You are on page 1of 16

Arabian Journal of Geosciences (2021) 14:362

https://doi.org/10.1007/s12517-021-06637-y

ORIGINAL PAPER

A probability-based risk assessment of roof strata weighting


in longwall panels
Sunny Murmu 1 & Gnananandh Budi 1

Received: 15 September 2020 / Accepted: 23 January 2021


# Saudi Society for Geosciences 2021

Abstract
Weighting of the roof strata in a longwall panel poses serious stability concerns at the working face. Therefore, it is essential to
estimate weighting in terms of its interval. Furthermore, difficulty arises during the prediction of roof weighting as rock mass
contains inherent uncertainties. In order to cater to this need, probabilistic analysis has been employed. The article aims to
decipher the suitability of Monte Carlo simulation technique for the assessment of caving of roof strata. In the current study,
firstly, regression models have been developed to estimate periodic weighting interval and first significant main roof weighting
interval using data from thirty (30) longwall panels in India. The regression models were then used as deterministic mathematical
models for Monte Carlo simulation. It has been observed that the Monte Carlo simulation technique is an acceptable means to
evaluate the uncertainties arising from varying geo-mining conditions. In addition to it, sensitivity analysis results exhibited that
the tensile strength and unit weight of the main roof is the key factor that governs the interval and severity of the periodic
weighting and the first significant main roof weighting.

Keywords Longwall mining . Periodic weighting . First significant main roof weighting . Statistical analysis . Monte Carlo
simulation

Introduction serious implications in terms of safety, to men and equipment,


environment degradation, financial performance, exposure to
Longwall method has been globally accepted as the most legal action and corporate reputation (Galvin 2016). The
profitable method of underground excavation for deep coal cavability of the longwall overlying strata is quantified by
deposits. The caving phenomenon, namely the periodic two terms; the periodic weighting interval (PWI) and the main
weighting, and the first significant main roof weighting are roof weighting interval (MRWI). The longer the length of the
the two most anticipated events in the life of a longwall panel intervals, the higher is the order of the severity of roof loading
as far as the safety and stability of the longwall face is con- at the longwall face. Several redundant occurrences like coal
cerned. Due to inherent uncertainties and peculiarities associ- wall block spalling, roof falls, water seepage, dynamic loading
ated with the nature and behaviour of the overlying roof strata, on the shield, several shield closure, bursting of hoses and
high risk is involved in correctly estimating the consequences damage to the structural components of the shields can be
that arise from the aforementioned events. In the area of observed during the weighting events.
ground control, loss in stability at the longwall face can render An umpteen number of researchers have attempted to un-
derstand the geo-mechanics of caving behaviour of longwall
Responsible Editor: Zeynal Abiddin Erguler overlying roof strata. A brief overview suggests that during
the extraction of a longwall panel when a void of sufficient
* Sunny Murmu dimension is created, the overlying strata above the panel gets
sunny.murmu1@gmail.com perturbed due to the re-distribution of the pre-mining stresses
(Peng and Chiang 1984). This results in inducing tensile frac-
Gnananandh Budi turing of the roof layers, delamination and opening up of pre-
anandbudi@gmail.com
existing natural discontinuities and shearing and slipping in
1
Department of Mining Engineering, Indian Institute of Technology
the bedding planes and the freshly formed fractures, thereby
(Indian School of Mines), Dhanbad 826004, Jharkhand, India inducing caving (Brady and Brown 2004). During the
362 Page 2 of 16 Arab J Geosci (2021) 14:362

commencement of production from the set-up entry, after the classify the severity of roof weighting by employing Monte
face is advanced to a sufficient distance, the immediate roof Carlo simulation and provide insights for meaningful roof
separates from the main roof and falls into the gob. This is weighting control suggestions during panel operation.
known as the local fall or the first weighting of the immediate Caving of the overlying roof strata is a complex process
roof (Peng and Chiang 1984; Singh and Singh 2009). This and results from interaction amongst several factors (Table 2).
process continues until a sufficient distance is reached for The degree of the severity of caving depends on its interval
the onset of the first weighting of the main roof. At this stage, and hence is the key to control the stability of the longwall
the main roof acts as a simply supported beam experiencing face. Therefore, the severity of loading at the face by estimat-
maximum tensile stress and bending at the mid span. As a ing PWI and MRWI is studied through the Monte Carlo (MC)
result of loss of support from the immediate roof, the main simulation technique.
roof undergoes shear and tensile fracturing thereby inducing
dynamic load at the face support, i.e. shields. As the main roof
breaks and subsides at the caved immediate roof, the shields Roof weighting and pressure arch
experience load relaxation. This is the first significant
weighting of the main roof. Subsequent to this event, the main There exists pressure arches in the rock mass surrounding the
roof cantilevers behind the face and periodically breaks into longwall excavation, both on strike and dip direction (Fig. 2).
the gob. This is known as the periodic weighting. Figure 1 These pressure arches can also be collectively called as stress
renders a better representation of the weighting events. shell (Xie et al. 2009). The two arches when combined togeth-
Several studies have been conducted to study and estimate er form an ellipsoidal stress shell in 3D. The principal stresses
the PWI and MRWI (Table 1). It has been observed that most of the pressure arch are greater than the rock mass lying inside
of the studies have been conducted through analytical and outside of the shell. There are three components to the
methods or physical modelling which bear several limitations pressure arch: (a) the skew back, (b) the vault and (c) the
of numerous assumptions. Nevertheless, numerical modelling shoulders. The shape of the pressure arch is not constant but
has provided better insights in understanding the caving be- changes all along the process of mining (Wang et al. 2018).
haviour of overlying roof strata. The empirical methods have The stress level is highest at the skewback followed by at the
been done through observation in a particular site and cannot vault and least at the shoulders. The portion of these compo-
be widely applicable. Nonetheless, empirical study through nents that rests in the coal seam is known as abutments. The
assessing data from numerous mines has been found lacking. destressed zone lies within the inner boundary of the pressure
Additionally, classification of weighting severity using field arch and is known as the negative camber of the shell. The
observed sufficient data is lacking. Therefore, this study aims overlying strata above the outer boundary of the pressure arch
to develop empirical models to ascertain PWI and MRWI by do not impose stress on the negative camber but get transmit-
assessing 30 longwall panels representing different geological ted to the abutments through the arches. Therefore, the pres-
and geo-mining conditions. Along with it, it also aims to sure arch act as a primary support that bears the partial load of

