COAL MINE SEISMICITY AND BUMPS HISTORICAL CASE STUDIES - PDF document

COAL MINE SEISMICITY AND BUMPS: HISTORICAL CASESTUDIES AND CURRENT FIELD ACTIVITYJohn L. EllenbergerKeith A. HeasleyNational Institute for Occupational Safety and HealthPittsburgh Research LaboratoryPittsburgh, PAABSTRACTThe National Institute for Occupational Safety and Health(NIOSH) has continued the research role of the former U.S.Bureau of Mines to develop techniques that will reduce thehazards in the mining work place associated with coal bumps.Current research focuses on both analyzing historical seismicdata from bump-prone operations and utilizing a mine-wideseismic network to investigate the exact strata failuremechanics associated with bump-prone geology. Theanticipated outcome of this research will be reduced bumpincidences through advanced engineering concepts anddesigns which implement the new understanding of stratabehavior.The analysis of the historic seismic data consists ofcorrelating observed mining seismicity with the geologic andgeometric parameters at the sites. The primary seismicparameters are the timing, location and magnitude of arecorded seismic event. These parameters are correlated withsuch mining parameters as: the overburden, the size of theimmediate gob, the size of the district gob area, etc. Thisdetailed analysis of historical seismic data has provided aninformative quantifiable relationship between many of thespecific mining parameters and the induced seismicity.The second aspect of the coal bump research is theinstrumentation of an appropriate field site to determine themain roof, floor, and gob behavior associated with bumpbehavior. The chosen field site is a deep-cover longwall minewith competent geology in a historically bump-prone area.The primary field instrumentation is a three-dimensional, full-waveform, seismic array with both surface and undergroundsensors surrounding an active multi-panel district. Thepurpose of this seismic array is to determine the timing, theexact location, and the mechanism (tensile fracture, beddingplane slip, etc.) of the failure of the strata surrounding theactive and multi-panel gobs. The preliminary results presentedin this paper help to define the strata failure areas around thelongwall panel.Coal mine bumps, sometimes referred to as outbursts,bursts, or bounces, have been recognized as a serious problemin mining for more than 75 years. For the purposes of thispaper, a bump is defined as a sudden release of geologic strainenergy that results in the expulsion of coal from a rib or pillarin a catastrophic manner. Beginning as early as the 1930's, theU.S. Bureau of Mines conducted research relating to causesand potential mechanism to avoid bumps in coal mines (Current research being conducted by the National Institute forOccupational Safety and Health (NIOSH) remains focused onthe reduction and elimination of bumps.In the past, bumps have been acknowledged as having agreater likelihood of occurrence at depths greater than 300 m(1,000 ft), in the presence of strong roof and/or floor, and whenan unusually strong massive unit exists in the main roof (). Miners working under one or more of these conditions needto be constantly aware of the possibility of a coal bump.Recently, the U.S. coal industry experienced severalconsecutive years with no fatal accidents resulting from bumps,until November of 1996, when three fatalities and fiveadditional serious injuries occurred in a two week period.Coal bumps are often associated with seismic events thatare large enough to be registered by regional seismic networks.However, not every potentially hazardous bump generates aregional seismic event, nor does every mine-induced, regionalseismic event manifest itself as a coal outburst at the seamlevel. In reality, coal bumps are just one subset of mine-induced seismicity, and like mine-induced non-bump events,they can exhibit a wide range of magnitudes and energy release.But, knowing that coal bumps are a subset of mine seismicity,it seems reasonable that analyzing mine seismicity in generalmay provide a better understanding of the causes and control ofmine bumps. Also, there are considerably more documentedmine-induced, non-bump, seismic events than there aredocumented coal bumps. In fact, the seismic network in Utahhas found the most active seismic area since 1962 to be thevicinity of the active coal mining with hundreds of mining-induced events recorded each year (). With this number of events, it