Research Article A Comparison of Mine Seismic ...

Research ArticleA Comparison of Mine Seismic Discriminators Based onFeatures of Source Parameters to Waveform Characteristics

Ju Ma Guoyan Zhao Longjun Dong Guanghui Chen and Chuxuan Zhang

School of Resources and Safety Engineering Central South University Changsha 410083 China

Correspondence should be addressed to Longjun Dong rydong001csueducn

Received 2 October 2014 Accepted 27 December 2014

Academic Editor Marcin A Lutynski

Copyright copy 2015 Ju Ma et al This is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

To find efficient methods for classifying mine seismic events two features extraction approaches were proposed Features ofsource parameters including the seismic moment the seismic energy the energy ratio of S- to P-wave the static stress droptime of occurrence and the number of triggers were selected counted and analyzed in approach I Waveform characteristicsconsisting of two slope values and the coordinates of the first peak and the maximum peak were extracted as the discriminatingparameters in approach II The discriminating performance of the two approaches was compared and discussed by applying theBayes discriminant analysis to the characteristic parameters extracted Classification results show that 835 of the original groupedcases are correctly classified by approach I and 971 of original grouped cases are correctly classified by approach IIThe advantagesand limitations pertaining to each classifier were discussed by plotting the event magnitude versus sample number Comparativeanalysis shows that the proposed method of approach II not only has a low misjudgment rate but also displays relative constancywhen the testing samples fluctuate with seismic magnitude and energy

1 Introduction

Mining excavations induce elastic and then inelastic defor-mation within the surrounding rock mass A seismic event isa sudden inelastic deformationwithin a given volume of rockHaving recorded and processed a number of seismic eventswithin a given volume of interest Δ119881 over time Δ119905 one canthen quantify the changes in the strain and stress regimes andin the rheological properties of the rock mass deformationassociatedwith the seismic radiation [1 2] However a varietyof dynamic processes in mines which radiate seismic wavesare detected by the seismic monitoring systems and in gen-eral seismograms generated by a development or productionblast and a shear fracturing or a sudden slip on a surfaceof weakness are the majority of records [3ndash8] As recordedquarry blasts may mislead scientific interpreting and lead toerroneous results in the analysis of seismic hazards in minesstandard processing of seismic monitoring data require theseevents to be separated An automatic classifier is necessaryto be developed to reduce the dramatically arduous task offinding to which class each recorded event belongs [9ndash17]

Many researches have been carried out on the topicof source location during the last decade All techniquesfor source location methods can be classified under twoconditions that require prior knowledge of the sonic speed ofthe structure [18] andmethods that do not require such infor-mation [19ndash21] However few researches focused on the dis-crimination of mine seismic events in the past Generally twosteps features extraction and statistical identification couldbe separated from all those limited amount discriminatorsParameters characterizing the source (such as magnitudepotency moment energy static stress drop apparent stressand apparent volume) and parameters directly extractedfrom the seismograms (including amplitude polarizationfrequency correlation coefficient and travel time) are thetwo categories of discriminating features Statistical method-ologies like fisher discriminant classifier logistic regressionunascertained measurement and neural networks have beenoccasionally carried out onmine seismic events identificationand classification

Malovichko selected the time of day the repetition ofwaveforms the high-frequency versus the low-frequency

Hindawi Publishing CorporationShock and VibrationVolume 2015 Article ID 919143 10 pageshttpdxdoiorg1011552015919143

2 Shock and Vibration

0 600(m)

N

(a)

netADC4

netADC4

netADC4 netADC4 netADC4 netADC4

netADC4 netADC4

23 24 25 26

netSP+ netSP+

netSP+ netSP+

11 12 13 14 15 16 17 18 19 20 21 22

10 9 8 7 6 5 4 3 T2 2 T1 1

Server

(b)

Figure 1 The microseismic monitoring system installed at Yongshaba mine (a) isometric view of the ground surface tunnels and themicroseismic monitoring system and (b) the configuration of the seismic network used in the project

radiation and the radiation pattern as features and thenestablished the GaussianMaximumLikelihoodClassificationmethod for the classification [2] This method providesa way to identify signals of different type but the greatamount of computation leads to low efficiency Vallejos andMcKinnon proposed the identification of seismic records inseismically activemines by considering the logistic regressionand the neural network classification techniques An efficientmethodology was presented for applying these approaches tothe classification of seismic records [3] However the calcu-lation of seismic source parameters requires precise signalprocessing namely expertise-required and time-consumingP- and S-wave hand-picking Besides the statistical classifi-cationmodels that only rely on source parameters show greatdifference in accuracy when applied to different mine sites

Liu et al proposed a synthesis method to identify micro-seismic event based on the triggering principle of short termaverage to long term average [9] But very small fluctuationsin the threshold would cause a great rate of misclassificationsince the threshold setting completely depends on experi-ence Zhu et al decomposed the microseismic signals into5 layers to gain specified frequency bands using waveletanalysis Based on the box fractal dimensions and thosespecified frequency bands 23-dimensional values of patternrecognition feature vector were established The supportvector machine which was adopted to train classify and rec-ognize shows a correct identification rate of 94 [10] Jianget al presented a three-step strategy to achieve the classifyingof local multichannels microseismic waveforms The authorextracted the time-frequency the amplitude distributionand the correlation coefficient as features and establishedan effective judgment mechanism [11] However the studiesdid not provide much attention to the huge workload ofcalculation in practical application Their input data couldnot be obtained directly from the monitoring system andtheir calculation algorithm is very complicatedThe efficiencyof identification still needs further improvement since anyalgorithm must be as simple as possible in order to run ona small low-power microprocessor

A classification method of mine blasts and microseismicevents suing the starting-up features in seismograms wasproposed It is a method that presents less computationwith a relatively low misjudgment rate In this study theBayes discriminant classifier is applied to further analysethe ldquostarting-up featuresrdquo The differential discriminatingperformances acquired by means of extracting the featuresof respectively source parameters and waveform were com-pared and discussed