Table 1 Previous studies


pertaining to periodic and first Models to estimate first weighting and periodic weighting of main roof
significant main roof weighting
Empirical Analytical Numerical Simulation

• Bilinski and Konopko (1973) • Obert and Duvall (1967) • Gale and Nemcik (1998)
• Sarkar and Dhar (1993) • Kuznetsov et al. (1973) • Singh and Singh (2009)
• Sarkar (1998) • Peng and Chiang (1984) • Singh and Singh (2010a)
• Singh et al. (2004) • Mukherjee (2003) • Singh and Singh (2010b)
• Banerjee et al. (2016) • Ju et al. (2015) • Manteghi et al. (2012)
• Banerjee et al. (2016) • Hosseini et al. (2014)
• Luan et al. (2019) • Jabinpoor and Jafari (2013)
• Zhao et al. (2019) • Shabanimashcool et al. (2014)
• Xu et al. (2015)
• Banerjee et al. (2016)
• Gao et al. (2014)
• Mohammadi et al. (2019)
• Li et al. (2019)
Arab J Geosci (2021) 14:362 Page 3 of 16 362

Fig. 1 Roof caving process


depicting first significant main
roof weighting and periodic
weighting (modified after Peng
and Chiang 1984)

the strata. Hence, it can be said that the fracturing of the main idea about the particular mine from where it has been collect-
roof is due to its own weight located in the negative camber, ed. Thus, compilation of data set is a cumbersome task. In
i.e. the low stressed zone gives rise to periodic weighting and India, the establishment of longwall mining has completed
is not the cause of great dynamic pressure that comes onto the almost four decades. Until now, thirty-three mines have been
face. Only the imbalance of the pressure arch results in dy- worked out with the longwall powered support technology. In
namic and violent strata behaviour thus giving rise to rock this study, 30 sets of data have been collected from 30
bursts, air blasts and shock bumps. The face and the gate roads longwall panels spread across the Indian sub-continent. The
lie within the low stressed zone of the pressure arch. minimum number of data (n) set has been found statistically
Therefore, the primary role of the powered support is to main- sufficient using the following equation:
tain the integrity of the face by providing resistance only to the  0 2
negative camber. z :σ
n≥ ð1Þ
MOE

where z′ is the value corresponding to the desired confidence


Data acquisition interval, σ is the sample standard deviation and MOE is the
margin of error. The z′ value corresponding to 95% confi-
In the domain of ground control, statistical data obtained from dence interval, standard deviation and MOE has been taken
field are highly valuable. It can be said that one panel can as 1.96, 4.26 and 2 m for PWI. Similarly for MRWI, the
provide a single data set. Therefore, each data set gives an values of z′, σ and MOE are 1.96, 25.40 m and 9 m that have
362 Page 4 of 16 Arab J Geosci (2021) 14:362

Table 2 Factors influencing the


cavability of roof strata Cavability of overburden Main roof characteristics • Strength properties
roof • Thickness
• Density
• Stiffness
Immediate roof characteristics • Strength properties
• Thickness
• Density
• Stiffness
• Bulking factor
Discontinuity and structural • Joint direction
characteristics • Persistence
• Filling
• Aperture
• No. of joint sets
• Dip angle
• Stiffness
• Faults, dykes and folds
Seam characteristics • Strength
• Seam inclination
• Presence of cleats
In situ stress condition • Horizontal and vertical stress regime
Panel geometrical characteristics • Face length
• Panel length
• Height of extraction
Operational characteristics • Rate of face advance
• Canopy to face width
• Setting and yielding pressure of powered
support

Fig. 2 Cross sectional view of the


pressure arches (modified after
Wang et al. (2015))
Arab J Geosci (2021) 14:362 Page 5 of 16 362

been assumed, respectively. Based on the above assumptions Table 4 Descriptive statistics of the predictor and predicted variables
the minimum data set required for the analysis of PWI and Variables Count Range Minimum Maximum Mean Std. Dev
MRWI were 18 and 30. Therefore, the requirement for mini-
mum number of data set is met statistically. The MOE for PWI (m) 30 19 8 27 16.77 4.26
MRWI is more as compared to PWI and the information avail- MRWI (m) 30 112 25 137 74.13 25.40
able for PWI is more when compared to MRWI. Therefore, γ (T/m3) 30 0.39 2.04 2.43 2.23 0.08
this research provides more confidence for estimating PWI G (GPa) 30 3.5 0.49 3.99 1.85 0.98
rather than MRWI. Tm (m) 30 43.8 5 48.8 17.28 8.44
Each longwall panel is unique in itself and represents a par- σt,P (MPa) 30 6.04 0.13 6.17 1.15 1.15
ticular geological or geo-mining setting. Some critical factors σt,M (MPa) 30 3.74 2.21 5.95 4.17 0.90
have been identified from Table 2 and have been utilized for
development of empirical models to predict PWI and MRWI.
The statistical data collected from these mines have created a
unique opportunity to assess cavability and make future guide-
lines. A sample and the descriptive statistics of the data set are & Does the addition of the variable increase the residual sum
shown in Tables 3 and 4. of squared error?
& Does the inclusion of the variable increase the adjusted R2
or not?
& Is the inclusion of the variable statistically significant as
Deterministic analysis: statistical model per partial F-tests?