would seem possible to develop statisticallysignificant relationships between mining parameters and theoccurrence of the mine-induced seismic events that would notbe evident using the much smaller subset of actual coal bumps.This paper presents a correlation analysis between theobserved mining-induced seismicity and the geologic andgeometric parameters at bump-prone sites. The analysisprovides a number of results which support previous researchand a number of results which may question conventionalwisdom. The paper also presents some of the preliminaryresults from a recent three-dimensional seismic systeminstalled at a bump-prone site.The first seismic study location, Site A, operates in acoalbed varying from 2.4 to 3.6 m (8.0 to 12.0 ft) in thickness.The immediate roof varies, generally consisting of a laminatedgray shale; however, in places, the immediate or main roofvaries from a weak, highly fossilized and slickensided blackshale to a strong siltstone, or sometimes, a sandstone channelsystem.This mine progressed through 10 longwall panelswithout incident, then encountered bumps related toincreasing depth and a sandstone channel system thatresulted in abrupt changes in the physical characteristics ofthe immediate roof. In the following longwall district,multiple-seam interactions in conjunction with sandstonechannels appeared to manifest coal bumps (). Informationin this current report covers the bumps encountered in thefinal longwall district of the mine where neither sandstonechannels nor multiple-seam mining was evident.At least six different gate road designs were utilized atSite A. The coal pillar configuration historically used wasa three-entry yield-abutment design, later being changed toa four-entry yield-abutment-yield design. Pillar dimensionsfrequently varied from one gate road to the next. However,all the three-entry systems were configured similarly, withthe abutment pillar placed adjacent to the longwall panel inthe headgate. In the earlier three-entry designs, thedimensions of both pillars were increased in successivedesigns, to account for increasing overburden (). In thecurrent report, the pillar design was consistent across thestudy area, and had evolved to a four-entry, yield-abutment-yield configuration (figure 1). Topographic relief variedsharply across the panels included in this study, varyingfrom 150 to 660 m (500 to 2,200 ft). Figure 1. Panel and face locations of regional and Local events at Mine A Bump/Seismic EventsFigure 1 shows the face locations at the time when themine records listed 48 different events (the figure shows lessthan 48 face positions because, on several occasions, morethan one event occurred in one day). These events wereinitially recorded on the foreman’s reports because coal wasejected from the face and/or large vibrations were felt from thetailgate/gob area. A large number of theses events appear tohave been smaller local events where some coal was ejectedfrom the face, but no large ground vibrations were evident.On the other hand, 22 of these bump/seismic events registereda magnitude of 2.8 or greater on regional seismic monitoringnetworks. These 22 regional seismic events and the mineparameters at the time of the event are presented in table 1.Analyzing this small database of mine-induced seismic eventsproduced a number of notable correlations.Effect of Overburden on Mine SeismicityDeep cover and/or strong lithologic units have long beenrecognized as necessary contributors to coal bumps (). Withthis in mind, the distribution of the seismic events from Site Awith respect to overburden was produced as shown in figure2. (In order to normalize the number of seismic eventsoccurring at a particular overburden range with relation to theamount of panel that was mined at that overburden range, the“event rate” was determined as the number of events thatoccurred per 300 m (1,000 ft) of panel advance at thatoverburden. For example, 5 events occurring in 2,200 m(7,200 ft) of face advance under overburden ranging from 300to 375 m (1,000 to 1,250 ft), would produce an event rate of0.69 (5/7.2) per 300 m (1,000 ft).) Clearly, the seismic eventsat this site are biased towards the deeper cover. In particular,all of the events occurred when the overburden was greaterthan 500 m (1,625 ft) although over one-third of the extractionarea was under less than this amount of cover.Table 1. Summary of Bump/Seismic Events.Intensity(UK,VPI)Overburden, PanelDistance tooverburden, mSidedistance