2 Data

21 Engineering Background The Yongshaba Mine is locatedat Gaobang which is about 85 km northeast of Guiyang(26∘381015840N 106∘371015840E) Guizhou Province PR China It is themain operation base of the Guizhou Kailin Co Ltd withphosphate production capacity over 2000000 tonsyear Thecurrent exploitation stopes are mainly scattered on levels of1090m to 840m with a relative depth of 500m to 700mbelow the surface Tens of millions of tons has cumulativelyexcavated employing open stoping mining method sinceinitial operating in the 1950s

The study region covers a volume of approximately3000m times 300m times 350m between the depths of 300m and700m below the surface The underground microseismicmonitoring system used to inform the evolution of themicrofracture behavior consists of 26 uniaxial and 2 triaxialgeophones (Figure 1) The geophone manufactured by IMSholds a natural frequency of 14Hz and a sampling rate of6000Hz Signals from various dynamic processes fracturingin rock mass production and development blasts impactsand vibration of machinery all are being recorded by themicroseismic monitoring system

22 Databases The sample databases used in this studyconsist of two parts the subset of the signals of blast eventsand the subset of the signals of normal microseismic eventsEach of the signals in the blast subset has been confirmedto be consistent with the blast operations according to

Shock and Vibration 3

+ ++ +

(a)

++ minus

minus

(b)

Figure 2 Two typical seismic source processes inmines blast (a) and shear fracturing (b) (Malovichko 2012) Radiation pattern of an idealizedexplosion and of a strike-slip earthquake along a vertically dipping fault are shown correspondingly (Bormann 2002) The arrows show thedirections of compressional (outward polarity + red shaded) and dilatational (inward polarity minus green shaded) motions in P-wave

the time and location records provided by the miningworkersThe subset of themicroseismic events can be dividedinto two parts those related to discrete large-scale rockmass failures and those not resulting in observable rock massdamage The observable ones also have been confirmed tobe consistent with the fact For the unobservable ones theywere processed manually by three independent processorsTo aid in determining whether an event is a microseismicor not the guide lines of an unequivocal classificationwere listed in Table 1 Only when the three discriminatingresults are exactly equivalent can the event be recruited intothe database Those events that could not be determinedwhich type belongs to by manually approaches were notenrolled into the sample database The databases used in thispaper contain a total of 103 seismic records with all seismicparameters calculated from which 56 are labeled as normalevents and the others are tagged as blasts

3 Discriminating Features

31 Approach I Source Parameters A seismic event is thesudden release of potential or stored energy in the rockThe released energy is then radiated as seismic wave [7]The two typical seismic source processes in mines (the blastsand the shear fracturing of the microseismic events) andits approximate radiation patterns are shown in Figure 2It is obvious that the explosion radiates predominantly P-waves outward directing compressional directions while theshear fracturing or slip on a surface radiates S-waves thatare stronger than P-waves These characteristics can be usedto identify the type of source process and to discriminatebetween blasts and microseismic events [2]

311 Seismic Moment The basic characteristics of radiationcan be described by a seismic moment tensor This tensor

represents a set of fictitious dipoles acting on a point in thesource areaThemoment tensor makes it possible to describethe low-frequency amplitudes and polarities of seismicwavesInverting the moment tensor from the observed waveformsand analysis of its components allows the discriminationof blasts from slip-type and even from crush-type events[2 13 14]The logarithm of the seismicmoment is consideredas the feature of radiation

Figure 3(a) shows that the values of the log(119872) at thepoints with the highest probability density are 100 for blastsand 80 for microseismic events The seismic moment isa relative useful performing discriminating feature as itprovides a relatively large separation between normal eventsand blasts

312 Seismic Energy The energy release during rock frac-turing and frictional sliding comes from the transformationof elastic strain into inelastic strain [14] This transformationmay occur at different rates ranging from slow creep-likeevents to very fast dynamic seismic events The averagevelocity of deformation at the source is up to a few metersper second Unlike the dynamic sources of the same sizethe slow type events have long time duration at the sourceand thus radiate predominantly lower frequency waves Interms of fracture mechanics the slower the rupture velocityis the less energy the event radiates A quasi-static rupturewould radiate practically no energy Similarly for blastingsmaller blasts make smaller changes to the rockmass ormoregradual stress changes so that the response is less dynamic[12ndash16]

Observations from the probability density distribution(Figure 3(b)) indicate that the seismic energy is one of thebest performing discriminating features The value of log(119864)distributed fromminus2 to 2 formicroseismic events and 0 to 6 forblasts More than 65 of the blasts events could be accuratelyidentified in terms of this single indicator


Table 1 Guidelines for mine seismic events classification manually

Characteristics DescriptionThe presence of delays of the blast

00

02

04

06

08

10

Blasting seismogram

Time (s)

A monotonically

Similar signals within30

20

10

00

minus10

times 10minus4

minus20

Velo

city

(mmiddotsminus

1)

decreasing tail

a single waveform

with upper envelopeBlasts especially stope firings have multiple delays which areexpressed in the seismogram as similar signals repeating closelywithin a short time interval The practice of decides whether an eventis a blast or a microseismic event is based on the repetition featureBesides seismograms capturing a blast will have a monotonicallydecreasing tail commonly For the waveform of a microseismic signala large amplitude difference exists between the maximum peak andthe peak closest behind without a gradual decrease

The steeper rise in the energy curve of the normalevent

Time (s)

00

02

04

06

08

10

Microseismic seismogram

Energy curve

Velo

city

(mmiddotsminus

1)

120

80

times 10minus5

40

00

minus40

minus80

100

80

times 10minus7

60

40

20

00

with energy curve Seismograms capturing a microseismic event usually associated withshear fracturing and have an S-wave arrival more obvious than in thecases of blasts due to the source of the latter are usually in the focalmechanism of expansion and compression Furthermore blastingevents will typically have a consistent gradual rise in their energycurve while seismic events will tend to have a steeper more distinctrise in the energy curve and the energy curves for blasting events canbe loosely compared to a positive sloping flat line

