You are on page 1of 5

International Journal of Mining Science and Technology 29 (2019) 139–143

Contents lists available at ScienceDirect

International Journal of Mining Science and Technology


journal homepage: www.elsevier.com/locate/ijmst

Investigation of factors influencing roof stability at a Western U.S.


longwall coal mine
Meriel Young a,⇑, Gabriel Walton a, Elizabeth Holley b
a
Department. Geology & Geol. Engineering, Colorado School of Mines, Golden, CO 80401, USA
b
Department Mining Engineering, Colorado School of Mines, Golden, CO 80401, USA

a r t i c l e i n f o a b s t r a c t

Article history: The coal mine roof rating (CMRR) was developed to bridge the gap between geological variation in under-
Received 14 June 2018 ground coal mines and engineering design. The CMRR accounts for the compressive strength of the
Received in revised form 29 July 2018 immediate roof, the shear strength and intensity of any discontinuities present, and the moisture sensi-
Accepted 15 August 2018
tivity of the immediate roof. The CMRR has been widely used and validated in Eastern US coal mines, but
Available online 3 December 2018
it has seen limited application in the Western US. This study focuses on roof behavior at a Western coal
mine (Mine A). Mine A shows significant lateral geological variation, along with localized faulting and a
Keywords:
laterally extensive sandstone channel network. The CMRR is not used to predict roof instability at the
Coal mine roof stability
CMRR
mine. It is, therefore, hypothesized that there are other factors that are correlated with roof instability
Western U.S. coal in underground coal mines that could potentially also be considered in conjunction with the CMRR.
Case study This hypothesis was tested by collecting 30 CMRR measurements at Mine A. At each measurement loca-
tion, a binary record of the roof condition (stable or unstable) was made, and other parameters such as
depth of cover, presence of faulting, and sandstone channels were also recorded. ANOVA tests showed
that the CMRR values and the roof conditions were not strongly correlated, indicating that the CMRR
input criteria are not fully predictive of roof stability at this mine. The CMRR values showed statistically
significant correlations (p less than 0.05) with faulting as well as with location at an intersection. For
areas that had previously experienced roof fall but were currently stable, faulting was correlated with
roof condition (p less than 0.05) only when the condition was classified as unstable.
Ó 2018 Published by Elsevier B.V. on behalf of China University of Mining & Technology. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction and background mine roof stability and be used in engineering design [4]. It is
widely used in the Eastern US as an input for the analysis of long-
Roof fall is one of the greatest hazards facing underground coal wall pillar stability (ALPS) in underground coal mines. It provides
miners [1]. In 2017, 91 lost-time injuries occurred due to roof fall an excellent start with respect to roof stability assessment; how-
in US underground coal mines. An additional 48 roof falls occurred ever, it is arguably not fully comprehensive, nor is it widely used
with no lost days in 2017 [2]. The geology, geometry, timing and in the Western US.
frequency of roof falls can usually give some indication of the cause Currently the CMRR includes the following parameters as inputs:
of failure. Documentation of these variables on a mine-wide basis uniaxial compressive strength (UCS), spacing and persistence of dis-
can make it possible to characterize the combination of factors continuities, cohesion and roughness of discontinuities, moisture
which contribute to a high number of roof falls. Most ground con- sensitivity, presence of a strong bed on the bolted interval, number
trol failures are related to geology, and it is therefore critical to of layers in the roof (bolted interval), groundwater presence, and
understand as much as possible about variation in the mine geol- surcharge of overlying beds. The CMRR is focused on the character
ogy [3]. of the units in the immediate roof. If the roof geology is uniform
The coal mine roof rating (CMRR) classification was developed throughout the mine, the CMRR values throughout the mine should
by U.S. National Institute of Occupational Safety and Health be the similar, thus implying the roof stability is approximately the
(NIOSH) researchers Molinda and Mark to quantify the geological same everywhere in the mine. The CMRR currently focuses only on
description of mine roof into a single value that could indicate geological features. This means that any parameters relating to
stress (such as depth of cover) are not included. It is likely that the
⇑ Corresponding author. current CMRR is best used in conjunction with other analyses to fully
E-mail address: merielyoung@mymail.mines.edu (M. Young). evaluate the potential for roof instability [5,6].

https://doi.org/10.1016/j.ijmst.2018.11.019
2095-2686/Ó 2018 Published by Elsevier B.V. on behalf of China University of Mining & Technology.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
140 M. Young et al. / International Journal of Mining Science and Technology 29 (2019) 139–143