tosolid fromtailgate, mDistanceto setuproom, mDistance torecoveryroom, mALPSstabilityfactorDistanceto firstsolid, mEnergy(UK, VPI)7/31/19943738L-70140211448220.588229258/1/19943738L-7-34140211638030.588039258/3/19943.5738L-7-46140211937730.58773246510/5/19943.6731L-7-82314021954120.5812299812/23/199N/A, 3.5592L-8186167686910970.74869n/a, 24651/12/19953.5661L-846167610139530.6595324651/12/19953661L-846167610139530.659539251/12/19953661L-846167610139530.659539251/19/19953.7679L-8-24167610858810.6388136481/30/19953.7687L-8-107167611498170.6281736482/4/19952.8, 3.3668L-8-140167612057610.65761625, 16652/4/19952.8, 3.3668L-8-140167612057610.65761625, 16653/11/19954592L-8-472167615274390.7443965683/11/19953.4592L-8-472167615274390.7443920267/22/19952.9, 2.9592L-921619519689980.74968760, 7608/5/19952.8661L-991195110908760.6587662510/25/1993.4, 4.1574L-10433222510039630.779632026, 79904/19/19963.7566L-11277249912667000.7870036485/4/19963.2, 3.7609L-11158249913905760.725761369, 36485/13/19962.7, 3.5653L-11232249914974690.66469513, 24655/13/19962.5, 3.5653L-11232249914974690.66469347, 24655/16/19962668L-11107249915284380.65438130Seismic data were collected and compared from both the University of Kentucky (UK) and Virginia Polytechnic Institute (VPI) systems, single values representing UK data are presented first, when both stations recordevents, UK then VPI are provided. Magnitude type (M) is unknown. 0.20.40.60.81.2Overburden (m)Figure 2. Event rate per 300 m (1,000 ft)vs overburden at Mine A 0.20.40.60.81.21.41.61.8Distance to Solid ( m Figure 4. Event rate per 300 m (1,000 ft) vs distanceto solid in meters at Mine A Distance to Nearest Panel End (m)Figure 3. Distribution of events occurring at Mine Arelative to the nearest panel end of the longwall panels. EventsALPS Stability Factor of Gateroad PillarsFigure 5. Percentage distribution of events by ALPS stabilityfactor of gateroad pillarsDistance to Nearest Panel EndAnother observation that one may make from reviewingfigure 1 is that the face had progressed through a significantportion of the panel when the seismic events occurred.Similarly, few of the seismic events occurred when the facewas very close to the end of a panel. Figure 3 shows theseismic events plotted against face distance to the nearest endof the panel. With this figure and figure 1, it can be seen thatno events occurred during the first 900 m (3,000 ft) of faceadvance in any of the panels in the district. And even takinginto account a number of events that occurred closer to the endof the panels, over 95 percent of the bumps occurred greaterthan 300 m (1,000 ft) from either end of the panels.Side Distance to SolidIn other bump-prone mines, it has been suggested that thewidth of the district gob (over multiple panels) has a distinctinfluence on the occurrence of bumps. In this paper, we referto the total width of the district gob area, as measured from thetailgate across the adjacent gob from previous panels to thestarting edge of the longwall district, as the "side distance tosolid." At this site, no seismic events were observed until thedistance to the side boundary of the district exceeded 1,200 m(4,000 ft) (figure 4). In fact, panels L-5 and L-6 had moreoverburden than panels L-7 through L-11 where the recordedbumps occurred, but the side distance to solid was only 600and 900 m (2,000 and 3,000 ft), respectively.Relationship between Seismic Events and ALPSPillar StabilityIn the past, the occurrence of bumps was shown to berelated to low gateroad stability factors as calculated by theAnalysis of Longwall Pillar Stability (ALPS) program ((The ALPS stability factor is essentially a measure of thestrength-to-load ratio of the gateroad pillar system.) In thisanalysis, the ALPS stability factor was correlated with theseismic activity. In order to accomplish this correlation, theoverburden above each abutment pillar was measured and theALPS stability factor was calculated. Next, factors for theindividual pillars were then grouped into 75 m (250 ft)increments of overburden. Finally, the observed seismicevents were associated with the ALPS stability factor of thetailgate pillar configuration for the face location at the time ofthe event and the results were charted as shown in figure 5.This histogram and table 1 show that the first seismic eventwas not encountered until the gateroad ALPS stability factorwas 0.78, and that all other events occurred when the ALPSstability factor was less than 0.78. (Since the pillar