Safety Science 123 (2020) 104555

Contents lists available at ScienceDirect

Safety Science
journal homepage: www.elsevier.com/locate/safety

Estimation of human error rate in underground coal mines through T

retrospective analysis of mining accident reports and some error reduction
strategies
Pramod Kumara, Suprakash Guptab, , Yuga Raju Gundab
⁎

a
Department of Mathematics, INSH, Shree Ramswaroop Memorial University (SRMU), Lucknow 225003, India
b
Department of Mining Engineering, Indian Institute of Technology (BHU), Varanasi 221005, India

ARTICLE INFO ABSTRACT

Keywords: Human error is one of the leading causes of workplace accidents and a pressing threat to system safety. The
Accident analysis management and control of human errors need their identification in the system activities and prioritisation of
Fuzzy set errors. Knowledge of error rates or probabilities is essential for their prioritisation and is a prerequisite for
Human error designing control measures of human error to enhance system safety. Traditional approaches estimate the
Underground coal mine
human error rate based on experts’ opinion, which suffers from the inherent uncertainty of human judgments.
System safety
Human factors
This research proposes a methodology for estimation of human error rate from a retrospective analysis of ac-
cident reports using fuzzy mathematical concepts. The proposed approach uses accident reports of underground
coal mines for assessing the human error rates of essential mining activities, identifying the critical activities and
error types. It also suggests some error reduction strategies for devising an intervention to accidents.

1. Introduction responsible for 60–90% of all the accidents; it is a pressing threat to

system safety and a major cause of an accident, property damage, sig-
The effect of risk in the industrial scenario starts form appetite nificant personal injury and sometimes fatality (Di Pasquale et al.,
stage, continued with tolerance stage and exposed to the threshold 2015, 2013; Feyer and Williamson, 1998; Griffith and Mahadevan,
stage to form an accident. In recent times, highly complicated operating 2011; Hobbs and Williamson, 2003; Rushworth et al., 1999; Zhang
procedures, continual market pressure, and harsh working environment et al., 2019). So, in all the high-risk industries, human error has been a
have transformed mining operations risky and more prone to mishaps. grave concern to safety personnel, and it is a widely studied and
So, it is a big challenge to mining officials and employees to work burning topic for personal safety (Academies and of Sciences and
safely, effectively, and deal with all the inherent risk. Medicine, 2018). Unfortunately, analysing mining accidents in India
Indian mining industries suffer from recognisable fatality rates from human error aspects is limited. This knowledge gap motivated the
compared to all other causes of deaths in India (Ministry of Statistics expansion of this research, which undertakes a quantitative analysis of
and Programme Implementation, 2017). As per the latest statistics of mine accidents from a human factors perspective.
mines, the fatality rate per 1000 persons employed in Indian coal mines Management of human error in various stages of a system necessi-
has varied between 0.36 and 0.16 in the last decade (2006 to 2015) tates its detailed analysis (Swain et al., 1963). Human reliability ana-
(Directorate General of Mines Safety, 2015). The fatality rate figures lysis (HRA) is a technique which tries to assess the human contribution
show a marginal decreasing trend (slope of the trend line is −0.014) to risk. Several HRA techniques like Technique for Human Error Rate
which indicates that the traditional measures of safety and productivity Prediction (THERP) (Swain and Guttmann, 1983), Accident Sequence
in Indian coal mines has reached its limit of effectiveness and fresh Evaluation Program (ASEP) (Swain, 1987), Cognitive Reliability and
inputs are needed for further improvement. Error Analysis Method (CREAM) (Hollnagel, 1998), Simplified Plant
Many experts indicate that the safety performance of modern mines Analysis Risk Human Reliability Assessment (SPAR-H) (Gertman et al.,
can only be improved further by assessing and managing unsafe human 2005), Information, Decision and Action in Crew context – IDAC
behaviour (Patterson and Shappell, 2010; Paul and Maiti, 2007; Yin (Chang and Mosleh, 2007), Phoenix (Ekanem et al., 2016), have been
et al., 2017). Many studies agree that unsafe human behaviour is developed to explain and quantify the human behaviour which can

⁎
Corresponding author.
E-mail addresses: sgupta.min@iitbhu.ac.in (S. Gupta), gundayugaraju.rs.min17@itbhu.ac.in (Y.R. Gunda).