The time of occurrence

0 4 8 12 16 20 24

000

005

010

015

020

025

Frac

tion

of o

ccur

renc

es

Hour of the day

BlastsMicroseismic events

Another way to eliminate blasts from the microseismic catalogue is toapply the time filters (ie generally mines have prescribed blastingtimes and the events that does not occur at the blasting time aremarked as microseismic events) Two main daily blasting shifts areobserved from the typical diurnal chart of Chinese mines hoursbetween 10sim16 (stope firings) and 23sim1 (development firings) each ofwhich triggers an increase in seismicity It has to be classifiedreferring to other features for the cases that recorded during theblasting time and located close to the blasting area

The dominant frequency

0 50 100

150

200

250

300

350

400

450

500

00


Am

plitu

de

Frequency (Hz)

100

80

60

40

20

times 10minus6

Fp = 6687

Fp = 2211 A large number of actual observations and analysis show that blastsor explosions usually radiate higher frequency waves compared tonormal microseismic events The statistics show that the value of thedominant frequency of the microseismic events varies from 10 to100Hz and from 70 to 260Hz for blasts In addition blasting eventswill typically not be well matched to the Brunersquos model curve asplotted in the Stacked Spectra Plot within TRACE (the softwareprovided by IMS) Spectral analysis is often considered as a relativeeffective method to distinguish blasts from microseismic events


Table 1 Continued

Characteristics Description

The maximum amplitude

0 10 20 30 40 50


Sample

15

10

times 10minus3

05

00

minus05

minus10

Max

imum

ampl

itude

(mmiddotsminus

1)

Themaximum amplitude of the waveform which characterizes thesize of energy released is also considered as an important referencefor the recognition of blasts and microseismic events Generally themaximum vibration velocity of microseismic events is about10minus5msdotsminus1 far less than 10minus3 msdotsminus1 of the blasting However thisapproach cannot be considered as universally applicable since it ispossible to have a scenario where a blast triggers shear fracturing thatradiates stronger seismic signals than the blast itself

313 S P Energy Ratio There are several seismic phases butonly S-waves and P-waves are commonly recorded in mininginduced seismicity P-waves or compressive waves are thefirst ones which arrive at receivers When seismic wavespropagate they carry energy from the source of the shakingoutward in all directions The ratio between S-wave energyand P-wave energy can be an indicator of seismic sourcemechanism [4] Seismic energy is proportional to the integralof the square of the vectorial sum of the velocity waveformand can be calculated separately for the P-wave and S-wave[7] For fault-slip type events in seismology (earthquakes)there is considerablymore energy in the S-wave than in the P-wave with the ratio of the S-wave energy to the P-wave energyfrequently greater than 10 [12] Urbancic et al noted thatfor nonshear seismic source mechanisms there would be adeficiency in S-wave energy or relatively more P-wave energythan for shearing events For nonshearing eventmechanismssuch as strain-bursting tensile failure and volumetric rockmass fracturing the ratio of S-wave energy to P-wave energyis frequently in the range of 3 or less [14]

Observations from Figure 3(c) point out that the mostprobability of the log(119864S119864P) falls in the interval of 00 to20 for microseismic events and minus10 to 40 for blasts It wasobvious that the S P energy ratio performs not well in thismine site since too small separation was provided

314 Static Stress Drop Shearer defined the term-stressdrop Δ120590 as the average difference between the stress acrossthe fault before and after an earthquake [17] There areseveral different methods of determining the stress dropof which some use records of ground velocity and groundacceleration Stress drops can vary considerably from eventto event For microseismic events in Yongshaba mine therange of log(SSD) is from 25 to 45 and 30 to 65 for blasts(Figure 3(d)) This parameter plays a tremendous role inYongshaba mines seismic discrimination

315 Time of Occurrence Mining-induced seismic eventsoccur most of the time in the ultimate proximity of mineworkings and concentrate during blasting time Statisticalresults show that almost all the blasts occur at 1000 to 1500and rarely occur in other time (Figure 3(e)) However alarge number of microseismic events take place at this timesimultaneouslyThe performance of the ldquotime of occurrencerdquois discounted

316 Number of Triggers The Number of triggers is affectedby many factors Mine layout geological features seismiclocations and the sensor array all are likely to affect thenumber of triggered sensors It should be guaranteed thatthe potential areas of microseismic occurring and blastingoperations are enclosed inside the sensor array Only inthis way can the ldquonumber of triggersrdquo reflect the pattern ofseismic wave propagation and the scale of seismic energyFigure 3(f) shows blasting events usually triggering moresensors than normalmicroseismic eventsThe discriminatingfeature of the ldquonumber of triggersrdquo provides a relativelymedium separation between normal events and blasts

32 Approach II Waveform Features The classificationmethod of mine blasts and microseismic events suing thestarting-up features in seismograms was proposed Signalsfrom databases of different event types were drawn into aunified coordinate system All waveform sections are startingat the point of each P-wave first arrival and ending in theirfirst peak points (Figure 4(a)) It is noticed that the starting-up angle of the two types tends to be concentrated intotwo different intervals Since it is difficult to calculate thestarting-up angle directly due to the inaccuracy of P-wavearrivalrsquos picking the slope value of the starting-up trend lineobtained from linear regression was proposed to substitutethe angle Two slope values associated with the coordinates


70 75 80 85 90 95 100 105 110 115 120000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

log(M)

(a)

0 1 2 3 4 5 6000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

minus2 minus1

log(E)

(b)

00 05 10 15 20 25 30 35 40000005010015020025030035040045050

Freq

uenc

y of

even

ts

minus10 minus05

log(ES EP )

(c)

20 25 30 35 40 45 50 55 60 65000

005

010

015

020

025

030

035

Freq

uenc

y of

even

ts

log(SSD)

(d)

0 2 4 6 8 10 12 14 16 18 20 22 24000

004

008

012

016

020

024

028

032

036

Freq

uenc

y of

even

ts

Time of occurrence


(e)


4 6 8 10 12 14 16 18 20000

003

006

009

012

015

018

021

024

Freq

uenc

y of

even

ts

Number of triggers

(f)

Figure 3 Distributions of the measured features for the database are displayed as histograms The seismic energy is the best performingdiscriminating feature as it provides maximum separation between normal events and blasts The feature ldquotime of occurrencerdquo provides theslightest separation between blasts and normal events



Time (s)

Velo

city

(mmiddotsminus

1)