designwas essentially the same within the analyzed longwall district, Overburden Energy per Event Percent of Energy Percent of EventsFigure 6. Distribution of energy per event, percent of energy, andpercent of events at Minethe differences in the ALPS stability factors correlate directlyto depth; therefore, the distribution in figure 5 is essentiallyidentical to the distribution of events with respect tooverburden as shown in figure 1.)Seismic Energy of the EventsUp until this point in the paper, only the occurrence of anevent has been analyzed and the relative size of the events hasbeen ignored. Of course, correlating the size, or energy, of theevents to the mining parameters may reveal some usefulinformation. The regional seismic events are typically reportedin Richter magnitude which is a logarithmic scale with relationto the size/energy of the event. In order to get a betterperspective on the relative magnitude of these events, thelogarithmic Richter magnitudes were converted to pure energyvalues using the following formula:log E + 9.05(1)where: log E = energy released (ergs) and = Richter magnitude This equation (6) has a number of assumptions built intothe derivation because of the empirical nature of the Richtercalculation. Therefore, the absolute energy values calculatedby the equation are somewhat debatable; however, for ourpurposes, the calculated energy values can be consideredconsistent in a relative sense. For the analysis, the averageenergy per event and the percentage of energy released for eachoverburden interval was calculated. The result is presented infigure 6. In terms of energy released, it is notable that theaverage energy per event was highest in the events thatoccurred in the range of 450 to 525 m (1,500 to1,750 ft) andthat average energy per event decreased at greater depths.From this first look a small number of seismic events atMine A, a number of interesting observations were presented:(1) all the bumps occurred under more than 450 m (1,500 ft) ofoverburden; (2) all the bumps occurred where the panel hadadvanced at least 900 m (3,000 ft) and the mined area was alateral distance of greater than 1200 m (4,000 ft) to solid coal;and (3) the energy released per event did not continue toincrease with increasing overburden. These observations areall informative and suggest some geometric limitationsassociated with the seismic events.Some caution needs to be used when looking at theseresults and interpreting the individual effect of the miningparameters. Looking at figure 1, it can be seen that thedeepest overburden generally occurs near the center of thelongwall district. This geometry forces a strong correlationbetween the depth of cover and the middle of the panels.Thus some of the above results concerning the miningdistance from the ends of the panels may just be a cross-correlation with depth.CASE STUDY B AND CAfter finding some useful correlations with the limitednumber of seismic events available at Mine A, it was decidedto continue analyzing seismic events in relation to miningparameters at a couple other mines with considerably morerecorded seismic activity. These two mines have a longhistory of longwall mining and have each extracted numerouspanels causing numerous seismic events to be registered bythe regional seismic system. For our analysis, all of theregional events located within approximately a 5 km (3 mi)radius of the mines was considered to be mine-induced.(Realizing that the average horizontal location error for theregional seismic events was approximately 1.5 km (1 mi), alarge area around the mines was included in order to collectall of the mining related seismic events from the regionaldatabase.) Further, to insure good quality of the eventdetection, only the events with a Richter magnitude of 2.0 orgreater were considered. The event selection was narrowedeven more by including only those events occurring within asix year period of active seismicity, this procedure resulted in623 events associated with the two mines.In order to perform the analysis with these mines, eachmonthly longwall extraction area was considered to be aseparate mining sample. The center of this sample area wasdetermined and the appropriate parameters: advance footage,overburden depth, distance from panel start, distance frompanel end and side distance to solid were assigned to the area.Then, each seismic event within the month of the extractionwas assigned to that sample area. This resulted in 436 