https://doi.org/10.1016/j.ssci.2019.104555
Received 19 September 2018; Received in revised form 6 September 2019; Accepted 17 November 2019
0925-7535/ © 2019 Elsevier Ltd. All rights reserved.
P. Kumar, et al. Safety Science 123 (2020) 104555

negatively impact a system, and predict its occurrence and severity (De human error has its roots in the environmental stimuli. The information
Felice and Petrillo, 2019; Dsouza and Lu, 2017). from the environmental stimuli is recognised by the sensory organs in
The HRA approach has evolved from attempts to create human- the sensory processing. (ii) The gathered information from the outside
error databases similar to those created for hardware components, to world is organised and interpreted in perceptual process. (iii) In the
the use of expert judgment techniques, and back to an improved var- cognitive process, the information from thought, experience, and the
iation of database use (Philippart, 2018). Currently, the HRA study senses are used to understand the meaning and acquire knowledge. (iv)
includes math approaches like fuzzy set theory and the Bayesian net- After cognition, the framework contains decision making which in-
work. Also, there are simulation‐based HRA methods which provide volves goal selection, planning, evaluating options, and selection,
dynamical modelling system based on a virtual environment (Wu et al., which are based on the values, preferences, and beliefs. Through all
2019). these four processes the cognitive processing resources like attention
Many of these HRA techniques are based on the database of the and memory are utilised. (v) Implementation and execution of the
nuclear industry and simulation with operators and has a limited scope decided response as an action.
of using in the mining context. Likewise, HRA techniques which use An error may occur at any of these processes and manifests into an
industrial databanks and expert’s judgement are highly predisposed to unintended action which could reduce or have the potential of reducing
the premise of the experts’ past experiences during the evaluation of the effectiveness, safety or system performance (Sanders et al., 1993). In
safety status of an industrial environment. Ambiguous data generated literature, many researchers have classified human error in a different
from expert’s judgement results in unreliable human error probability context, but the pioneer works are done by Rasmussen and Reason
value which may generate misleading results from subsequent appli- (Reason, 1990; Simpson et al., 2012). They stated the classification of
cations in probabilistic risk and safety analysis. Thus, the fundamental human error based on the causes of errors in psychological context
limitations of these techniques are: during the execution of response, i.e., skill-based (slip and lapses),
mistake (rule base mistake (RBM) and knowledge base mistake (KBM)),
• Insufficient human error data which applies to all sectors (generic and violation.
data)
• Methodological limitations related to the subjectivity of expert 3. Error rate and classification of accident data
judgment
• Uncertainty concerning the actual behaviour of people during ac- Human error influences the safety of a system in its design, devel-
cident conditions opment, installation, operation, and maintenance. It is a contributing
factor in industrial/occupational accidents, and further safety im-
Besides the complexity of human cognition process, the explicit provements can be achieved by managing it. Therefore, it is necessary
quantification of undesired responses (errors) is too hard for conven- to have an in-depth analysis of accidents to identify the critical errors
tional techniques. Indeed, problems exaggerate when the information is and critical activities, for designing countermeasures and upgrade
qualitative, inexact and uncertain (Konstandinidou et al., 2006). Ob- safety. The criticality of an error can only be assessed with a reliable
serving the problem of subjectivity and complexity of human beha- estimate of the error rate. Well known HRA methods like success like-
viours per se, many researchers (Akyuz et al., 2018; Baziuk et al., 2016; lihood index method using multi-attribute utility decomposition (SLIM-
Gupta et al., 2019, 2013; Gupta and Kumar, 2014; Konstandinidou MAUD) (Embrey et al., 1984), and A Technique for Human Event
et al., 2006; Li et al., 2010; Marseguerra et al., 2007; Ung et al., 2006; Analysis (ATHEANA) (Forester et al., 2004) even though simple to use,
Zio et al., 2009) find the application of fuzzy mathematical tools in HRA but solely depend on experts’ judgment to estimate error probability/
to grasp the uncertainty and fuzziness in data. error rate and suffer from the embedded subjective uncertainty. Kotek
Available information about mining accidents in India is mostly and Mukhametzianova have raised concern over the significant varia-
qualitative, inexact and uncertain. This paper presents a framework for tion of HEP values resulted from different methods and proposed a
quantifying the human error rate (HER) through a retrospective ana- more solid statistical approach for the calculation of employee-specific
lysis of accidents' reports using the concepts of fuzzy set theory to grasp HEP (Kotek and Mukhametzianova, 2012). Many researchers proposed
the embed uncertainty in the information. to use fuzzy mathematical tools to grasp the uncertainty in human
The underlying assumption of the proposed approach is that each judgement (Baziuk et al., 2016).
system activity may have a set of error forms, where only a few of them As the primary aim of this research is to reduce risk and enhance
are critical. The exposition of the paper is as follows. Section 2 describes mine safety by the retrospective analysis of mining accident reports,
the mechanism and categorisation of human error and as were used in this work emphasises only on two aspects of system safety, i.e., activity
this study. Section 3 explains the use of fuzzy mathematical tools to and human error. System operation comprises a set of human interac-
estimate the HER. Sections 4 and 5 explain the design and application tions called activities, and the accident statistics vary with the type of
of the proposed methodology in the mining system. In Section 6 we activity. Also, the accident reports hardly identify the type of human
discussed the findings of the research. Section 7 outlines the guidelines error in the human-system interactions that have led to the accident.
for the management and control of human error based on the identified Devising and implementing countermeasures require knowledge about
factors of workplace accidents. Finally, in Section 8 we made conclu- the category of error in system activities. Therefore, the accident re-
sions and limitations of the study. ports are classified based on activity type and error category. This work
presents a structured methodology for categorising errors based on
2. Understanding of human error and its categorisation accident data using fuzzy sets.