50times 10minus4

40

30

20

10

00

minus10

minus20

minus30

minus40

minus5040 80

times 10minus3120

(a)

1

14

12

34

y

y

P2i(x2i y2i)

K2

x

x

K1

P11

P12

P13

P14

(b)

Figure 4 Comparison chart of signal starting-up before first peak within blasts and microseismic events (frame (a)) and the schematicdiagram of data points selecting and the trend line constructed by linear regression (frame (b) solid circles represent sampling points andthe reds represent the selected) The x- and y-axis in frame (b) represent the time and the amplitude axis in frame (a)

00 05 10 15000

005

010

015

020

025

030


Freq

uenc

y of

even

ts

minus45 minus40 minus35 minus30 minus25 minus20 minus15 minus10 minus05

log(K1)

(a)


00000

004

008

012

016

020

Freq

uenc

y of

even

ts


log(K2)

(b)

Figure 5 Comparison chart of slope value of the two starting-up trend lines within waveforms of blasts andmicroseismic events Histogramsare used to illustrate the behavior of the discriminating features

of the first peak and the maximum peak were extracted as thecharacteristic parameters

Set 1198701as the absolute value of the slope of the trend line

of the waveform section that is from the P-wave arrival tothe first peak and set 119870

2as the absolute value of the slope

of the trend line of the waveform section that is from the P-wave arrival to the maximum peak The two trend lines areconstructed by linear regression based on four points Themethod of the features extraction and their performance aredisplayed in Figures 4 and 5

4 Method

Bayes discriminant is a branch of modern statistics with thebasic hypothesis that some certain cognitions of the studyingcollectivity had been received before extracting samplesGenerally using the priori probability to describe the level ofawareness and then the posterior probability was obtained bymodifying it [22ndash26]

Suppose G = (X1X2 Xp)119879 is a collectivity with 119901

member indexes and 1198911(X) and 119891

2(X) are the distribution


density functions of the two collectivities G1 which refers toblasts and G2 which refers to microseismic events IndexesX1 X2 X3 X4 andX5 represent the characteristic parame-ters respectively

The priori probability of G1 and G2 are calculated by thefollowing formulas

1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)

where 1198991and 1198992are the number of training samples belonged

to the collectivity G1 and G2Set Σ1 Σ2 and Σ as the covariance matrix of G1 G2

and G The Bayes discriminant function can be expressed asfollows when Σ1 = Σ2 = Σ

119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)

where 120583j is the mean vectors of Gj and the generalizedsquared distance function can be obtained as

1198892

119895

(X) = (X minus 120583j)119879

Σminus1(X minus 120583j) minus 2 ln119901119895 (3)

The a posteriori probability function can be obtained asfollows

119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)

Since pj119891119895(X) prop exp[minus051198892119895

(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)

Normally 1205831 1205832 and Σ are unknown and their estima-tion values 1 2 and Σ can be obtained from the trainingsamples then

1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)

The estimation of a posteriori probability function is

(Gj |X) =exp [minus051198892

119895

(X)]

exp [minus0511988921

(X)] + exp [minus0511988921

(X)] (7)

Bayes discriminant criterion can be expressed as

X isin G1 when (G1 | X) ge (G2 | X)

X isin G2 when (G1 | X) lt (G2 | X) (8)

To estimate the reliability of the discriminator the resub-stitutionmethod was used to calculate the misdiscriminationrate All the training samples were regarded as the newsamples and resubstituted into the classifier The rate ofmisjudgment can be evaluated by the following index

=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912

is the number of training samples regarded to beG2 which belong to collectivity of G1 actually and 119899

21is the

number of training samples discriminated as G1 but belongsto G2 in fact

Table 2 Prior probabilities for groups

Type Prior Cases used in analysisUnweight Weighted

Blast 05 47 470Microseismic event 05 56 560Total 10 103 1030

5 Results

Theaimof the present study is to compare the two approachesfor event accurate identification of different classesTheBayesdiscriminant models for signal identification are establishedafter developing the theory discussed above to the 103 sets ofsamples selected The prior probabilities for different groupsand the classification function coefficients are listed in Tables2 and 3

As can be seen from the classification functions coef-ficients the features of source parameters are in the fol-lowing order according to their importance in the accurateidentification of microseismic events the seismic momentthe static stress drop the seismic energy the S P energyratio the number of triggers and the time of occurrenceAccording to their importance in the accurate identificationof microseismic events the waveform characteristics are inorder of the time of the first peak arrival (119909

11) the amplitude

of the maximum peak (11991021) the slope value of the first trend

line (1198701) the slope value of the second trend line (119870

2)

the amplitude of the first peak (11991011) and the time of the

maximum peak arrival (11990921)

The classification results show that 835 of originalgrouped cases are correctly classified by approach I and971 of original grouped cases are correctly classified byapproach II (Table 4) The results show that the secondfeatures extraction approach (waveform characteristics) hashigher accuracy Although the input data used for classifi-cation in approach II cannot be given directly by the moni-toring systems the calculation process of those characteristicparameters are not complicated

These misclassified events are labeled in Figure 6 accord-ing to the order of moment magnitude Figure 6 shows themisclassified cases of both approaches falling without specificmagnitude intervals It is concluded that the error rate is notaffected by the seismic magnitude or the scale of seismicenergy Although approach I has a higher misclassified ratewith a more complex computation it has been widely used inmany mines Approach II is a new straight-forward methodwith lower error rate and got very good application in theYongshaba mine But in other mines or in more complexcases (mines existing variety of dynamic processes whichradiate seismic waves such as shock and vibrations inducedby orepass) the feasibility of this approach may need furtherimprovements

6 Conclusion

(1) The considerations and the criteria of manual identifi-cation of blasts and microseismic events were summarized


Table 3 Classification function coefficients

Approach I log(119872) log(119864) log(119864S119864P) log(SSD) TOOa NOTb ConstantBlast 18229 minus9827 2964 18790 0625 minus1495 minus11242Microseismic 18248 minus9761 2410 16133 0557 minus1596 minus100464Approach II log(119909

11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702

) ConstantBlast minus92588 3878 minus8471 minus3704 minus33644 minus4304 minus186187Microseismic minus91487 2263 minus10915 minus0344 minus32830 minus7053 minus193783aTOO presents the time of occurrence bNOT presents the number of triggers