eventsassociated with Mine B and 187 events associated with MineEvent Rate Versus OverburdenThe first parameter to be examined for the properties wasthe event rate versus the overburden and this is plotted infigure 7. Similar to Mine A, there is very little activity until 0.05.010.015.020.025.030.0150-225225-300300-375375-450450-525525-600600+Overburden (m)Energy (x10 ergs ) Mine B Mine CFigure 8. Average energy per event (x10distributed by overburden 150-225225-300300-375375-450450-525525-600600+Overburden (m)Event Rate per 300 m (1,000 ft) Mine B Mine CFigure 7. Event rate per 300 m (1,000 ft) versusoverburden at Mines B and Cthe overburden reaches a lower limit of 375 m (1,250 ft). Also as seen at Mine A, the event rate is seen to decrease afterthe maximum rate around 450 m (1,500 ft) of overburden.Intuitively, the seismic event rate might be anticipated toincrease with increasing overburden. The data from all threemines seems to conflict with this hypothesis. All of thesemines reach a maximum event rate between 450 and 525 m(1,500 and 1,750 ft) of overburden and then the rate distinctlydeclines at greater depths.Energy Released Versus OverburdenThe average energy per event by overburden range forMines B and C was calculated using the same equationpresented for Mine A and is plotted in figure 8. Here again,the sample does not indicate that the seismic energy continuesto increase with depth as one might initially anticipate. Inmine C, the highest measured average energy is encountered Figure 9. Mine outline showing locations of surface and undergreound geophones when the overburden is between 300 and 450 m (1,000 and1,500 ft), and in Mine B, the highest measured average energyoccurs when overburden is in the range of 450 to 525 m (1,500to1,750 ft). In both mines, the energy per event generallydrops at greater depths (the only exception being Mine C atgreater than 600 m (2,000 ft) of overburden). When this dropin the average energy per event is coupled with the drop inevent rate with depth (figure 9), it can be determined that over75 percent of the measured seismic energy occurs in theoverburden range of 375 to 525 m (1,250 to 1,750 ft). This brief analysis of mine-induced seismicity hasproduced some interesting results. First, the seismic dataclearly show that very few detectable events occur with lessthan 300 m (1,000 ft) of overburden, and it is not until greaterthan 375 to 450 m (1,250 to 1,500 ft) that the majority ofseismic activity occurs. This behavior almost exactly matchesthe expected occurrence and depth ranges for coal bumps, andsupports the possibility that reducing the overall mineseismicity will similarly reduce the bump frequency.Second, the seismic data indicate that both the event rateand the event energy do not continue to increase with depth,but actually decrease slightly above approximately 450 m(1,500 ft) of cover. This result initially seems to be counterintuitive, but upon further thought, it is not totallyunreasonable to hypothesis that the fracture processes arounda longwall panel may reach a steady state, or even decline,above a certain depth. Certainly, the upper strata will have atendency to bend versus fracturing as it becomes more remotefrom the extraction area. Also, the actual fracturing aroundthe panel may become more plastic and less brittle, therebyreleasing less seismic energy, with higher confinement stressesat greater depths. This aspect of the data poses someinteresting questions and certainly needs to be investigatedfurther.MICROSEISMIC FIELD SITEThe NIOSH project which is presently investigating bumphazards is titled; "Coal Bump Reduction through AdvancedMine Design". The basic research approach of this project isto instrument a deep, bump-prone longwall mine anddetermine the main roof, gob, and floor behavior using a three-dimensional, microseismic system. This microseismic system"listens" to the rock and determines the timing and location ofthe failure of the rock strata surrounding the longwall. Byanalyzing the observed rock failure, researchers can betterunderstand the caving of the massive main roof, thecompaction and load acquisition of the gob, the failure of thefloor, and the stress redistribution in the coalbed andsurrounding strata. With this knowledge, mines can be betterdesigned to reduce dangerous bump occurrences.Throughout the first two years of this project, amicroseismic system was installed over a longwall district ata deep western coal mine. The overburden at this