The human action (or behaviour) that fails to produce the desired 3.1. Translation of accident report to the fuzzy set of human error
outcome of their task can be defined as human error (Reason, 1990).
Also, human failure (or error) and success in performing a task are the This work assumes that human errors are one of the major causes of
results of the same cognitive processes. So, based on cognitive literature a workplace accident. Committing an error (Slips, Lapses, KBMs, RBMs,
(Al-Tarawneh, 2012; Anderson et al., 2004; Chang and Mosleh, 2007; and Violations) or combination of errors may lead to an accident. Here,
Pan et al., 2017; Sgobba et al., 2017; Simpson et al., 2012; Whaley accident cases are classified based on the five categories of error and are
et al., 2012; Wickens et al., 2015) the cognitive framework (Fig. 1.) that expressed as a fuzzy set. The elements of this fuzzy set are ordered pairs
contains a complete set of cognitive mechanisms underlying human of the number of accident cases (accident frequency) and its member-
failure is as follows. (i) As human behaviour is context-dependent, ship value. The no. of accident cases of an error category (‘E-category

2
P. Kumar, et al. Safety Science 123 (2020) 104555

Fig. 1. Cognitive framework underlying human error.

error' where E is Slips/Lapses/KBMs/RBMs/Violations) for a year is µ a~ij . The value of the µ a~ij varies between [0,1] depending on the number
calculated from the accident reports of that year. An accident case re- of present cues in the incidence sequence of the accident aij matches
port reflecting a favourable set of cues or anchor points of ‘E-category with the set of cues or anchor points of ‘E-category error'. Therefore,
error', is counted against the ‘E-category error'. An accident case report when ‘n’ accidents are reported in the ith year, and analysis of accident
might reveal the presence of more than one category of error and data reveals the likelihood of ‘E-category error' in the jth accident is µ a~ij ,
likewise counted against each error category. then a fuzzy set of ‘E-category error' for the ith year is formulated as,
~
3.2. Calculation of membership value
Ai, E category {
= (ai1 category , µ a~i1 category ) , (a i 2 category , µ a~i2 category ), ,

(ain category , µ a~in category )}

The membership value of ‘E-category error' in an accident is mea-
sured as a ratio of the number of present cues in the accident sequence This work classifies accident cases into five categories. Then the
matches to the total number of cues in the cue–set of the ‘E-category accident data for the ith year is expressed by five fuzzy sets, namely,
error'. A likely set of cues or anchor points for all the five categories of
~
errors has been prepared through extensive literature review and en- {
Ai, slip = (ai1 slip , µ a~i1 slip ) , (a i 2 slip , µ a~i2 slip ), (
, ain slip , µ a~in slip )}
listed in Table 1. (Anderson et al., 2004; Chang and Mosleh, 2007;
Kirwan, 2017; PERROW, 1999; Reason, 1990; Sgobba et al., 2017; ~
Simpson et al., 2012; Whaley et al., 2012; Wickens et al., 2015). It
Ai, lapse = {(ai1 lapse , µ a~i1 lapse
), (ai2 lapse , µ a~i2 lapse
), (
, ain lapse , µ a~in lapse )}
contains nine cues or anchor points for slips and lapses, while there are
ten cues or anchor points for RBM, KBM, and violations. ~
Ai, KBM = (ai1{ KBM , µ a~i1 KBM ) , (a i 2 KBM , µ a~i2 KBM ), ,

(ain KBM , µ a~in )}

3.3. Development of fuzzy sets of error
KBM

~
An accident, aij (jth accident of the ith year) might have occurred due Ai, RBM = (ai1{ RBM , µ a~i1 RBM ) , (a i 2 RBM , µ a~i2 RBM ), ,
to the ‘E-category error' with a possible value of µ a~ij , i.e., ‘E-category
error' could be a causal factor for the accident aij with a possibility of
(ain RBM , µ a~in RBM )}

3
 P. Kumar, et al.

Table 1
List of probable cue-sets or anchor points for different error types.
S. No. Slip Error Lapses Error RBM KBM Violation