Table 4 Classification results

TypePredicted group membership Predicted group membership

(approach I) (approach II) TotalBlast Microseismic events Blast Microseismic events

CountBlasts 39 8 46 1 47Microseismic events 9 47 2 54 56

Blasts 830 170 979 24 1000Microseismic events 161 839 36 964 1000

835 of original grouped cases are correctly classified by approach I and 971 of original grouped cases are correctly classified by approach II

0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)

Figure 6 Events that are misclassified frame (a) misclassified by approach I and frame (b) misclassified by approach II The hollow onesrepresent the misclassified cases

based on experiencesTheguide lines listed in this study coverwell the majority of all the encountered situations in seismicdata processing The sample databases of the two typesrsquoevents were established by three independent processors withreference to these guidelines

(2) Two features extraction approaches were proposedFeatures of source parameters including the seismicmomentthe seismic energy the energy ratio of S- to P-wave thestatic stress drop the time of occurrence and the number oftriggers were selected counted and analyzed in approach I

Waveform characteristics consisting of two slope values andthe coordinates of the first peak and the maximum peak wereextracted as the discriminating parameters in approach II

(3) The discriminating performance of the twoapproaches were compared and discussed by applying theBayes discriminant analysis to the characteristic parametersextracted The classification results show that 835 of theoriginal grouped cases were correctly classified by approachI and 971 of original grouped cases were correctly classifiedby approach II


(4) Comparative analysis shows that the misclassifiedcases of the two approaches all fall without specificmagnitudeintervals The error rate is not affected by the seismic magni-tude Although approach II has got very good application inthe Yongshabamine the feasibility of this approachmay needfurther improvements in other mines or in more complexcases mines existing in variety of dynamic processes whichradiate seismic waves such as shock and vibrations inducedby orepass

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

The authors gratefully acknowledge the financial support ofNational Natural Science Foundation of China (no 5137424411447242) and the Fundamental Research Funds of CentralSouth University (no 2013zzts058)

References

[1] A J Jager and J A Ryder A Handbook on Rock EngineeringPractice for Tabular Hard Rock Mines Creda CommunicationsCape Town South Africa 1999

[2] D Malovichko ldquoDiscrimination of blasts in mine seismologyrdquoin Proceeding of the DeepMining Australian Centre for Geome-chanics Perth Australia 2012

[3] J A Vallejos and S D McKinnon ldquoLogistic regression andneural network classification of seismic recordsrdquo InternationalJournal of Rock Mechanics and Mining Sciences vol 62 pp 86ndash95 2013

[4] S J Gibowicz and A Kijko An Introduction to Mining Seismol-ogy Academic Press San Diego Calif USA 1994

[5] K Larsson Seismicity in Mines A Review Lulea University ofTechnology 2004

[6] P BormannNewManual of Seismological Observatory PracticeGFZ Potsdam Germany 2002

[7] A J Mendecki Seismic Monitoring in Mines Chapman amp HallLondon UK 1997

[8] L-J Dong X-B Li Z-L Zhou G-H Chen and J Ma ldquoThree-dimensional analytical solution of acoustic emission sourcelocation for cuboid monitoring network without pre-measuredwave velocityrdquo Transactions of Nonferrous Metals Society ofChina vol 25 no 1 pp 293ndash302 2015

[9] C Liu C-A Tang J-H Xue and G-F Yu ldquoComprehensiveanalysis method of identifying and calibrating micro-seismicevents attributes in coal and rock massrdquo Journal of Mining ampSafety Engineering vol 28 no 1 pp 61ndash65 2011

[10] Q-J Zhu F-X Jiang Y-M Yin Z-X Yu and J-L WenldquoClassification of mine microseismic events based on wavelet-fractal method and pattern recognitionrdquo Chinese Journal ofGeotechnical Engineering vol 34 no 11 pp 2036ndash2042 2012

[11] F-X Jiang Y-M Yin Q-J Zhu S-X Li and Z-X Yu ldquoFeatureextraction and classification ofminingmicroseismicwaveformsvia multi-channels analysisrdquo Journal of the China Coal Societyvol 39 no 2 pp 229ndash237 2014

[12] P Duplancic Characterization of caving mechanism throughanalysis of stress and seismicity [PhD thesis] University ofWestern Australia Perth Australia 2001

[13] Y Abolfazlzadeh Application of Seismic Monitoring in CavingMines Case Study of Telfer Gold Mine Laurentian UniversityOntario Canada 2013

[14] T I Urbancic B Feignier and R P Young ldquoInfluence of sourceregion properties on scaling relations for 119872 lt 0 eventsrdquo Pureand Applied Geophysics vol 139 no 3-4 pp 721ndash739 1992

[15] M R Hudyma D Heal and P Mikula ldquoSeismic monitoringinminesmdasholdtechnologymdashnewapplicationsrdquo in Proceedings ofthe 1st Australasian Ground Control in Mining Conference pp201ndash218 University of New South Wales Sydney Australia2003

[16] J Boatwright and J B Fletcher ldquoThepartition of radiated energybetween P and S wavesrdquo Bulletin of the Seismological Society ofAmerica vol 74 pp 361ndash376 1984

[17] PM Shearer Introduction to Seismology Cambridge UniversityPress Cambridge Mass USA 1999

[18] L J Dong X B Li and G Xie ldquoAn analytical solution foracoustic emission source location for known P wave velocitysystemrdquoMathematical Problems in Engineering vol 2014 Arti-cle ID 290686 8 pages 2014

[19] X B Li and L J Dong ldquoAn efficient closed-form solutionfor acoustic emission source location in three-dimensionalstructuresrdquo AIP Advances vol 4 no 2 Article ID 027110 2014

[20] L J Dong and X B Li ldquoThree-dimensional analytical solutionof acoustic emission or microseismic source location undercube monitoring networkrdquo Transactions of Nonferrous MetalsSociety of China vol 22 no 12 pp 3087ndash3094 2012

[21] L Dong and X Li ldquoA microseismicacoustic emission sourcelocation method using arrival times of PS waves for unknownvelocity systemrdquo International Journal of Distributed SensorNetworks vol 2013 Article ID 307489 8 pages 2013