Variable selection Based on the above defined criteria, the independent vari-
ables that have been finally selected are γ, G and σt,P for PWI
In order to develop models for prediction of PWI and MRWI, model and γ, Tm and σt,M for MRWI model. The effective
several factors have been identified (Table 2). The employment tensile strength σt,P and σt,M can be calculated as per Singh
of all the factors would improve the precision of the model but et al. (2004). The tensile strength of the main roof had to be
increase the complexity. In addition to it, the relationship be- converted to attribute the rock mass tensile strength and has
tween several factors with PWI and MRWI is also difficult to been scaled down by 0.5 and multiplied by RQD. Moreover,
decipher through statistical analysis due to the inherent noise and the shear modulus has been derived using standard constitu-
non-linearity. In order to make the models simple and feasible, tive laws pertaining to Poisson’s ratio and deformation mod-
only the most critical parameters have been selected based on ulus. The deformation modulus of the main roof has been
literature review and the ease with which it can be obtained from achieved by converting the intact rock Young’s modulus
the field. Several statistical criteria have been applied for selec- using the following equation (Zhang and Einstein 2004).
tion of the most appropriate variable that are statistically signifi- Em
cant to predict PWI and MRWI. The following statistical criteria ¼ 100:0186*RQD−1:91 ð2Þ
Er
have been applied for variable selection:

Table 3 Data sub-set used for


developing the regression models Mine Panel PWI (m) MRWI (m) G (GPa) γ (T/m3) Tm (m) σt,P (MPa) σt,M (MPa)
for estimating PWI and MRWI
PVK5 21 12 45 2.40 2.25 15.94 0.13 4.09
ALP 1 22 83 2.36 2.15 26 1.12 5.34
VK7 1 14 55 0.58 2.19 14.1 0.55 3.59
VK7 4 17 81 0.67 2.18 14.1 0.56 3.68
JK5 3 15 53 1.70 2.23 20.8 0.68 4.12
Jhanjra AW3 18 65 2.72 2.21 12.97 1.71 3.97
Jhanjra AW8 16 83 3.35 2.23 21.6 1.52 4.70
JK5 1 18 75 1.55 2.25 13.84 0.58 3.46
Khottadih 1 23 104 2.01 2.15 17.41 1.20 5.28

Here G is shear modulus; γ is unit weight; Tm is the thickness of main roof, σt,P is the effective tensile strength of
the main roof during periodic weighting and σt,M is the effective tensile strength of the main roof during first
significant main roof weighting
362 Page 6 of 16 Arab J Geosci (2021) 14:362

The parameters which have been finally selected justify the be 0.79 and 0.71, respectively. In view of the fair R2 obtained for
physical law of roof weighting as γ attributes to the vertical stress Eqs. 3 and 4, it signifies that there are several other parameters
(stress factor) that gets exerted on the face due to the weight of which can be included in the regression models for better predic-
the main roof and not the strata overlying the pressure arch. The tion of PWI and MRWI, respectively. A future work could include
shear modulus and tensile strength attributes to the resistance the identification of the parameters and refining the models.
offered by the main roof to shear and tensile fracture formation
and the thickness of the main roof displays its massiveness. Validation

Multicollinearity diagnostic The validation for the PWI and MRWI models has been done by
its application to predict the PWI and MRWI for four longwall
The presence of multicollinearity amongst the independent var- panels, namely, Balrampur Colliery of South Eastern Coalfields
iables leads to unreliable estimate of the regression weights or Limited (SECL) panel P1, Jhanjra Colliery of Eastern Coalfields
coefficients. This further leads to model formation that are unable Limited (ECL) panels AE1 and AW4, Venkatesh Khani 7 mine
to rightly predict the state variable for a new set of data. of Singareni Collieries Company Limited panel 9. The brief
Therefore, the diagnosis of multicollinearity amongst the inde- summary of the geo-mining conditions of the panels is presented
pendent variables is assessed by determining their variance infla- in Table 8. A quick assessment of the model’s predictive capa-
tion factor (VIF) and tolerance (Liu, Song, and Li 2015). The bility can be made by looking at Fig. 3a and 3b. It is noticeable
threshold value for VIF and tolerance has been suggested from from Fig. 3 that the field measured values and the predicted
Dormann et al. (2013). In Table 5, it can be noticed that no values of PWI and MRWI are in close agreement to each other
predictor variables cross the threshold value of VIF greater than as compared to the other existing equations in the literature. The
10 and tolerance less than 1. Thus, it can be inferred that the error percentage achieved by the proposed equations is always
variables lack multicollinearity and can be utilized to develop lesser then the equations that are available in the literature.
the regression models. Therefore, it can be deduced that the models can be used for
ascertaining the PWI and MRWI of a particular panel during
Multiple regression its design stage.