1. Action deviates from intention due to the The omission of actions steps at execution Performance is goal-oriented and structured Choose the wrong action plan among Failure to follow a posted sign, e.g.,
presence of attention capture stage that one has intended to perform. through a stored rule. various procedures due to inexperience or caution, danger, No smoking, etc.
time pressure.
2. The error just happened unintentionally. Error occur in more familiar situation The action did not go as plan due to the failure Unable to decide on the specific noble Taking shortcuts to fulfil the desired goal.
primarily due to overconfidence. of the attempted rule. conditions.
3. Errors occur at the time of execution of the Little conscious effort has been put for Failures to achieve the intended goal. Plan inadequate or inappropriate Entry to an unauthorised area.
act. storage of an action plan.
4. The performance is the routine type and Memory failure occurs to follow a routine Occur at intention formation stage due to not The problem is not well understood, and Choosing the wrong way of an act
well-practised in familiar surroundings. action sequence. following the right rule or following the bad no formal procedure exists. deliberately for personal benefits
rule.

4
5. Unconscious control like smooth, The forgetting intention, place losing, The plan is inadequate or inappropriate as the There is a mismatch between intended Not implementing safety barriers for
automated, and highly integrated checklist, material left at the workplace due situation is misdiagnosed before applying the consequences and prior intention. predicted risk.
behaviour. to attentional capture. rule.
6. No pre-planned has done or required for the Goal fixing and interpretation have not done Repetition of trip and fall behaviour. Unfamiliar environment. Motivating by another to break the rules.
attempted task. properly.
7 Performance is governed by stored patterns Delayed interpretation or incorrect Future caution/ warnings have been ignored The employee has been vested with the Wilfully disregard the rules and
of pre-programmed instructions in prediction in case multiple goals present at and not considered to frame rules. responsibility of tasks requiring prior procedures
analogue structures. the same time. knowledge and training.
8 Failure to make an appropriate attentional Confusion over ideas or intrusion of old Failure to report the observed signed/ The decision-maker has no rule to cope up Reported duty in the intoxicated state.
check. habits in routine type activity symptoms to the competent authority. the problem.
9. Inadvertent or improper act. Repetition due to inattention or over- Overgeneralization of rarely good fit rules. Mishandling during maintenance has Not paying due attention to solving the
attention damaged sophisticated and complicated issues that develop a rebellious attitude
parts. in employees.
10 Improper attempt or shortcut has been chosen Unknowingly taking the risk to do Hostile environment provokes to adopt
to save time something and failure to assess the gravity shortcut.
of a situation.
Safety Science 123 (2020) 104555
P. Kumar, et al. Safety Science 123 (2020) 104555

~
Ai, violation = (ai1{ violation , µ a~i1 violation ) , (a i 2 violation , µ a~i2 violation ), , Step 3: Study the system activities and design a generalised classi-
fication method of human error. The classification scheme must not
(ain violation , µ a~in violation )} be system-specific and should be based on the common human ac-
tion or response process specific.
3.4. Calculation of error rate Step 4: Enlist the possible cues or anchor pointers for each error
type.
Now, only the accident cases of ith year those have µ a~ij category > 0 are Step 5: Develop the accident database from the accident reports for a
considerable length of time (Say last ten years).
counted against the number of ‘E-category error' for the ith year.
Step 6: Categorise accidents by associated system activity and sub-
Therefore, error rate of 'E-category error' for the ith year is expressed as
sequently based on the category of error that might have led to an
a fuzzy number
accident.
ERi, E category {
= n (Ai category ), µ a~i category } = {ERi category , µ ERi category
} Step 7: Calculate the possibility for an error category of an accident
as a ratio of the number of present cues in the accident sequence
when, n (Ai category ) is the number of reported accident cases of the matches to the total number of cues in the cue–set of that error
ith year having µ a~ij category > 0 and, for a conservative estimate of safety, category as listed in Table 1. This exercise generates a data set of
µ a~i category is taken as the max (µ a~ij category ) . possible value of the error categories for an accident. Sort the gen-
So, there are five fuzzy numbers of the error rate for the ith year and erated data into activity-wise and error category-wise. Represents
are expressed as: this data set as a fuzzy set where each element represents an acci-
~ dent case and possible value of an error category. Therefore, for
ERi, slip = {n (Ai slip ), max(µ a~i slip )} = {ERi slip, µ ERi slip }
each system activity, five fuzzy sets are generated for the five ca-
~ tegories of error. A total number of developed fuzzy sets are five
ERi, lapse = {n (Ai lapse ), max(µ a~i )} = {ERi lapse , µ ERi }
lapse lapse
times the number of system activity.
~ Step 8: Now construct the fuzzy sets of the error rate of a system
ERi, KBM = {n (Ai KBM ), max(µ a~i )} = {ERi KBM , µ ERi }
KBM KBM
activity whose elements are ordered pairs of the frequency of an
~
ERi, RBM = {n (Ai error category in a year and the maximum possible value of that
RBM ), max(µ a~i )} = {ERi RBM , µ ERi }
RBM RBM
error category in that year.
~
ERi, violation = {n (Ai Step 9: Calculate the relative cardinality of fuzzy sets of error rate.
violation ), max(µ a~i )}
violation
The relative cardinality value represents the average human error
= {ERi violation , µ ERi violation
} rate for an error category in system activity.
Step 10: Analyse the calculated error rate and find out the safety-
Analysis of accident data collected over k years will result in a finite
critical error types and system activities.
fuzzy set of error rate (ER ) having k elements. For slip type error it is
expressed as:
5. Application of the proposed framework
ERslip
= {(ER1 Application of the proposed method to an industrial system has been
slip , µ ER1 ), (ER2 slip , µ ER2 ), ,(ERk slip , µ ERk )}
illustrated with the underground mining system. Underground coal
slip slip slip