[22] S Weisberg Applied Linear Regression John Wiley amp SonsHoboken NJ USA 3rd edition 2005

[23] L Dong and X Li ldquoComprehensive models for evaluatingrockmass stability based on statistical comparisons of multipleclassifiersrdquo Mathematical Problems in Engineering vol 2013Article ID 395096 9 pages 2013

[24] A Bjorck Numerical Methods for Least Squares ProblemsSIAM Philadelphia Pa USA 1996

[25] A R Webb Statistical Pattern Recognition John Wiley amp Sons2nd edition 2002

[26] G J Mclachlan Discriminant Analysis and Statistical PatternRecognition John Wiley amp Sons Malvern UK 1992

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



0 600(m)

N

(a)

netADC4

netADC4

netADC4 netADC4 netADC4 netADC4

netADC4 netADC4

23 24 25 26

netSP+ netSP+

netSP+ netSP+

11 12 13 14 15 16 17 18 19 20 21 22

10 9 8 7 6 5 4 3 T2 2 T1 1

Server

(b)

Figure 1 The microseismic monitoring system installed at Yongshaba mine (a) isometric view of the ground surface tunnels and themicroseismic monitoring system and (b) the configuration of the seismic network used in the project

radiation and the radiation pattern as features and thenestablished the GaussianMaximumLikelihoodClassificationmethod for the classification [2] This method providesa way to identify signals of different type but the greatamount of computation leads to low efficiency Vallejos andMcKinnon proposed the identification of seismic records inseismically activemines by considering the logistic regressionand the neural network classification techniques An efficientmethodology was presented for applying these approaches tothe classification of seismic records [3] However the calcu-lation of seismic source parameters requires precise signalprocessing namely expertise-required and time-consumingP- and S-wave hand-picking Besides the statistical classifi-cationmodels that only rely on source parameters show greatdifference in accuracy when applied to different mine sites

Liu et al proposed a synthesis method to identify micro-seismic event based on the triggering principle of short termaverage to long term average [9] But very small fluctuationsin the threshold would cause a great rate of misclassificationsince the threshold setting completely depends on experi-ence Zhu et al decomposed the microseismic signals into5 layers to gain specified frequency bands using waveletanalysis Based on the box fractal dimensions and thosespecified frequency bands 23-dimensional values of patternrecognition feature vector were established The supportvector machine which was adopted to train classify and rec-ognize shows a correct identification rate of 94 [10] Jianget al presented a three-step strategy to achieve the classifyingof local multichannels microseismic waveforms The authorextracted the time-frequency the amplitude distributionand the correlation coefficient as features and establishedan effective judgment mechanism [11] However the studiesdid not provide much attention to the huge workload ofcalculation in practical application Their input data couldnot be obtained directly from the monitoring system andtheir calculation algorithm is very complicatedThe efficiencyof identification still needs further improvement since anyalgorithm must be as simple as possible in order to run ona small low-power microprocessor

A classification method of mine blasts and microseismicevents suing the starting-up features in seismograms wasproposed It is a method that presents less computationwith a relatively low misjudgment rate In this study theBayes discriminant classifier is applied to further analysethe ldquostarting-up featuresrdquo The differential discriminatingperformances acquired by means of extracting the featuresof respectively source parameters and waveform were com-pared and discussed

2 Data

21 Engineering Background The Yongshaba Mine is locatedat Gaobang which is about 85 km northeast of Guiyang(26∘381015840N 106∘371015840E) Guizhou Province PR China It is themain operation base of the Guizhou Kailin Co Ltd withphosphate production capacity over 2000000 tonsyear Thecurrent exploitation stopes are mainly scattered on levels of1090m to 840m with a relative depth of 500m to 700mbelow the surface Tens of millions of tons has cumulativelyexcavated employing open stoping mining method sinceinitial operating in the 1950s

The study region covers a volume of approximately3000m times 300m times 350m between the depths of 300m and700m below the surface The underground microseismicmonitoring system used to inform the evolution of themicrofracture behavior consists of 26 uniaxial and 2 triaxialgeophones (Figure 1) The geophone manufactured by IMSholds a natural frequency of 14Hz and a sampling rate of6000Hz Signals from various dynamic processes fracturingin rock mass production and development blasts impactsand vibration of machinery all are being recorded by themicroseismic monitoring system

22 Databases The sample databases used in this studyconsist of two parts the subset of the signals of blast eventsand the subset of the signals of normal microseismic eventsEach of the signals in the blast subset has been confirmedto be consistent with the blast operations according to


+ ++ +

(a)

++ minus

minus

(b)













00

02

04

06

08

10

Blasting seismogram

Time (s)

A monotonically


20

10

00

minus10

times 10minus4

minus20

Velo

city

(mmiddotsminus

1)

decreasing tail

a single waveform



Time (s)

00

02

04

06

08

10


Energy curve

Velo

city

(mmiddotsminus

1)

120

80

times 10minus5

40

00

minus40

minus80

100

80

times 10minus7

60

40

20

00



0 4 8 12 16 20 24

000

005

010

015

020

025

Frac

tion

of o

ccur

renc

es

Hour of the day




0 50 100

150

200

250

300

350

400

450

500

00


Am

plitu

de

Frequency (Hz)

100

80

60

40

20

times 10minus6

Fp = 6687



Table 1 Continued



0 10 20 30 40 50


Sample

15

10

times 10minus3

05

00

minus05

minus10

Max

imum

ampl

itude

(mmiddotsminus

1)









70 75 80 85 90 95 100 105 110 115 120000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

log(M)

(a)

0 1 2 3 4 5 6000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

minus2 minus1

log(E)

(b)

00 05 10 15 20 25 30 35 40000005010015020025030035040045050

Freq

uenc

y of

even

ts

minus10 minus05

log(ES EP )

(c)

20 25 30 35 40 45 50 55 60 65000

005

010

015

020

025

030

035

Freq

uenc

y of

even

ts

log(SSD)

(d)

0 2 4 6 8 10 12 14 16 18 20 22 24000

004

008

012

016

020

024

028

032

036

Freq

uenc

y of

even

ts

Time of occurrence


(e)