During a visit to a Western U.S. longwall coal mine (Mine A), trate the relationships between CMRR and depth of cover, as well
significant lateral geological variation was observed in the roof. CMRR and surface topography. Additionally, ANOVA was used to
There was also moisture sensitive mudstone at various locations, evaluate the relationships between the CMRR and each of the
as well as locally present sandstone and slickensides. The mine parameters collected (e.g., CMRR vs. depth of cover, CMRR vs. loca-
geologists have mapped the sandstone as a network of channels tion at an intersection, etc.).
spanning the entire mine, and it has been informally observed that Next, the correlation between roof condition and each parame-
where the sandstone channel is present in the immediate roof, the ter was evaluated. ANOVA was used to analyze the correlation with
roof is less stable. The CMRR is not used to predict roof stability at depth of cover and surface topography as these parameters were
Mine A. recorded numerically. However, ANOVA cannot evaluate the corre-
The observations outlined above suggest that there may be lation between two categorical variables. In order to examine the
other factors in addition to those included in the CMRR that could correlation between roof condition (recorded as stable or unstable)
also predict roof instability. Where the input parameters currently and other parameters which were also recorded as binary (e.g.,
considered by the CMRR are constant, there may be other factors sandstone channel presence as yes or no), the fisher exact test
that are correlated with roof instability in underground coal mines was used [7]. This test evaluates the null hypothesis that there is
that could be used in conjunction with the CMRR for roof stability no association between the independent and dependent variables.
assessment. By investigating a single case study, this paper pro- If the p value generated is less than 0.05, the null hypothesis may
vides an initial discussion of the CMRR’s comprehensiveness and be rejected at 95% confidence.
how it may be improved. A much larger data set would be required Two sets of analyses were performed for the sites where a roof
in order to make any definitive conclusions. fall had occurred but was now considered to be stable: first, the
statistics were run with these locations recorded as stable, and sec-
ond, the statistics were run with these locations recorded as
2. Methods unstable.

To test the hypothesis that the parameters currently considered


by the CMRR do not fully explain roof instability, CMRR values
3. Results and discussion
were collected at 30 sites at Mine A. At each site, the roof condition
was recorded as stable or unstable. A suite of other parameters
Fig. 1 shows the results from the ANOVA analysis evaluating
were recorded in addition to the CMRR inputs, including slope
correlation between the CMRR and roof condition. The p value
grade of surface topography, presence of faulting, localization of
for the analysis was calculated to be 0.5, indicating there is little
the faulting, presence of sandstone channels, location at an inter-
correlation between roof condition and CMRR at Mine A. Thus it
section, rib and floor condition, and evidence of horizontal stress-
is likely that there are other factors in addition to those included
induced damage. At Mine A, there were eight locations for which
in the CMRR that are indicative of roof instability at Mine A.
drill core and core logs were still accessible. Since the CMRR may
Figs. 2–6 show the results from the ANOVA analyses in Matlab
be calculated from both drill core and underground exposures,
evaluating correlation between CMRR and the other parameters
the CMRR at these locations was calculated using both methods.
recorded at Mine A. The p values for each of these analyses are
The other sites at Mine A were selected randomly to represent
summarized in Table 1.
the spectrum of geological conditions at the mine, including the
The data presented above show that the CMRR value is corre-
presence or absence of sandstone channels in the immediate roof.
lated with the presence of faulting and the location of the data col-
The CMRR component values were recorded using the methods
lection site at an intersection. The CMRR values are not correlated
outlined by Molinda and Mark [4]. The roof condition was also
with the depth of cover, gradient of surface topography, and pres-
recorded as stable or unstable at each site. There is undoubtedly
ence of the sandstone channel in the immediate roof. The CMRR
a spectrum between stable and unstable; however, in order to
does not account for the depth of cover surface topography or
examine correlation, binary measurements were deemed simplest.
large-scale geological features, such as sandstone channels. The
It is important to note that this assessment is subjective and cre-
CMRR does not explicitly account for roof being weaker at an inter-
ates a significant source of uncertainty because the difference
section. However, the roof at intersections may be recorded as hav-
between stable and unstable conditions could be subtle in some
ing a lower strength, cohesion, and friction angle than the roof in
cases. Where there was evidence of significant skin failure or high
the surrounding area, in which case the CMRR would implicitly
levels of support installed, the roof was recorded as unstable.
capture this parameter. Although the CMRR does not explicitly
Where minimal support was required and there was little evidence
record fault displacement or slip, these features appear to be cap-
of skin failure or instability, the roof was recorded as stable. The
tured in the CMRR parameter which records the presence of
depth of cover and slope grade of surface topography were deter-
slickensides.
mined from pre-existing mine maps. The other parameters
described above (if applicable) were recorded individually at each
location.
The data from Mine A were analyzed for correlation between
the CMRR and roof condition. A strong correlation would suggest
that the CMRR includes the most influential parameters that are
predictive of roof instability at Mine A. If there is little correlation
between CMRR and roof condition, it is likely that there are other
parameters that influence roof instability at Mine A and merit fur-
ther investigation. An analysis of variance (ANOVA) was performed
in Matlab to test the null hypothesis that the roof condition and
CMRR are not related. For a p value less than 0.05, the null hypoth-
esis may be rejected at 95% confidence, indicating that any
observed correlation between stability and CMRR values is not Fig. 1. Comparative boxplot result from the ANOVA analysis of CMRR and roof
likely due to random chance. Scatter plots were generated to illus- condition, p = 0.5.
M. Young et al. / International Journal of Mining Science and Technology 29 (2019) 139–143 141