minereaches 900 m (3,000 ft) and the lithology contains severalnotably competent units. In particular, the 180 m (600 ft)thick Castlegate sandstone, known for forming vertical cliffsin the surrounding escarpments, is approximately 165 m (550ft) above the coal seam. The coal bed at the mine ranges from2.4 to 6.0 m (8 to 20 ft) in thickness with an extractionthickness of 2.4 to 3.0 m (8 to 10 ft), and the geologyimmediately surrounding the seam consists of thinner ((10 ft)) layers of siltstones, mudstones, shales, sandstones andcoal.The microseismic system at the mine consists of 23geophones surrounding the coal mine both underground andon the surface. At present, the underground seismic arrayconsists of 14 geophones in the mains and bleeders around thelongwall panels (figure 9). On the surface over the mine,another nine geophones are distributed above the panels. Atthe mine office, a data analysis workstation receives the datafrom both the underground and surface geophones. These dataare automatically analyzed in order to calculate the eventlocations, and then the locations are displayed in relation to amine map on the computer screen for use by mine personnel.At the time this paper was written, some 7,500 seismicevents had been recorded from the first three quarters of panel2. As an initial step in analyzing this information, thelocations of the events were normalized to the advancinglongwall face. The results of this normalization are shown inplan view in figure 10 and in a cross section parallel to theadvance direction in figure 11. This is only preliminary dataand the reader should understand that the exact location ofthese events may change in the future as the location algorithmis optimized based on field calibration data; however, therelative location of the events will remain fairly consistent andcan be used at this time for some initial inferences.From figures 10 and 11, it can be seen that the seismicactivity is generally well in front of the face. This matchesseismic data from other sites () and is interpreted to be theshear failure of the strata due to the forward stress abutmentzone. From the plan view (figure 10), the seismic activity canbe seen to curve back over the gateroads with a little moreseismic activity located over the tailgate pillars than theheadgate pillars. This result is consistent with strata failureabove and below the yielding gateroad pillars, with theincreased activity around the tailgate pillars due to the sideabutment stress. From the side view (figure 11), it can be seenthat the seismic activity is occurring both above and below theseam level. This response matches other field sites and isconsistent with a front abutment that is vertically symmetricabout the coal seam. Also, in figure 11, a slight angle (fromvertical) of the seismic events back into the gob above thepanel can be seen. This type of angle would be expected withthe arching of the principal stress over the caved area in thegob. 1.Rice GS. Bumps in Coal Mines of the Cumberland Field,Kentucky and Virginia-Causes and Remedy. U.S. Bureauof Mines RI 3267, 1935, 36 pp.2.Iannacchione AT, Zelanko JC. Occurrence andRemediation of Coal Mine Bumps: A Historical Review.Paper in Proceedings: Mechanics and Mitigation ofViolent Failure in Coal and Hard-Rock Mines. U.S.Bureau of Mines Spec. Publ. 01-95, 1995, pp. 27-67.3.Zelanko JC, Heasley KA. Evolution of ConventionalGate-Entry Design for Longwall Bump Control: TwoSouthern Appalachian Case Studies. Paper inProceedings of the Mechanics and Mitigation of ViolentFailure in Coal and Hard-Rock Mines. U.S. Bureau ofMines Spec. Publ. 01-95, 1995, pp. 167-180.4.Arabasz WJ, Nava SJ, Phelps TW. Mining Seismicity inthe Wasatch Plateau and Book Cliffs Coal MiningDistricts, Utah, USA-Overview and Update. Proceedingsof the 15th International Conference on Ground Controlin Mining, Golden, CO, August 13-15, 1996, 28 pp.5.Mark C. Pillar Design Methods for Longwall Mining,U.S. Bureau of Mines IC 9247, 1990, 53 pp.6.Kanamori, H., J. Mori, E. Hauksson, T. Heaton, L. K.Hutton, and L. M. Jones. Determination of EarthquakeEnergy Release and ML Using Terrascope. Bull. Seis.Soc. Am. No. 83, 1993, pp. 330-346.7.Hatherly P, Lou X, Dixon R, McKavanagh B, Berry M,Jecny Z, Bugden C. Roof and Goaf Monitoring for StrataControl in Longwall Mining. Final Report ACARP 1 Figure 10. Plan view of event locations, normalized to face location Figure 11. Cross sectional view of event locations, normalized to face locations