The relationship between each independent variable and their


respective dependent variable has been deciphered through sim- Probabilistic analysis
ple scatter plots. The relationship having the highest coefficient
of determination (R2) has been selected and hence has been Background
included in the regression model. In general, all the independent
variables have been observed to follow a non-linear relationship There exist inherent uncertainties and variability in the geo-mining
with the predicted variable. Therefore, the equations proposed and geotechnical properties of the overlying roof strata of longwall
are as follows: panels making it difficult to accurately predict the PWI and
MRWI. In other words, it can be plainly stated that the stiffness
PWI ¼ 100:39 þ 3:25*σt;P þ 0:95*G−39:92*γ ð3Þ or strength of rock specimen obtained from the same sample when
1:22 tested in the laboratory shows wide variation. In addition to it, the
MRWI ¼ 48:57*ðT m Þ0:08 * σt;M *ðγ Þ−1:98 ð4Þ
thickness of the immediate and main roof layers, depth of the
The summary of the regression is shown in Tables 6 and 7. The seam, height of extraction, seam thickness, etc. are not constant
R2 values for the PWI and MRWI model have been worked out to throughout the panel length. The empirical approaches are simple
equations that provide first-hand information regarding cavability.
These deterministic models utilize a single set of input data to
Table 5 Multi- predict a state variable. However, there are several limitations to
collinearity diagnostic Variables Tolerance VIF
such approach, as different assumptions are made based on intui-
parameter values for the
independent variables PWI tion, experience and engineering judgement. This kind of ap-
γ 0.91 1.09 proach leads towards several risks and uncertainties. Therefore, it
G 0.39 2.58 becomes a necessity to carry out probabilistic analysis, which will
σt,P 0.37 2.73 provide a better estimation of PWI and MRWI so that risks can be
MRWI circumvented as much as possible. In order to cater to this require-
γ 0.41 2.42 ment, the probability approach has been employed to predict PWI
Tm 0.45 2.25 and MRWI. There are several probabilistic methods available in
σt,M 0.57 1.74 the literature but the Monte Carlo simulation (MCS) technique has
been utilized in this study. Lately, the application of MCS
Arab J Geosci (2021) 14:362 Page 7 of 16 362

Table 6 The analysis of variance


(ANOVA) for PWI and MRWI Sum of squares Degree of freedom Mean square F statistics Significance value

PWI
Regression 386.84 3 128.95 27.09 0.000
Residual 104.70 22 4.76
Total 491.54 25
MRWI
Regression 3.03 3 1.01 17.68 0.000
Residual 1.26 22 0.06
Total 4.29 25

technique in the field of rock engineering and mining sciences is probabilistic-based computerized method which is able to ad-
seen a surge. Morin and Ficarazzo utilized the MCS technique dress complex problems (Metropolis and Ulam 1949). It takes
along with the Kuz-Ram fragmentation model to ascertain the into account the effect of uncertainty and risk for a variety of
fragmentation of blasted rock in open pit mine (Morin and predictive or forecasting models, in the field of finance,
Ficarazzo 2006). Sari and Sari et al. were able to explain the decision-making and project management. It incorporates
variability of rock mass properties through this innovative sam- the variability of input parameters through random sampling
pling technique (Sari 2009; Sari et al. 2010). Moreover, Ghasemi from statistically derived probability distributive function.
et al. quantified the uncertainty that arose while determining the The probability distribution functions can be in the form of a
factor of safety for coal pillars in underground mine (Ghasemi frequency histogram or a density function that are capable of
et al. 2010). Sari et al. simulated the backbreak obtained from best describing the data distribution of the input parameters
blasting in surface mines by using this stochastic MCS modelling unlike the conventional forecasting or empirical models where
(Sari et al. 2014). An overview of this approach can be found in only a few constant values are obtained. During the MC sim-
Blair (2011). In addition to it, the MCS technique has also been ulation, a random variable is selected for each input parameter
applied in the domain of slope stability analysis (Manteghi et al. according to their range of estimates and is subsequently fed
2012; Mojtahedi et al. 2019). Few researches have been done to into the deterministic model. The result obtained is recorded
estimate the risks related to flyrock distance in quarry blast and the simulation is reported for a pre-determined number of
(Armaghani et al. 2016; Ghasemi et al. 2012). Furthermore, times. Post simulation, a large set of result, is acquired from
Murmu et al. estimated the probability of occurrence of various the model, which have been obtained as a result of random
magnitude of peak particle velocity arising from bench blast near input values. These results delineate the likelihood or proba-
Chittorgarh Heritage Fort (Murmu et al. 2018). bility of the risk that can occur in the physical system under
study.
There are primarily two objectives that can be achieved
Monte Carlo simulation through the MCS. Firstly, the quantitative examination of
existing uncertainties or variables in the physical system there-
The MCS technique was first introduced by Metropolis and by making it susceptible to risk. Secondly, the role of contri-
Ulam and has now gained wide acceptability as a full-fledged bution key parameters responsible for variability in the system
through sensitivity analysis. A flowchart of the whole meth-
Table 7 Summary of the regression for the two proposed models odology along with MCS technique is shown in Fig. 4.
Variables Coefficients Standard error t statistics p value

PWI Limitation of the proposed method


Intercept 100.39 15.22 6.60 0.00
σt,P 3.25 1.55 2.09 0.00 Although the regression-based Monte Carlo simulation tech-
γ −39.92 6.77 −5.90 0.18 nique serves well for prediction of PWI and MRWI; however,
G 0.95 0.70 1.36 0.04 there are certain limitations to the study:
MRWI
Intercept 3.88 2.22 1.75 0.09 & The database utilized for the development of the empirical
γ −1.98 2.34 −0.85 0.40 models is limited to 30 data sets. This can be further in-
Tm 0.08 0.14 0.57 0.57 creased to add to the confidence level of the models.
σt,M 1.22 0.26 4.59 0.00 & Several other independent variables can be added in the
future study.
362 Page 8 of 16 Arab J Geosci (2021) 14:362

Table 8 Brief summary of the geo-mining conditions of the longwall panels used for model validation

Name of mine Balrampur, P1 Jhanjra, AE1 Jhanjra, AW4 Venkatesh


and panel Khani 7, 9

Name of the Passang RVIIA RVIIA Top Seam


seam
Seam thickness 2.4 2.8–5.78 2.8–5.78 11.42
(m)
Working height 2.4 2.2 2.6 3
(m)
Average depth 50 95 92 272
(m)
Em (GPa) 3 12.6 10.4 4.12
σt,P (MPa) 0.62 1.93 1.54 0.80
σt,M (MPa) 3.02 5.18 4.46 5.21
Ti (m) 16 12.96 14.05 20.87
Tm (m) 14 15.72 23.96 15.85
Fl (m) 150 120 120 120
Nature of Coarse-grained to medium-grained Shale, sandstone Shale, fine-grained sandstone, laminated formation, Coal
immediate sandstone, with shale lamination with coal at the top coal
roof
Nature of main Medium grained sandstone Shale, sandstone Laminated formation, coarse-grained to Grey
roof medium-grained sandstone, intercalation, shale, sandston-
coal e, shale
PWI (m) 19 21 14 15
MRWI (m) 80 68 67 104
Powered 4*450 4*550 4*550 4*360
support
capacity (t)