The average ER per year is calculated as the relative cardinality of mining activities are hazardous and have highlighted safety issue.
the fuzzy set ERE category ,i.e., Indian mines are obligated to submit the accident reports to the
Director General of Mines Safety (DGMS). This study uses accident re-
|ERE category |
Avg . no. oferrorperyear = ERE category = ports of all types of accident associated with various mining activities. A
|ER| group of five typical underground coal mines, scattered in different
{(µ ERi category
) × (ERi category )} parts of Indian, has been selected for the safety studies. The selected
=
Noofyears (1) group of mines includes two mines reporting comparatively high, an-
other two mines recording comparatively low and the rest one testi-
where |ERE category| is the cardinality of the finite fuzzy set of error rate fying average rate of reported accidents in the last ten years. Safety
of E-category error, |ER| is the cardinality of the finite set of error rate of status of these mines represents the safety scenario of atypical under-
E-category error, i.e. the number of elements in the finite set and ground coal mine in India. Accidentcasereports have been collected
ERE category is the relative cardinality of the finite fuzzy set of error from the safety division of these mines. Following the methodology
rate of E-category error. Therefore, the average error rate of slip described in Section 3, collected data is classified activity-wise for
(number of slip category error per year) is calculated as the relative subsequent analysis. Essential activities of an underground coal mine
cardinality of the finite fuzzy set of error rate of slip and expressed as are grouped into the following six-unit operations.
(µ ER1 ) × (ER1 slip ) + (µ ER2 ) × (ER2 slip)
Drilling/Blasting: It includes all the drilling activities in mines. Blast
holes are charged with explosives and blasted for the material ex-
slip slip

+ + (µ ERk slip
) × (ERk slip ) cavation. Holes are also drilled related to stability issues, e.g. roof
ERslip =
K bolting, roof stitching, induced caving.
Loading/Unloading: It involves all the loading/unloading activities
that are accomplished in mines.
4. Design of the proposed methodology
Transportation: It includes to and fro transportation of man and
material.
The proposed methodology is based on the retrospective analysis of
Supporting: Supporting activity includes dressing of loose rocks and
accident data collected from the industry. Fig. 2 presents a flowchart of
various activities for reinforcement of underground structures.
the methodology, and the steps are outlined below.
Maintenance: Maintenance activity includes all the routine, sched-
uled and breakdown maintenance of facilities, utilities, equipment, and
Step 1: Select the system for performing safety analysis.
infrastructures.
Step 2: Study the operation of the selected system and classify the
Miscellaneous: All other activities that are carried out in mines and
system operation into a manageable number of system activities.
not included above.

5
P. Kumar, et al. Safety Science 123 (2020) 104555

Fig. 2. Flowchart of the proposed methodology for HER estimation.

Table 2
Analysis of a mine accident following the proposed methodology.
Name/working status Date/shift/ Mining Description and cause of the accident Duration of Existing anchoring pointsfor different error categories
location of the activity enforced
accident absence Error Serial number of Membership
category cues (refer value
Table 1) present

Doman Bhuia /Loader 27.02.2008/ Loading He was loading blasted coal without wearing 46 days Slip 2, 3, 4, 9 4/9
FB. No.-439, 4.00 A.M mining boot. The belt conveyor was stopped at Lapses 3 1/9
Joining date: 3rd shift/ that time. Suddenly belt starts and a coal KBM 3, 5 2/10
15.02.2002 U/G-GM, boulder falls from the belt on his right foot and RBM 1, 3, 5 3/10
MAP-3 Seam gets an injury. Violation 5 1/10
26LW 11 dip.