4 6 8 10 12 14 16 18 20000

003

006

009

012

015

018

021

024

Freq

uenc

y of

even

ts

Number of triggers

(f)




Time (s)

Velo

city

(mmiddotsminus

1)

50times 10minus4

40

30

20

10

00

minus10

minus20

minus30

minus40

minus5040 80

times 10minus3120

(a)

1

14

12

34

y

y

P2i(x2i y2i)

K2

x

x

K1

P11

P12

P13

P14

(b)


00 05 10 15000

005

010

015

020

025

030


Freq

uenc

y of

even

ts


log(K1)

(a)


00000

004

008

012

016

020

Freq

uenc

y of

even

ts


log(K2)

(b)







4 Method








1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)




119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)


1198892

119895

(X) = (X minus 120583j)119879



119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)


(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)


1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)


(Gj |X) =exp [minus051198892

119895

(X)]


(X)] + exp [minus0511988921

(X)] (7)





=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912


21is the





5 Results



11) the amplitude



2)





6 Conclusion





11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










+ ++ +

(a)

++ minus

minus

(b)













00

02

04

06

08

10

Blasting seismogram

Time (s)

A monotonically


20

10

00

minus10

times 10minus4

minus20

Velo

city

(mmiddotsminus

1)

decreasing tail

a single waveform



Time (s)

00

02

04

06

08

10


Energy curve

Velo

city

(mmiddotsminus

1)

120

80

times 10minus5

40

00

minus40

minus80

100

80

times 10minus7

60

40

20

00



0 4 8 12 16 20 24

000

005

010

015

020

025

Frac

tion

of o

ccur

renc

es

Hour of the day




0 50 100

150

200

250

300

350

400

450

500

00


Am

plitu

de

Frequency (Hz)

100

80

60

40

20

times 10minus6

Fp = 6687



Table 1 Continued



0 10 20 30 40 50


Sample

15

10

times 10minus3

05

00

minus05

minus10

Max

imum

ampl

itude

(mmiddotsminus

1)









70 75 80 85 90 95 100 105 110 115 120000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

log(M)

(a)

0 1 2 3 4 5 6000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

minus2 minus1

log(E)

(b)

00 05 10 15 20 25 30 35 40000005010015020025030035040045050

Freq

uenc

y of

even

ts

minus10 minus05

log(ES EP )

(c)

20 25 30 35 40 45 50 55 60 65000

005

010

015

020

025

030

035

Freq

uenc

y of

even

ts

log(SSD)

(d)

0 2 4 6 8 10 12 14 16 18 20 22 24000

004

008

012

016

020

024

028

032

036

Freq

uenc

y of

even

ts

Time of occurrence


(e)


4 6 8 10 12 14 16 18 20000

003

006

009

012

015

018

021

024

Freq

uenc

y of

even

ts

Number of triggers

(f)




Time (s)

Velo

city

(mmiddotsminus

1)

50times 10minus4

40

30

20

10

00

minus10

minus20

minus30

minus40

minus5040 80

times 10minus3120

(a)

1

14

12

34

y

y

P2i(x2i y2i)

K2

x

x

K1

P11

P12

P13

P14

(b)


00 05 10 15000

005

010

015

020

025

030


Freq

uenc

y of

even

ts


log(K1)

(a)


00000

004

008

012

016

020

Freq

uenc

y of

even

ts


log(K2)

(b)







4 Method








1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)




119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)


1198892

119895

(X) = (X minus 120583j)119879



119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)


(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)


1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)


(Gj |X) =exp [minus051198892

119895

(X)]


(X)] + exp [minus0511988921

(X)] (7)





=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912


21is the





5 Results



11) the amplitude



2)





6 Conclusion





11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












00

02

04

06

08

10

Blasting seismogram

Time (s)

A monotonically


20

10

00

minus10

times 10minus4

minus20

Velo

city

(mmiddotsminus

1)

decreasing tail

a single waveform



Time (s)

00

02

04

06

08

10


Energy curve

Velo

city

(mmiddotsminus

1)

120

80

times 10minus5

40

00

minus40

minus80

100

80

times 10minus7

60

40

20

00



0 4 8 12 16 20 24

000

005

010

015

020

025

Frac

tion

of o

ccur

renc

es

Hour of the day




0 50 100

150

200

250

300

350

400

450

500

00


Am

plitu

de

Frequency (Hz)

100

80

60

40

20

times 10minus6

Fp = 6687



Table 1 Continued



0 10 20 30 40 50


Sample

15

10

times 10minus3

05

00

minus05

minus10

Max

imum

ampl

itude

(mmiddotsminus

1)









70 75 80 85 90 95 100 105 110 115 120000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

log(M)

(a)

0 1 2 3 4 5 6000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

minus2 minus1

log(E)

(b)

00 05 10 15 20 25 30 35 40000005010015020025030035040045050

Freq

uenc

y of

even

ts

minus10 minus05

log(ES EP )

(c)

20 25 30 35 40 45 50 55 60 65000

005

010

015

020

025

030

035

Freq

uenc

y of

even

ts

log(SSD)

(d)

0 2 4 6 8 10 12 14 16 18 20 22 24000

004

008

012

016

020

024

028

032

036

Freq

uenc

y of

even

ts

Time of occurrence


(e)


4 6 8 10 12 14 16 18 20000

003

006

009

012

015

018

021

024

Freq

uenc

y of

even

ts

Number of triggers

(f)




Time (s)

Velo

city

(mmiddotsminus

1)

50times 10minus4

40

30

20

10

00

minus10

minus20

minus30

minus40

minus5040 80

times 10minus3120

(a)

1

14

12

34

y

y

P2i(x2i y2i)

K2

x

x

K1

P11

P12

P13

P14

(b)


00 05 10 15000

005

010

015

020

025

030


Freq

uenc

y of

even

ts


log(K1)

(a)


00000

004

008

012

016

020

Freq

uenc

y of

even

ts


log(K2)

(b)







4 Method








1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)




119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)


1198892

119895

(X) = (X minus 120583j)119879



119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)


(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)


1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)


(Gj |X) =exp [minus051198892

119895

(X)]


