1 Localization and Discrimination of Microseismic/AE Sources …
9
noticed that the starting-up angles of the two types tend to be concentrated into two
different intervals. The inaccuracy of the P-wave arrival’s picking led to the difficulty
of calculating the staring-up angle directly. The slope value of the starting-up trend
line obtained by linear regression was selected to substitute the angle. Two slope
values associated with the coordinates of the first peak and the maximum peak were
extracted as the characteristic parameters. The discrimination model was established
based on the Fisher discriminant analysis and the accuracy of classification reached
97.1%.
1.3.1.3 Discrimination of Mine Seismic Events and Blasts Using
Machine Learning [21]
Five typical parameters of blasts, the probability density functions of blast time,
and probability density functions of origin time difference for neighbouring blasts
were extracted as discriminant indicators. The discrimination models were established based on the Fisher classifier, naïve Bayesian classifier, and logistic regression,
which were performed on three databases from Australia and Canada. The proposed
discrimination models have explicit and simple functions, which are convenient for
workers in mines or researchers. Back-test, applied results, cross-validated results,
and analysis of ROC curves in different mines have demonstrated that established
discrimination models have a reasonably good performance.
1.3.1.4 Discrimination of Mining Microseismic Events and Blasts Using
Convolutional Neural Networks and Original Waveform
The source parameters used in the statistical methods are extracted from the original
waveforms according to criteria or experience, which may lead to a partial loss of
information or mixed subjective factors affecting the real expression. The waveform
can provide the most complete source information as the original information. It is
usually argued that the P-wave caused by blasts is simple than the P-wave produced
by the rock microfracture, and the energy of blasts mainly concentrated in the first
half of the timeline, sometimes there will be obvious in the wave window of several
same waveforms in the blasts. Compared with blasting waveform, the microseismic
waveform has obvious S-wave and the energy of the microseismic events distributes
in all time periods. Therefore, a wave classification model is established based on the
difference between microseismic events and blasts in waveform using the Convolution Neural Network (CNN). It not only decreases the influence from individual
differences in experience, but also removes the errors induced by source and waveform parameters. Results proved that the established discriminant method improved
the efficiency and accuracy of microseismic data processing for monitoring rock
instability and seismicity.
9
noticed that the starting-up angles of the two types tend to be concentrated into two
different intervals. The inaccuracy of the P-wave arrival’s picking led to the difficulty
of calculating the staring-up angle directly. The slope value of the starting-up trend
line obtained by linear regression was selected to substitute the angle. Two slope
values associated with the coordinates of the first peak and the maximum peak were
extracted as the characteristic parameters. The discrimination model was established
based on the Fisher discriminant analysis and the accuracy of classification reached
97.1%.
1.3.1.3 Discrimination of Mine Seismic Events and Blasts Using
Machine Learning [21]
Five typical parameters of blasts, the probability density functions of blast time,
and probability density functions of origin time difference for neighbouring blasts
were extracted as discriminant indicators. The discrimination models were established based on the Fisher classifier, naïve Bayesian classifier, and logistic regression,
which were performed on three databases from Australia and Canada. The proposed
discrimination models have explicit and simple functions, which are convenient for
workers in mines or researchers. Back-test, applied results, cross-validated results,
and analysis of ROC curves in different mines have demonstrated that established
discrimination models have a reasonably good performance.
1.3.1.4 Discrimination of Mining Microseismic Events and Blasts Using
Convolutional Neural Networks and Original Waveform
The source parameters used in the statistical methods are extracted from the original
waveforms according to criteria or experience, which may lead to a partial loss of
information or mixed subjective factors affecting the real expression. The waveform
can provide the most complete source information as the original information. It is
usually argued that the P-wave caused by blasts is simple than the P-wave produced
by the rock microfracture, and the energy of blasts mainly concentrated in the first
half of the timeline, sometimes there will be obvious in the wave window of several
same waveforms in the blasts. Compared with blasting waveform, the microseismic
waveform has obvious S-wave and the energy of the microseismic events distributes
in all time periods. Therefore, a wave classification model is established based on the
difference between microseismic events and blasts in waveform using the Convolution Neural Network (CNN). It not only decreases the influence from individual
differences in experience, but also removes the errors induced by source and waveform parameters. Results proved that the established discriminant method improved
the efficiency and accuracy of microseismic data processing for monitoring rock
instability and seismicity.