘进爆破过程中所产生的影响,达到确保油气管道安全的目的。本文对现场监测数据进行地表爆破振动特性的分析得到在掘进爆破过程不同类型的炮孔产生了不同程度的波的叠加。通过频谱分析得到随着测点与掌子面的距离不断增大,振动速度主频率也相应提高,单段装药量的增加,相应的峰值振动速度的主频率反而降低的规律。并且EMD分解能选择性的滤掉不需要的波,而通过重构波的幅值包络图能分析出每段雷管实际起爆时间及之间的延迟。通过实测数据的回归拟合分析,得出该工程并不符合萨道夫斯基的经验公式,需要采取其他方法进行爆破振动峰值预测。本文选择BP神经网络、GA-BP、PSO-SVM、 GA-SVM四种预测方法,通过实测数据进行训练,并得到每种方法的预测结果。通过预测结果对比,误差相对百分比依次为GA-BP、PSO-SVM、GA-SVM、BP神经网络。其中GA-BP、PSO-SVM预测误差百分比均很小,然而GA-BP的拓扑结构依然是BP神经网络的拓扑结构,具有一定的不稳定性,而支持向量机克服了传统的神经网络所固有且无法避免的缺陷,并且大多数工程中都是小样本预测。综合而言采用PSO-SVM方法进行预测,效果最佳。
[Abstract]:The project background is the driving blasting of the exit section of the Xianxian tunnel in the west of this paper. Because of the passage of oil and gas pipeline under the tunnel and the intersection of the tunnel direction and the natural gas pipeline direction. The vertical distance is about 55m.The tunnel is constructed by drilling and blasting method, which will inevitably affect the surrounding rock formations and oil and gas pipelines. We need to control the impact on the excavation blasting process. In order to ensure the safety of oil and gas pipeline, this paper analyzes the vibration characteristics of surface blasting with the monitoring data on the spot. It is concluded that different types of blasting holes have different degrees of wave superposition in the process of tunneling blasting. The results show that the distance between the measuring point and the palm surface is increasing. The main frequency of the vibration velocity also increases correspondingly, and the main frequency of the corresponding peak vibration velocity decreases with the increase of the charge quantity in the single section. Moreover, the EMD decomposition can selectively filter out the unwanted wave. Through the reconstruction of the amplitude envelope diagram, the actual detonator initiation time and the delay between the detonators can be analyzed. Through the regression fitting analysis of the measured data, it is concluded that the project does not accord with the empirical formula of Sadolski. It is necessary to adopt other methods to predict the peak value of blasting vibration. In this paper, the BP neural network and four prediction methods of PSO-SVM and GA-SVM are selected and trained by the measured data. The relative percentage of the error is GA-BPU PSO-SVMN GA-SVMBP neural network, in which GA-BP. The percentage of PSO-SVM prediction error is very small, but the topological structure of GA-BP is still the topological structure of BP neural network, which has some instability. Support vector machine (SVM) overcomes the inherent and unavoidable defects of traditional neural networks, and most of the projects are small sample prediction. In summary, PSO-SVM method is the best method for prediction.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U455.41

【参考文献】