& The limits of the probability distribution are a function of The heart of the MC simulation is generating random var-
the database. Therefore, more data renders a more com- iables for input parameters by selecting an appropriate sam-
prehensive model. pling technique. The Latin hypercube sampling technique is
& Not all scenarios generated during the MCS are practical chosen over the conventional Monte Carlo sampling as it re-
in the field. Therefore, it calls for sound judgement in quires less computation time and yields lesser standard error.
deciphering the impractical scenarios so that it does not In addition to it, it stratifies the cumulative distribution func-
influence the overall engineering measures during tion curve of the input parameter into equal intervals from
weighting. which random values are acquired. This helps in making sim-
ulation converge quicker as compared to when any other sam-
The @RISK Palisade software has been employed for the pling technique is used.
MC simulation. The software works as a MS-Excel™ add-in, The steps that have been undertaken to conduct MC simu-
making use of the MC functions and random distributions in lation are listed below:
the form of spreadsheet models. It accommodates for basic
data fitting using the maximum likelihood estimators. In this & The sample data pertaining to each input parameter have
study, MCS has been applied in two deterministic models, i.e. been complied.
the PWI model and MRWI model, i.e. Eqs. 3 and 4, respec- & The deterministic models which best describe the physical
tively. All the input parameters, namely γ, G, σt,P, Tm and σt,M system of cavability, i.e. the PWI and MRWI models,
have been fitted to the aforestated variables. Table 9 illustrates have been developed as represented in Eqs. (3) and (4).
the details of the probability distribution that have been & The probability density functions (PDF) which best de-
assigned to each input parameter. Figure 5 portrays the fre- scribe the sample data of distribution of the input param-
quency histogram concerning each input parameter. The eters have been defined.
goodness of fit test to ascertain appropriate distribution fit & In the MC simulation, 10,000 iterations have been execut-
for the input parameters has been determined by employing ed for random realization generations for input
the chi-square (χ2) test. parameters.
Arab J Geosci (2021) 14:362 Page 9 of 16 362

Fig. 3 Comparison of the


proposed models with the existing
equations of literature pertaining
to a PWI and b MRWI

& The generated realizations are fed into the two determin- to it, large number of scenarios have been covered through sev-
istic models. eral random combination of input variables to obtain the state
& Output as PWI and MRWI has been obtained in the form variable, which are beyond the control of the design engineer.
of PDF and cumulative distribution function (CDF). Therefore, it can be comprehended that the MCS technique is a
good tool to simulate roof strata weighting.
The output of the state variables obtained from the MCS has The severity level associated with the PWI is classified in
been depicted in the form of PDF in Fig. 6a and 6b. Furthermore, Table 10 based on the observation of the condition of the
the CDFs of the measured, predicted and simulated data have longwall working face of several panels during the periodic
been compared to assess the credibility of the MC simulation weighting. Table 11 enlists the condition of the longwall working
(Fig. 7a and 7b). It can be noticed that the CDFs are in close face during the occurrence of periodic weighting for various
proximity to each other displaying good agreement. In addition panels. Three categories, namely, low, moderate and critical
362 Page 10 of 16 Arab J Geosci (2021) 14:362

severity, have been defined based on the magnitude of the PWI.


Furthermore, the weighting characteristics have been classified
too which provides an insight as to the length of main roof
overhang interval that has to be maintained for smooth operation
of a longwall panel. Based on the MC simulation, it can be
inferred that the probability of periodic weighting to attain low,
moderate and critical severity has been worked out to be 0.50,
0.24 and 0.26, respectively. This implies that 50% of the periodic
weighting are critical in nature if longwall panels have to be
developed in the geological conditions under the study.
Therefore, effective measures of roof strata softening techniques
like hydro fracturing, deep hole blasting of strata from surface
and overhang blasting have to be selected based on the environ-
mental constraints in order to prevent irregular caving or longer
length of overhang behind the shields. These techniques if
adopted would further reduce the load coming on the face. The
data pertaining to the condition of the longwall face during first
significant weighting of the main roof could not be obtained.
Therefore, its severity level could not be classified. However,
intuitively it can be said that longer the MRWI greater the risk
at the longwall working face. The probability of MRWI to be less
than 75 m has been obtained as 0.56. The summary of the MCS
is presented in Table 12.

Sensitivity analysis

The effect or influence of predictor parameters over the predicted


variable can be quantified through the sensitivity analysis. It is a
method to decipher the relative importance of predictor variables.
The @RISK palisade software conducts twofold sensitivity anal-
ysis: regression and correlation sensitivity. The regression-based
sensitivity utilizes the stepwise regression method which allows
for estimating the contribution to variance corresponding to each
input parameter. It breaks down the variance of the state variable of
the probabilistic model into fractions. These fractions can be im-
puted to the input parameters or set of input parameters. Therefore,
the measure of the fractions is directly elucidated as the measure of
sensitivity. The greater the magnitude of the contribution to vari-
Fig. 4 Metho`dology of regression-based MC simulation
ance corresponding to any input parameter, the greater is its influ-
ence on the state variable. The results of the sensitivity analysis can

Table 9 Best fitted probability


density functions (PDF) estimated Input variable Fitted distribution
for the input variables used in MC
simulation PWI
γ Weibull (k (shape parameter): 6.38; λ (scale parameter): 0.5)
G Lognormal (mean: 2.08; Std. dev: 1.35)
σt,P Triangular (min: 0.01; most likely: 0.91; max: 2.15)
MRWI
γ Weibull (k (shape parameter): 6.38; λ (scale parameter): 0.5)
Tm Lognormal (mean: 17.31; Std. dev: 7.61)
σt,M Triangular (min: 1.95; most likely: 4.16; max: 6.16)
Arab J Geosci (2021) 14:362 Page 11 of 16 362