Classified accident records of each activity have been organised with an accident report description of mine-I for the year 2002.
further into error category. As perceived from the accident reports and A membership value gives a real number on the scale 0 to 1 and also
matching with the listed cues of Table 1, reported incidents are tagged represents the degree of belief for the occurrence of an error and
with different error category. The possibility of an error category in a leading to an incident. For example, the report of the incidence given in
reported accident has been represented as a fuzzy membership value. It Table 2, indicates four out of the nine listed cues of slip category error
is calculated as the ratio of the existing number of cue matches with are present. Therefore, the membership value for slip category error of
cues listed in Table 1 against an error category (as perceived from ac- this incidence is (4/9) ≈ 0.44. When the same category of error is the
cidents reports) and the total number of cues listed inTable1 against cause of repeated accidents within a year in an activity, the maximum
that error category. The calculated ratio indicates the role of an error value of the membership is used for further analysis. The maximum
category to arise an incident. Table 2 illustrates the above methodology value is preferred as a conservative estimate for safety. Similarly, the

6
P. Kumar, et al. Safety Science 123 (2020) 104555

Table 3 countermeasures while critical system activity specifies the application

Calculation of average rate of slips in loading activity of mine-I for the year area of countermeasures. Significant findings of this analysis identify
2001–10. the areas requiring attention.
Year No of Possibility of µ ERi (µ ERi ) × (ERi Avg. Table 4 presents the results of the analysis of collected accident
slip slip )
incidence error
slip
no. of reports from mine-I. It shows that slips are the most prevalent category
(ERi slip )
( µa~ij slip ) slips of error (ERAverage = 0.745) followed by KBM. Therefore, counter-
per year
measures for mine-I should be designed primarily for management and
2001 02 0.56, 0.44 0.56 1.12 0.78 control of slip and KBM. Cases of violations are not an issue, and RBM
2002 01 0.44 0.44 0.44 and lapses are also not a grave concern to the safety management team.
2003 02 0.36, 0.52 0.52 1.04 Human error in maintenance and drilling-blasting activities are the
2004 02 0.28, 0.45 0.45 0.90
cause for concern. Slips in drilling and blasting have the highest overall
2005 02 0.33, 0.44 0.44 0.88
2006 01 0.12 0.12 0.12
ERAverage value of 1.22. LowERAverage in supporting and miscellaneous
2007 02 0.38, 0.62 0.62 1.24 activities is a reflection of safe operating procedures. Status of human
2008 01 0.44 0.44 0.44 error in loading-unloading and transportation activities are comparable
2009 02 0.44, 0.59 0.59 1.18 and not of serious nature. Cases of lapses in supporting activity are very
2010 02 0.22, 0.22 0.22 0.44
rare. In summary, the safety management team of mine-I should focus
on devising countermeasures against slips and KBM in Drilling-blasting
accident reports of all the five case study mines have been analysed. and maintenance activities.
Table 3 presents an analysis of reported accidents of mine-I in Analysis of accident data collected from the case study mines is used
loading activity possibly due to slips. It reads the number of reported to assess the status of safety from the human error aspect and to design
accidents with their membership value, the maximum membership standard for error rate in various underground coal mining activities in
value and the average number of slip category error in loading activity the Indian scenario. Table 5 represents the average safety scenario of
per year. For example, in the year 2007, two reported incidences in Indian underground coal mines. Each cell gives an average number of
loading activity have occurred due to slip category error with possibi- errors per year of a category in an activity. It is apparent from Table 5
lity values 0.38 and 0.62. Therefore, the membership value of slip ca- that slips, and KBMs are predominant in underground coal mines of
tegory error in loading activity for the year 2007 was chosen 0.62, as it India. Lapses and RBMs are the second-highest causal factors of acci-
was the maximum membership value among two incidences. The dents in mines. The rate of violations is the least and not a significant
average rate of slip category error in loading activity calculated from concern about low safety level in underground coal mines. Loading-
the Eq. (1) is 0.78 per year. unloading is the most safety-critical activity. Human error in trans-
Following the above methodology, the average error rate (ERAverage ) portation and miscellaneous activities contributes a large number of
of all error category (slip, lapses, KBM, RBM and Violation) in all the accidents. This finding is also supported by the DGMS report 2015,
system activities are calculated and presented in Table 4 for Mine-I. which states that a significant number of fatal accidents in Indian coal
Average error rate (ERAverage ) in all the five case study mines are cal- mines were caused during the operation of transportation machinery
culated activity-wise and error category-wise. Overall average error (Directorate General of Mines Safety, 2015). ER in drilling-blasting and
rate (ERStandard )is calculated from the average error rates of the case maintenance activities are not the headache of the safety manager. The
study mines as given in Table 5 and represents the safety scenario in lowest ER in supporting activity mirror adherence to safe practices.
terms of HER status in average Indian underground coal mines. This can Therefore, safety management policy of Indian underground coal mines
be used as a standard to compare the safety status concerning human should be oriented to designing countermeasures primarily against slips
error in various system activities. This information is helpful to develop and KBM in loading-unloading and transportation activities. Overall
a suitable intervention for the control and management of human error average ER calculated from the average ER of five case study mines may
and safety in mines. serve as a standard for ER (ERStandard ) . Mean ER values may serve as a
standard for the Indian mining industry. These values may be used to
compare the safety status of a mining system to fix the future course of
6. Results and discussion action. From the findings of this study and with the help of the available
literature on ‘managing and control of human error’ a set of guidelines
Management and augmentation of system safety comprise primarily has been framed for reducing the human error and presented in section
the identification of the potential type of error and activity. 7.
Identification of critical error helps in designing effective

