Research On Monitoring System Design And Risk Prediction For Rock Burst

Rock burst is an important part of coalmine accidents. In our country, the mines suffered fromrock burst are characterized by large quantity and wide distribution. And the frequency of rock burstis closely interrelated with mining depth. However, compared with other coalmine accidents, harm ofrock burst hasn’t been fully recognized by the public. So, it is necessary and important to have aresearch about rock burst. This thesis focuses on its prediction from the view of practical monitoringand the theoretical inference. The main work follows as bellow:(1) To overcome the shortage of networking mode and not real-time performance in the presentunderground sound monitoring system, a novel Underground Sound Transmitter(UST) isdesigned using LM3S9B96processor, which utilizes lwIP protocol to transmit data. Therefore,data is transmitted without any distance restriction. In addition, this UST not only has highsampling precision, but also owns good real-time performance.(2) The traditional least square method is susceptible to the errors in rock burst source location. Andits location result differs greatly from the real rock burst source location. On the basis of Newtoniteration method, this thesis proposes single rank inverse Broyden iterative method based on leastsquare. This new method not only overcomes the weakness of high computation cost and beingsensitive to initial value in Newton iteration method, but also improves prediction accuracycompared with least square method.(3) Considering the high cost of current rock burst prediction algorithms, this thesis puts forward acascade based cost sensitive classification algorithm. Along with instance-weighting、resamplingand threshold-moving, these algorithms are applied to the prediction of risk level of rock burst.The result shows the total costs of cost sensitive algorithms are fewer than the normal methodwithout cost sensitive. Particularly, the cascade based cost sensitive classification algorithmperforms best.