Fig. 5 The PDF of input parameters incorporated into the MC simulation


362 Page 12 of 16 Arab J Geosci (2021) 14:362

Fig. 6 PDF of the output obtained


from MCS of a PWI and b
MRWI

be seen in Table 13. It is evident that for PWI, the weight of the Conclusion
main roof majorly governs the weighting phenomena whereas for
MRWI, the tensile strength plays a pivotal role as the main roof The stability of the longwall working face is of prime concern for
has to overcome it to undergo tensile fracturing. the design engineer as it directly attributes to the economic loss,
In the correlation-based sensitivity analysis, the rank order safety of men and equipment. Therefore, an attempt has been
correlations are determined between the input and the state made to cater to these uncertainties by employing the
variable during the simulation process. The rank order corre- probabilistic-based approach. The important features studied un-
lation coefficient can take a value anywhere between − 1 and der the current research are summarized below. A comprehen-
+ 1. The magnitude of the correlation coefficient explains the sive database pertaining to 30 longwall panels representing vary-
influence of the input over the output whereas the sign con- ing geo-mining conditions has been prepared. Some critical fac-
vention unravels the nature of the relationship between the tors like tensile strength and thickness, unit weight and modulus
same. The results of the correlation-based sensitivity analysis of elasticity of main roof have been selected to estimate the
comply with the results obtained from the regression-based weighting intervals (PWI and MRWI) utilizing regression. The
sensitivity analysis. The PWI is influenced by in descending predictive capability of the regression models is put forward by
order γ, σt,P and G, respectively. For the MRWI, the σt,M plays calculating the R2, 0.79 and 0.71, for PWI and MRWI model,
a major role followed by the γ and Tm. respectively. Furthermore, the statistical significance of the
Arab J Geosci (2021) 14:362 Page 13 of 16 362

Fig. 7 Comparison between the


CDFs of measured, predicted and
simulated values of a PWI and b
MRWI

models has been achieved by comparing it to the other existing MRWI) show lognormal distribution incorporating random com-
analytical and empirical models available in the literature. The bination of varying predictor variables. The comparison between
models developed have been found to show lesser margin of the CDFs of measured, predicted and simulated results has been
error as compared to other equations. The regression models made, which displayed the credibility of the MCS based on its
have been used as deterministic mathematical model for the close proximity to each other. Furthermore, the condition of the
probabilistic analysis. The Monte Carlo simulation (MCS) tech- longwall working face during periodic events has been described
nique has been employed by determining the PDF of the input for all the longwall panels under the study. Consequently, the
parameters and random variable generation using Latin hyper- severity level of periodic weighting has been classified. Through
cube sampling method. The PDF of the state variables (PWI and the MCS it has been determined that the probability of periodic

Table 10 Probability of
occurrence of PWI of different PWI (m) Severity level Weighting characteristics Probability of occurrence
severity level
< 17 Low Readily controllable 0.50
17 to 19 Moderate Operationally controllable 0.24
>19 to 27 Critical Uncontrollable 0.26
362 Page 14 of 16 Arab J Geosci (2021) 14:362

Table 11 Longwall face


condition during weighting Name of the Panel PWI Face condition during weighting
events mine Name (m)

Moonidih A4 8 No significant strata control problem


Jhanjra W4 10 No significant strata control problem
Jhanjra W2 11 No significant strata control problem
Jhanjra W1 12 No significant strata control problem
PVK 5 21 12 No significant strata control problem
GDK 10 A 3B 12 No significant strata control problem
JK 5 3 12.5 High face convergence, severe strata control problem (inadequate support
rating)
JK 5 3A 15 No significant strata control problem
VK 7 9 15 No significant strata control problem
Jhanjra AW9 16 No significant strata control problem
JK 5 5 16 No significant strata control problem
VK 7 4 17 No significant strata control problem
Jhanjra AW8 17 Face deterioration and frequent stoppage during weighting,
JK 5 1 17.5 High face convergence and roof degradation
Jhanjra AW3 18 Face deterioration and frequent stoppage during weighting
Jhanjra AW1 18 Stable roof till 420 m face advance and after that 22 legs got bent, 7
canopies damaged, significant no. of hoses damaged, long downtime
Rajendra P2 18 The panel practiced blasting from surface to reduce overhang length.
Therefore, the weightings were operationally controlled.
Balrampur P1 19 Damage to support during weighting
Jhanjra AE1 19 65 leg circuits bled during weighting, 56 leg circuits yielded
GDK 10A 11 20 High face convergence and roof degradation, frequent face stoppage during
weighting
Jhanjra AW2 20 Face deterioration and frequent stoppage during weighting
Khottadih 2 20 Sudden collapse of strata, damage to supports at several instances
Khottadih 3 20 Face collapse at 800 m of face advance
GDK 10A 3A 21 No significant strata control problem
GDK 10 A 12 21 No significant strata control problem
Adriyala 1 22 Seepage, face spalling observed within 20 shields distance
Khottadih 1 25 Sudden collapse of strata, damage to supports at several instances
Churcha P1 27 Significant intensity of air blasts, ultra excessive rapid loading in powered
supports, bursting of seals and blowing of outer cylinder caps were
experienced during weighting. Ultimately, the face collapsed.