Table 4
Average rate of error in system activities of mine I.
Error category Drilling & Blasting Loading & Unloading Transportation Supporting Maintenance Miscellaneous Descriptive statistics

ERAverage ERAverage ERAverage ERAverage ERAverage ERAverage

Slip 1.22 0.78 0.81 0.25 1.16 0.25 Range: 0.25 – 1.22
Mean: 0.745
Lapse 0.61 0.38 0.33 0.10 0.85 0.18 Range: 0.10 – 0.85
Mean: 0.408
KBM 1.14 0.60 0.59 0.29 1.09 0.29 Range: 0.29 – 1.14
Mean: 0.667
RBM 0.75 0.53 0.36 0.17 0.73 0.23 Range: 0.17 – 0.75
Mean: 0.462
Violation 0.59 0.25 0.43 0.17 0.73 0.17 Range: 0.17 – 0.73
Mean: 0.390
Descriptive statistics Range: 0.59 – 1.22 Range: 0.25 – 0.78 Range: 0.33 – 0.81 Range: 0.10 – 0.29 Range: 0.73 – 1.16 Range: 0.17 – 0.29 Range: 0.10 – 1.22
Mean: 0.860 Mean: 0.508 Mean: 0.504 Mean: 0.196 Mean: 0.912 Mean: 0.224 Mean: 0.534

7
P. Kumar, et al. Safety Science 123 (2020) 104555

Table 5
Overall average error rate (ERStandard ) in underground coal mining activities at India.
Error category Drilling & Blasting Loading & Unloading Transportation Supporting Maintenance Miscellaneous Descriptive statistics

ERStandard ERStandard ERStandard ERStandard ERStandard ERStandard

Slip 0.71 1.14 1.05 0.45 0.66 0.71 Range: 0.45 – 1.14
Mean: 0.787
Lapse 0.51 0.81 0.75 0.24 0.49 0.75 Range: 0.24 – 0.81
Mean: 0.592
KBM 0.74 1.02 0.93 0.42 0.62 0.68 Range: 0.42 – 1.02
Mean: 0.735
RBM 0.47 0.80 0.73 0.31 0.53 0.53 Range: 0.31 – 0.80
Mean: 0.562
Violation 0.28 1.14 0.48 0.22 0.39 0.36 Range: 0.22 – 1.14
Mean: 0.478
Descriptive statistics Range: 0.28 –0.74 Range: 0.80 – 1.14 Range: 0.48 – 1.05 Range: 0.22 – 0.45 Range: 0.39 – 0.66 Range: 0.36 – 0.75 Range: 0.22 – 1.14
Mean: 0.542 Mean: 0.982 Mean: 0.788 Mean: 0.328 Mean: 0.538 Mean: 0.606 Mean: 0.631

