| Microseismic monitoring technology is a very effective monitoring tool in deep rock engineering,and the monitored microseismic signals contain rich information.It is of great practical significance to identify the monitored microseismic signals quickly and accurately and extract the information from them for effective rockburst risk warning.Based on this,this paper takes the Jinping large facility project as the research object,and closely focuses on microseismic signal identification,rock burst risk warning and risk prevention and control issues,and conducts a comprehensive research method such as microseismic monitoring system establishment,field monitoring analysis,algorithm model improvement and optimization,and engineering verification,etc.,and analyzes and obtains the site adjacent to the short distance cavern microseismic sensor fixed arrangement monitoring method and expansion of the same side with the palm surface mobile microseismic sensor arrangement The monitoring method was successfully established,the microseismic monitoring system was successfully established,the microseismic activity during excavation was analyzed,the damage mechanism at the local instability was studied,the optimized support method for risk prevention and control at the instability and the blasting construction plan for deep rock works were proposed,and the support vector machine signal classification model was optimized and improved based on the random forest and grid search method.The joint risk warning method of multi-parameter indicators is established and verified in practical use.The main research results obtained in this paper are as follows:(1)The basic characteristics and differences of four typical microseismic signals,namely rock rupture signal,blasting signal,anchor rig signal and electrical signal,were summarized,and it was found that the rock rupture signal and blasting signal were difficult to be distinguished only by visual observation of waveform characteristics;therefore,they were studied and analyzed in depth,and the time domain characteristics of rupture signal and blasting signal were extracted by recursive STA/LTA method and the frequency domain characteristics were extracted by integrated window function Fourier transform.(2)The analysis found that the RBF kernel function is better in the SVM algorithm selection,and combined with random forest and grid search to optimize the SVM,the BF-GS-SVM automatic classification model was established to learn the classification of rock rupture signal and blast signal eigenvalues,and the practical application proved that the model classification effect is superior.(3)The study shows that the prediction parameters of quantitative seismology and the characteristic indexes of microseismic activity,depending on the volume and energy index,can better respond to the strength of microseismic activity and the internal instability of the surrounding rock,and a multi-parameter index comprehensive early warning method is established based on the prediction parameters of seismology and the characteristic indexes of microseismic activity.(4)The number of microseismic events accumulates more in the early stage of excavation in Jinping large facility,and the external forces such as site construction disturbance is an important influencing factor to induce microseismic events.Among them,the microseismic activity is intense at the channel B0+7～10,whose rock microrupture form is mainly tensile rupture,accompanied by partial shear rupture and a few compression ruptures. |