Table 12 Summary of the MC simulation

Variables Min. Max. Mean Std. Dev 5% 95% Table 13 Sensitivity ranking of input variables

Variables Contribution to variance (%) Correlation coefficient


PWI
PWI (m) 7 34 16.62 3.34 11 22 PWI
γ (T/m3) 1.88 2.41 2.22 0.07 2.09 2.32 γ 65.79 (1) − 0.795 (1)
G (GPa) 0.02 7 2.02 1.21 0.5 4.65 G 14.50 (3) 0.332 (3)
σt,P (MPa) 0.01 2.15 0.91 0.45 0.25 1.75 σt,P 19.71 (2) 0.442 (2)
MRWI MRWI
MRWI (m) 25 140 71.28 18.67 40 102 γ 5.78 (2) − 0.18 (2)
γ (T/m3) 1.88 2.41 2.22 0.07 2.09 2.32 Tm 1.66 (3) 0.09 (3)
Tm (m) 3 50 17.31 7.61 7.79 31.51 σt,M 91.84 (1) 0.96 (1)
σt,M (MPa) 1.95 6.16 4.16 0.87 2.64 5.56
The values in bracket show the ranking
Arab J Geosci (2021) 14:362 Page 15 of 16 362

weighting to attain low, moderate and critical severity level are Galvin JM (2016) Ground engineering - principles and practices for un-
derground coal mining. https://doi.org/10.1007/978-3-319-25005-2
0.50, 0.24 and 0.26, respectively. Also, the weighting character-
Gao F, Stead D, Coggan J (2014) Evaluation of coal longwall caving
istics under which the longwall panel can achieve successful characteristics using an innovative UDEC Trigon approach.
completion have been defined. The probability of MRWI to be Comput Geotech 55:448–460. https://doi.org/10.1016/j.compgeo.
less than 75 m has been obtained as 0.56. Hence, according to the 2013.09.020
study, the application of effective roof softening methods has Ghasemi E, Shahriar K, Sharifzadeh M, Hashemolhosseini H (2010)
Quantifying the uncertainty of pillar safety factor by Monte Carlo
been suggested to decrease the load at the longwall working face simulation—a case study. Arch Min Sci 55(3):623–635
arising due to weighting. Ghasemi E, Sari M, Ataei M (2012) Development of an empirical model
In addition to it, regression and correlation-based sensitiv- for predicting the effects of controllable blasting parameters on
ity analysis has been conducted wherein ranks have been giv- flyrock distance in surface mines. Int J Rock Mech Min Sci 52:
en to the input parameters based on their influence on the state 163–170. https://doi.org/10.1016/j.ijrmms.2012.03.011
Hosseini N, Goshtasbi K, Oraee-Mirzamani B, Gholinejad M (2014)
variable. It has been concluded from the sensitivity analysis Calculation of periodic roof weighting interval in longwall mining
that the unit weight is a critical factor for periodic weighting using finite element method. Arab J Geosci 7(5):1951–1956. https://
and tensile strength of the main roof markedly governs the doi.org/10.1007/s12517-013-0859-8
first significant main roof weighting. Jabinpoor A, Jafari A (2013) Estimation of rock cavability in jointed roof
in longwall mining. In: Coal Operators’ Conference, pp 68–73.
https://doi.org/10.13140/RG.2.1.1091.6000
Acknowledgements The authors are grateful to the various managements
of ECL and SCCL mines from where data pertaining to the study could be Ju J, Xu J, Shan Z (2015) Mechanisms of the abnormal first weighting in
obtained. The article is a part of the Ph.D. work of the first author. “knife handle shaped face” with 7.0 m high supports. Int J Oil, Gas
Therefore, the findings and views that have been expressed are the opin- Coal Technol 9(3):348. https://doi.org/10.1504/ijogct.2015.069002
ions of the authors and not necessarily of the organization they serve. Kuznetsov ST, Pekarskii DG, Korovkin VT (1973) Determining the nor-
mal stresses in a uniform bent beam cantilever. Sov Min Sci 9(5):
478–482
Data availability The full data supporting the findings of the study can be
obtained from the corresponding author on request. Li J, Huang Y, Zhang J, Li M, Qiao M, Wang F (2019) The influences of
key strata compound breakage on the overlying strata movement
and strata pressure behavior in fully mechanized caving mining of
Declarations shallow and extremely thick seams: a case study. Adv Civ Eng
2019:11–11. https://doi.org/10.1155/2019/5929635
Conflict of interest The author declares that there are no conflicts of Liu X, Song G, Li X (2015) Classification of roof strata and calculation of
interest. powered support loads in shallow coal seams of China. J South Afr
Inst Min Metall 115(11):1113–1119. https://doi.org/10.17159/
2411-9717/2015/v115n11a15
Luan H, Liu X, Wang E, Liang T, Cao Z (2019) The effect of disconti-
References nuity density on key strata threshold span in longwall panels. J
Geophys Eng 16(2):289–296. https://doi.org/10.1093/jge/gxz006
Armaghani DJ, Mahdiyar A, Hasanipanah M, Faradonbeh RS, Manteghi H, Shahriar K, Torabi R (2012) Numerical modelling for esti-
Khandelwal M, Amnieh HB (2016) Risk assessment and prediction mation of first weighting distance in longwall coal mining - a case
of flyrock distance by combined multiple regression analysis and study. In: Coal Operators’ Conference, pp 60–68
Monte Carlo simulation of quarry blasting. Rock Mech Rock Eng Metropolis N, Ulam S (1949) The Monte Carlo method. J Am Stat Assoc
49(9):3631–3641. https://doi.org/10.1007/s00603-016-1015-z 44:335–341
Banerjee G, Kumbhakar D, Ghosh N (2016) Assessment of cavability Mohammadi S, Ataei M, Kakaie R, Mirzaghorbanali A (2019) A New
and categorization of coal measure roof rocks by parting plane ap- Roof Strata Cavability Index (RSCi) for longwall mining incorpo-
proach. In: Recent advances in rock engineering. Bangalore, pp rating new rating system. Geotech Geol Eng 6:1–18. https://doi.org/
301–308 10.1007/s10706-019-00857-6
Bilinski A, Konopko W (1973) Criteria for choice and use of powered Mojtahedi SFF, Tabatabaee S, Ghoroqi M, Soltani Tehrani M, Gordan B,
supports. In: The symposium on protection against roof falls
Ghoroqi M (2019) A novel probabilistic simulation approach for
(Katowice, 1973), Paper No. IV-1
forecasting the safety factor of slopes: a case study. Eng Comput
Blair DP (2011) A probabilistic analysis of vibration based on measured 35(2):637–646. https://doi.org/10.1007/s00366-018-0623-5
data and charge weight scaling. In: Proceedings of the Sixth EFEE
Morin MA, Ficarazzo F (2006) Monte Carlo simulation as a tool to
World Conference on Explosives and Blasting Techniques, Lisbon,
predict blasting fragmentation based on the Kuz-Ram model.
Portugal, pp 319–337
Comput Geosci 32(3):352–359. https://doi.org/10.1016/j.cageo.
Brady BHG, Brown ET (2004) Rock mechanics for underground mining.
2005.06.022
https://doi.org/10.1007/978-1-4020-2116-9
Mukherjee SN (2003) Mechanised longwall mining in India - a status
Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, Carré G, Marquéz
review. J Inst Eng 84(August):5–10
JRG, Gruber B, Lafourcade B, Leitão PJ, Münkemüller T, McClean
C, Osborne PE, Reineking B, Schröder B, Skidmore AK, Zurell D, Murmu S, Maheshwari P, Verma HK (2018) Empirical and probabilistic
Lautenbach S (2013) Collinearity: a review of methods to deal with analysis of blast-induced ground vibrations. Int J Rock Mech Min
it and a simulation study evaluating their performance. Ecography Sci 103:267–274. https://doi.org/10.1016/j.ijrmms.2018.01.038
36(1):27–46. https://doi.org/10.1111/j.1600-0587.2012.07348.x Obert L, Duvall WI (1967) Rock mechanics and the design of structures
Gale WJ, Nemcik J (1998) Prediction of stata caving characteristics and in rock. Wiley, New York
its impact on longwall operation. In: Coal Operators’ Conference, Peng SS, Chiang HS (1984) Longwall mining. John Wiley & Sons, New
Wollongong, Australia, pp 156–166 York
362 Page 16 of 16 Arab J Geosci (2021) 14:362