(X)] + exp [minus0511988921

(X)] (7)





=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912


21is the





5 Results



11) the amplitude



2)





6 Conclusion





11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Table 1 Continued



0 10 20 30 40 50


Sample

15

10

times 10minus3

05

00

minus05

minus10

Max

imum

ampl

itude

(mmiddotsminus

1)









70 75 80 85 90 95 100 105 110 115 120000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

log(M)

(a)

0 1 2 3 4 5 6000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

minus2 minus1

log(E)

(b)

00 05 10 15 20 25 30 35 40000005010015020025030035040045050

Freq

uenc

y of

even

ts

minus10 minus05

log(ES EP )

(c)

20 25 30 35 40 45 50 55 60 65000

005

010

015

020

025

030

035

Freq

uenc

y of

even

ts

log(SSD)

(d)

0 2 4 6 8 10 12 14 16 18 20 22 24000

004

008

012

016

020

024

028

032

036

Freq

uenc

y of

even

ts

Time of occurrence


(e)


4 6 8 10 12 14 16 18 20000

003

006

009

012

015

018

021

024

Freq

uenc

y of

even

ts

Number of triggers

(f)




Time (s)

Velo

city

(mmiddotsminus

1)

50times 10minus4

40

30

20

10

00

minus10

minus20

minus30

minus40

minus5040 80

times 10minus3120

(a)

1

14

12

34

y

y

P2i(x2i y2i)

K2

x

x

K1

P11

P12

P13

P14

(b)


00 05 10 15000

005

010

015

020

025

030


Freq

uenc

y of

even

ts


log(K1)

(a)


00000

004

008

012

016

020

Freq

uenc

y of

even

ts


log(K2)

(b)







4 Method








1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)




119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)


1198892

119895

(X) = (X minus 120583j)119879



119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)


(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)


1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)


(Gj |X) =exp [minus051198892

119895

(X)]


(X)] + exp [minus0511988921

(X)] (7)





=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912


21is the





5 Results



11) the amplitude



2)





6 Conclusion





11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










70 75 80 85 90 95 100 105 110 115 120000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

log(M)

(a)

0 1 2 3 4 5 6000

003

006

009

012

015

018

021

024

027

Freq

uenc

y of

even

ts

minus2 minus1

log(E)

(b)

00 05 10 15 20 25 30 35 40000005010015020025030035040045050

Freq

uenc

y of

even

ts

minus10 minus05

log(ES EP )

(c)

20 25 30 35 40 45 50 55 60 65000

005

010

015

020

025

030

035

Freq

uenc

y of

even

ts

log(SSD)

(d)

0 2 4 6 8 10 12 14 16 18 20 22 24000

004

008

012

016

020

024

028

032

036

Freq

uenc

y of

even

ts

Time of occurrence


(e)


4 6 8 10 12 14 16 18 20000

003

006

009

012

015

018

021

024

Freq

uenc

y of

even

ts

Number of triggers

(f)




Time (s)

Velo

city

(mmiddotsminus

1)

50times 10minus4

40

30

20

10

00

minus10

minus20

minus30

minus40

minus5040 80

times 10minus3120

(a)

1

14

12

34

y

y

P2i(x2i y2i)

K2

x

x

K1

P11

P12

P13

P14

(b)


00 05 10 15000

005

010

015

020

025

030


Freq

uenc

y of

even

ts


log(K1)

(a)


00000

004

008

012

016

020

Freq

uenc

y of

even

ts


log(K2)

(b)







4 Method








1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)




119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)


1198892

119895

(X) = (X minus 120583j)119879



119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)


(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)


1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)


(Gj |X) =exp [minus051198892

119895

(X)]


(X)] + exp [minus0511988921

(X)] (7)





=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912


21is the





5 Results



11) the amplitude



2)





6 Conclusion





11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











Time (s)

Velo

city

(mmiddotsminus

1)

50times 10minus4

40

30

20

10

00

minus10

minus20

minus30

minus40

minus5040 80

times 10minus3120

(a)

1

14

12

34

y

y

P2i(x2i y2i)

K2

x

x

K1

P11

P12

P13

P14

(b)


00 05 10 15000

005

010

015

020

025

030


Freq

uenc

y of

even

ts


log(K1)

(a)


00000

004

008

012

016

020

Freq

uenc

y of

even

ts


log(K2)

(b)







4 Method








1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)




119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)


1198892

119895

(X) = (X minus 120583j)119879



119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)


(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)


1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)


(Gj |X) =exp [minus051198892

119895

(X)]


(X)] + exp [minus0511988921

(X)] (7)





=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912


21is the





5 Results



11) the amplitude



2)





6 Conclusion





11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












1199011= 119875 (G1) =

1198991

1198991+ 1198992

1199012= 119875 (G2) =

1198992

1198991+ 1198992

(1)




119882119895(X) = (Σminus1120583j)

119879

minus 05120583j119879

Σminus1120583j + ln119901

119895 119895 = 1 2 (2)


1198892

119895

(X) = (X minus 120583j)119879



119875 (Gj | X) =119901119895119891119895(X)

11990111198911(X) + 119901

21198912(X)

(4)


(X)] then

119875 (Gj | X) =exp [minus051198892

119895

(X)]exp [minus051198892

1

(X)] + exp [minus0511988921

(X)] (5)


1198892

119895

(X) = (Xminusj)119879

Σminus1

(Xminusj) minus 2 ln119901119895 (6)


(Gj |X) =exp [minus051198892

119895

(X)]


(X)] + exp [minus0511988921

(X)] (7)





=11989912+ 11989921

1198991+ 1198992

(9)

where 11989912


21is the





5 Results



11) the amplitude



2)





6 Conclusion





11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












11

) log(11991011

) log(1198701

) log(11990921

) log(11991021

) log(1198702








0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample


minus15

minus10

minus05

(a)


0 10 20 30 40 50 60

00

05

10

15

20

Mom

ent m

agni

tude

Sample

minus15

minus10

minus05

(b)










Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Acknowledgment


References





























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Research Article A Comparison of Mine Seismic ...

Documents

Transcript of Research Article A Comparison of Mine Seismic ...