基于声发射信号聚类分析和神经网络识别的岩爆预警方法实验研究

Experimental study of rockburst early warning method based on acoustic emission cluster analysis and neural network identification

开展巷道岩爆室内模拟声发射监测实验,对岩爆过程中的声发射信号进行聚类分析,得到3种类型声发射信号,优选具有岩爆前兆演化异常、能量大、数量小等特征的信号类型作为岩爆前兆特征信号。岩爆发生前,岩爆前兆特征信号开始密集出现,明显的前兆异常。运用时间窗函数对岩爆前兆特征信号的时间密集性进行量化分析,提出衡量信号时序演化特征的量化指标即时间密集度f_t。岩爆发生前,前兆信号密集出现,出现了f_t≥3值,可将首次f_t≥3作为岩爆前兆信息,可将岩爆预警阈值a划定为3。基于岩爆前兆特征信号优选和预警阈值提取,借助BP神经网络,智能识取岩爆前兆特征信号,利用时间窗函数计算岩爆前兆特征信号f_t值,当岩爆前兆特征信号f_t值开始达到岩爆预警阈值(f_t≥a)时则进行岩爆灾害预警,建立了基于巷道岩爆模拟实验声发射数据的岩爆实时预警方法,为工程现场岩爆预警方法的建立提供了新思路。

Three types of AE signals are obtained by carrying out the tunnel rock burst simulation experiment and clustering rock explosion signal. The acoustic emission signals which have the characteristics of the precursory anomaly, large energy and small quantity, are preferred as precursor characteristic signals of rockburst. Before the rockburst, the precursor characteristic signal of rockburst appeared densely; it is shown obvious precursory anomaly. The window functionis used to carry out the quantitative analysis of temporal density of precursor signal and put forward the quantitative index——time intensity tf which can measure timing evolution characteristics of signal. Before the rockburst, the precursory signal appeared densely, tf is greater than or equal to 3. This situation firstly appears can be as the precursor information of rockburst; therefore the early warning threshold a can be set to 3. Based on optimization of rockburst precursor signal and early warning threshold extraction, with the help of BP neural network, intelligently identify rockburst precursor signal. The time window function is used to calculate tf value of precursor signal. Whentf value of precursor signal start to achieve rockburst warning threshold(tf ≥a), it starts to warn rockburst. Therefore, a real-time warning method of rockburst based on AE data of tunnel rockburst simulation experiment is established so as to provides a new idea for the construction of early warning method of rockburst.