基于波形特征和决策树分类算法的岩体破裂信号识别

Recognition of mining rock fracture signal based on waveform feature and decision tree classification algorithm

微震监测技术能够捕捉开采扰动下岩体响应信息,已被广泛应用于岩体稳定性分析与矿山安全生产管理。受矿山现场频繁生产活动的影响,微震监测系统能够捕捉到不同类型信号,导致噪音信号较多,无法及时有效地揭示开采扰动下岩体响应规律。本文依托阿舍勒铜矿微震监测,分析了微震系统采集典型信号波形参数特征的差异,提出了基于决策树分类算法的岩体破裂信号识别方法,并对其识别精度进行了对比分析。研究结果表明,电气噪音信号、爆破信号、机械振动信号、岩石破裂信号的持续时间、上升时间、振铃数、上升振铃数、最大振幅、主频等参数分布范围存在不同程度的重合,无法采用单一参数有效识别岩体破裂信号,消除噪音信号的影响。采用决策树分类算法构建岩体破裂信号识别模型,能够有效消除噪音信号的影响,识别准确率达97.8%,显著高于支持向量机(SVM)模型73.9%的准确率。研究成果对于快速圈定、预警岩体破坏高风险区域具有重要意义。

The microseismic monitoring technology has been widely used on the analysis of rock stability and management of mine safety production owing to its ability in catching the response information of rock mass under mining disturbance.Due to the impact of frequent production activities at the mine site,the microseismic monitoring system would catch so many different types of signals including lots of noise that the response law of rock mass under mining disturbance could not be revealed timely and effectively.This dissertation has analyzed the differences in the characteristics of typical signal waveform parameters according to the microseismic monitoring system built in Ashele Cooper Mine.Moreover,an identification method of rock mass rupture signal based on decision tree classification algorithm has been proposed,and made a comparative analysis of its recognition accuracy in this thesis.Besides the rock rupture signals,the other noise signals mainly contains electrical noise signals,mechanical vibration signals and blasting signals.The results indicate that these signals have different degrees of coincidence in the distribution range of parameters such as duration,rise time,ringing number,rising ringing number,maximum amplitude,and main frequency.Therefore,it is impossible to use a single parameter to effectively identify the rock mass rupture signal and eliminate the influence of noise signals.Additionally,the model basing on the decision tree classification algorithm performs better than SVM in terms of identifying the rock rupture signals,and its recognition accuracy rate has reached 97.8%.The results are of great significance for quickly delineating and early warning of high-risk areas of rock mass damage.