Sari M (2009) The stochastic assessment of strength and deformability workings. Int J Rock Mech Min Sci 47(1):1–16. https://doi.org/10.
characteristics for a pyroclastic rock mass. Int J Rock Mech Min Sci 1016/j.ijrmms.2009.09.001
46:613–626 Singh GSP, Singh UK (2010b) Numerical modeling study of the effect of
Sari M, Karpuz C, Ayday C (2010) Estimating rock mass properties using some critical parameters on caving behavior of strata and support
Monte Carlo simulation: Ankara andesites. Comput Geosci 36:959– performance in a longwall working. Rock Mech Rock Eng 43:475–
969 489. https://doi.org/10.1007/s00603-009-0061-1
Sari M, Ghasemi E, Ataei M (2014) Stochastic modeling approach for the Singh GSP, Singh UK, Banerjee G (2004) Cavability assessment model
evaluation of backbreak due to blasting operations in open pit mines. for longwall working in India. In: Proceedings of the Third Asian
Rock Mech Rock Eng 47(2):771–783. https://doi.org/10.1007/ Rock Mechanics Symposium, Kyoto, Japan, pp 295–300
s00603-013-0438-z Wang S, Li N, Li C, Zou Z, Chang X (2015) Instability mechanism
Sarkar SK (1998) Mechanised longwall mining - the Indian experiences. analysis of pressure-arch in coal mining field under different seam
Oxford and IBH Publishing Company Private Limited, New Delhi, dip angles. Dyna (Spain) 90(3):279–284. https://doi.org/10.6036/
India 7530
Sarkar SK, Dhar B (1993) Strata control failures at caved longwall faces Wang S, Wu X, Zhao Y, Hagan P (2018) Mechanical performances of
in India - experience from Rana to Churcha (1964 to 1990). In: pressure arch in thick bedrock during shallow coal mining.
Proceedings of Fourth Asian Mining Congress, Calcutta, India, vol Geofluids. 2018:1–13. https://doi.org/10.1155/2018/2419659
361–380 Xie GX, Chang JC, Yang K (2009) Investigations into stress shell char-
acteristics of surrounding rock in fully mechanized top-coal caving
Shabanimashcool M, Jing L, Li CC (2014) Discontinuous modelling of
face. Int J Rock Mech Min Sci 46(1):172–181
stratum cave-in in a longwall coal mine in the arctic area. Geotech
Xu T, Yang TH, Chen CF, Liu HL, Yu QL (2015) Mining induced strata
Geol Eng 32(5):1239–1252. https://doi.org/10.1007/s10706-014-
movement and roof behavior in underground coal mine. Geomech
9795-y
Geophys Geo-Energy Geo-Resour 1(3-4):79–89. https://doi.org/10.
Singh GSP, Singh UK (2009) A numerical modeling approach for assess- 1007/s40948-015-0010-2
ment of progressive caving of strata and performance of hydraulic Zhang L, Einstein HH (2004) Using RQD to estimate the deformation
powered support in longwall workings. Comput Geotech 36(7): modulus of rock masses. Int J Rock Mech Min Sci 41(2):337–341
1142–1156. https://doi.org/10.1016/j.compgeo.2009.05.001 Zhao M, Huang Q, Huang K (2019) Roof structure and support resistance
Singh GSP, Singh UK (2010a) Prediction of caving behavior of strata and during roof first weighting of shallow coal seams in Yushen mining
optimum rating of hydraulic powered support for longwall area. Electron J Geotech Eng 24(02):337–346

You might also like