Fig. 6. Scatter plot showing the relationship between CMRR and gradient of surface
topography, p = 0.9 (linear regression).

Fig. 2. Comparative boxplot result from the ANOVA analysis of CMRR and faulting,
p = 0.007.
Table 1
Summary of p values for ANOVA analysis of CMRR
and the parameters collected at each site.

Parameter evaluated with CMRR p value


Roof condition 0.5
Faulting 0.007
Sandstone channel 0.23
Intersections 0.005
Depth of cover 0.54
Surface topography 0.91

Table 2
Summary of p values for the analysis of roof condition and
the parameters collected at each site.
Fig. 3. Comparative boxplot result from the ANOVA analysis of CMRR and the
presence of the sandstone channel in the immediate roof, p = 0.23. Parameter evaluated with roof condition p value
Faulting 0.34
Sandstone channel 0.69
Intersections 0.69
Depth of cover 0.44
Surface topography 0.71

Fig. 4. Comparative boxplot result from the ANOVA analysis of CMRR and the
measurement location, p = 0.005.

Fig. 7. Comparative boxplot from the ANOVA analysis of roof condition and the
depth of cover, p = 0.44.

Fig. 5. Scatter plot showing the relationship between CMRR and depth of cover,
p = 0.54 (linear regression).

Perhaps the more important question is whether individual


parameters are correlated with actual roof condition. In order to
address this, the correlations between the roof condition and the
non-CMRR parameters were evaluated using the ANOVA analysis Fig. 8. Comparative boxplot from the ANOVA analysis of roof condition and the
and the Fisher exact test. The p values from each analysis are pre- surface topography, p = 0.71.
sented in Table 2, and the boxplot results from the ANOVA analyses
are shown in Figs. 7 and 8. condition and the presence of faulting, despite the fact that these
The statistical results do not show a correlation between roof parameters are well correlated with the CMRR. The depth of cover
condition and location at an intersection, or between roof and surface topography also show little correlation with the roof
142 M. Young et al. / International Journal of Mining Science and Technology 29 (2019) 139–143

condition; this could be because the geology is the primary control For the data points collected from stable sites that had previously
on roof condition at Mine A, overshadowing any effect of stress- experienced roof fall, Table 3 shows that rerunning the analysis with
related parameters. the roof classified as unstable caused the correlation between fault-
Lastly, the correlations between roof condition and the CMRR ing and roof condition to become statistically significant (for a sig-
(Fig. 1) and between roof condition and other parameters (Table 2) nificance threshold of a = 0.05). The p values for the correlation
were reanalyzed using the ‘‘unstable” classification for the six sites with sandstone channels and surface topography also decreased,
where the roof was currently considered stable but had seen signif- although these results are still not statistically significant. The p-
icant roof instability in the past (Fig. 9 and Table 3). values increased slightly for the correlations between roof stability
Figs. 10 and 11 show the boxplot results from these ANOVA and intersections, as well as between roof stability and depth of
analyses of the amended roof condition with depth of cover and cover. The results in Table 3 also illustrate how much uncertainty
slope angle respectively. is introduced through the binary recording of the roof condition.
Faulting was the only parameter to show a statistically signifi-
cant correlation with roof condition at Mine A, and only for the ver-
sion of the dataset wherein currently stable areas with previous
roof fall were classified as unstable. It is possible that there is insuf-
ficient data or a bias in the data collection locations, which resulted
in data that is not fully representative of Mine A. It is also possible
that the other parameters show more significant relationships with
roof condition at other mines. This potential variability between
mines makes it difficult to have a ‘‘one size fits all” approach. A
much larger data set spanning multiple mines is needed to make
any definitive conclusions about which additional parameters out-
side the CMRR could be used to predict roof failure, and how
stress-related parameters should (or should not) be combined with
Fig. 9. Comparative boxplot result from the ANOVA analysis of CMRR and roof
condition, p = 0.2.
the CMRR for roof stability assessment.

4. Conclusions
Table 3
Summary of p values for the analysis of roof condition and
the parameters collected at each site with stable areas that
The data from one Western U.S. longwall coal mine (Mine A)
had previously experienced roof fall classified as ‘‘unstable” show no statistically significant correlation between the CMRR
and the recorded roof condition using a binary stability classifica-
Parameter evaluated with roof condition p value
tion. This implies that there are other parameters in addition to
Faulting 0.03 those included in the CMRR that are indicative of roof instability
Sandstone channel 0.48
Intersections 0.72
at Mine A. The CMRR values did correlate with the presence of
Depth of cover 0.53 faulting and location at an intersection. Roof locations that are cur-
Surface topography 0.21 rently stable but had previously been unstable showed a statisti-
cally significant correlation with faulting when the unstable
condition was used in the statistical analysis. The analyses of areas
previously affected by roof fall illustrate the uncertainty intro-
duced as a result of recording a time-sensitive characteristic such
as roof condition as binary data (stable or unstable). The data also
suggest that faulting is the main control on roof stability at Mine A;
and the other parameters did not show any statistically significant
correlation with roof condition.

5. Future work

The same data collection and analysis procedure should be car-


ried out at other mines to broaden the data pool and allow compar-
Fig. 10. Comparative boxplot from the ANOVA analysis of roof condition and the ison of results with those obtained at Mine A. In order to determine
depth of cover, p = 0.40. the sensitivity of the CMRR classification system to each of its
input parameters, the CMRR values should be recalculated at each
location for all but one of the inputs (e.g. calculate the CMRR with-
out the moisture sensitivity parameter or the discontinuity inten-
sity parameter). The correlation should be evaluated between the
according CMRR output and roof condition, as well as with the
other parameters. This will give an indication of how each con-
stituent in the CMRR influences the CMRR output values. Lastly,
the dataset should be reexamined with gradational rather than
binary classification of roof stability.

Acknowledgments

Fig. 11. Comparative boxplot from the ANOVA analysis of roof condition and the This work was supported by a NIOSH Capacity Building grant
slope angle, p = 0.21. (No. 200-2016-90154) to Drs. G. Walton and E. Holley and
M. Young et al. / International Journal of Mining Science and Technology 29 (2019) 139–143 143

collaborators at the Colorado School of Mines. The data were col- [2] Mining MSHA. Industry Accident, Injuries, Employment and Production
Statistics and Reports. U.S. Department of Labor Mine Safety and Health
lected as part of an MS project by Meriel Young, who acknowledges
Administration; 2018.
personnel at Mine A for their support during the data collection [3] Peng SS. Topical areas of research needs in ground control–a state of the art
process. The authors also thank Mark Larson and Bo Kim at NIOSH review on coal mine ground control. Int J Min Sci Technol 2015;25(1):1–6.
for their continued support and input on ground control research [4] Molinda GM, Mark C. Coal Mine Roof rating (CMRR): A Practical Rock Mass
Classification for Coal Mines. Vol. 9387 of Information Circular. U.S. Department
efforts at the Colorado School of Mines. The authors would also like of Interior, Bureau of Mines; 1994. p. 83.
to thank Chris Mark of MSHA for accommodating Meriel Young’s [5] Calleja J. Rapid Rating using coal mine roof rating to provide rapid mine roof
visit to the Pittsburgh facility in Spring 2018 and taking the time characterization from exploration drilling. Proceedings of the Coal Operators’
Conference. Wollongong, Australia. University of Wollongong; 2006.
to discuss the CMRR with her in detail. [6] Hill D. Practical experiences with application of the coal mine roof rating
(CMRR) in Australian coal mines. In: International Workshop on Rock Mass
References Classification in Underground Mining. U.S. Department of Health and Human
Services: National Institute for Occupational Safety and Health; 2007. p. 65–72.
[7] Freeman JV, Campbell MJ. The Analysis of Categorical Data: Fisher’s Exact
[1] Barczak TM, Dolinar DR, Mark C, Signer SP, Tuchman RJ, Wopat PF. Proceedings:
Test. Scope; 2007. p. 16.
New Technology for Coal Mine Roof Support. Information Circular 9453. U.S.
Department of Health and Human Services: National Institute for Occupational
Safety and Health; 2000. p. 274.

You might also like