6.1. Verification and validation issues some errors are not detected because people are willing to accept only a
rough agreement between the actual state of the world and their cur-
Verification and validation are useful for the accuracy and useful- rent theory about it. With the fundamental premise that humans are
ness of quantification techniques. Verification implies that the system fallible, and errors are expected even in the best organisation, the fol-
works as it is supposed to work, i.e. over any number of repeated trials lowing section enumerates common causes and detection approaches
one gets appropriate output for a given set of inputs. It is possible to for error. Based on the findings of this study, the working context of
assign the proper error cues and explicit calculation of error rate with accidents, and the available literature on ‘managing and control of
the help of the developed framework for retrospective analysis of the human error’ a set of guidelines for error management has been framed
accident sequence. There might be minor differences in the number of (Kumar et al., 2016; Patterson and Shappell, 2010; Rasmussen, 1997;
error cues assigned for an accident sequence, which primarily depends Reason, 2000; Simpson et al., 2012). The proposed guidelines will likely
on the quality and quantity of information and the experience of the reduce the chances of an error occurrence and maximise the likelihood
analyst. The whole calculation part of the framework is easy and de- of error recovery.
terministic which helps the analysts to get the appropriate output for Slips: These are easier to correct when detected in time. Therefore,
any number of trials with the proper input. one practical approach to identifying slips are the actions that people
Validation establishes the credibility of HRA techniques so that they quickly recognise as unintended as soon as they are aware of them.
may be used with confidence. It needs some relevant data to check for Followingschemesarethe effective deterrents for slips.
the adequacy of the results. This data might be from expert opinions or
simulator experiments in the same sector of applicability. Even though • Slip errors occur in the absence of necessary attentional checks, so it
it is preferable to test the techniques against real field data, its gen- is needed to ensure at intervals that things are running as intended.
eration and collection are too hard. As the context under which the • Post attentional check is active in the case when an error cue signal
available generic data was generated is questionable, there are fewer generates to detect it. In some cases, generated signal cannot be
chances of validating this method explicitly. The most practical method captured immediately. In this case, one should apply the more ef-
for validation is to test the framework against real field data, which is a ficient way of mitigating it.
part of the future study. However, the following findings of this study • Slips can be detected by monitoring the feedback from each re-
are in close agreement with the available literature: (i) the slips and sponse and checking it against some correct response.
KBM are the predominant categories of errors in the system activities • Many times, our body automatically generates indicators of slips
(Burns and Bonaceto, 2018; Feng et al., 2019; Patterson and Shappell, and their corrective reflexes for awareness of what is being done and
2010), (ii) a large number of accidents are associated with transporta- what is currently intended. So, one should think with a conscious
tion activities (Directorate General of Mines Safety, 2015). All these mind about it.
testify the effectiveness of the proposed methodology. • It is also needed to design for guiding emergency action at various
Many studies like (Burns and Bonaceto, 2018; Feng et al., 2019; steps, e.g. when errors lead to a blocking of further progress. These
Patterson and Shappell, 2010; Rashid et al., 2013; Wiegmann and are called Error Forcing Conditions “Something that prevents the
Shappell, 2017) have used Reasons four levels framework of human behaviour from continuing until the problem has been corrected”
failure for accident analysis which requires specialist knowledge on (Lewis and Norman, 1986). Like good advice, a forcing function is
human factor. In contrary an analyst with little knowledge of system most valuable if it is not too late.
activities and human factor can perform the same analysis following the • It is needed to design barriers according to the scenario for miti-
proposed framework. gating the consequences of any slip occurrence.

Lapses: A familiar situation and overconfident are the primary

7. Guidelines for the management and control of human error in factors for the occurrence of lapses. Following typical control measures
an operating system are recommended for lapses.

The retrospective analysis of accident report investigates a past in-

cident and aims to identify the leading indicators of the failures in-
• Recovery of lapses is possible by comparing input and desired
output. It is suggested for implementing fast feedback rule for any
cluding human failure (Boring et al., 2017). However, error detection application.
and correction are not easy. Even an individual’s mental model to solve
a problem in wrong respect (Lewis and Norman, 1986) has played a
• Creation of some key paths so that one can capture a wrong
movement immediately after its execution.
significant role in the occurrence of errors. As error occurrence is
mainly based on behavioural, contextual and conceptual conditions,
• The high variable task has more chances of lapses as compared to

8
P. Kumar, et al. Safety Science 123 (2020) 104555

automatic pattern type. reduce the occurrence or to enable recovery of an error.

• Fatigued and pressured personnel expected to commit more of This methodology can be easily adapted to other systems with ad-
lapses (Moore-Ede, 1993). justment in error categories and a list of cues for different error types.
• Effective Supervision played a major role in detecting lapses. Also, accident reporting is required to be standardised in line with the
cue-list of errors. To fix the standard error rates in various system ac-
KBM: Knowledgebase mistake is much hard to detect than skill base tivities, a large volume of data which will fine-tune the standard values
error. Following guidelines are helpful to manage these errors. is required.

• It is recommended to develop the capability of decision making and Acknowledgement

diagnosis of a situation within workers to detect knowledge base
mistake in the workplace. We would like to acknowledge and thank those who provide data,
• A clear goal with an effective strategy is essential to manage KBM, expertise and suggestions in this work. We are thankful to the diligent
and this needs a proper homework. reviewers for their valuable suggestions that have immensely helped us
• Training and supervision of personnel play a significant role in the to improve the paper.
management and control of these errors.
• One should employ a more efficient approach for removing errors References
from the output stage when no clue was found to remove error at an
initial stage. Akyuz, E., Celik, E., Celik, M., 2018. A practical application of human reliability assess-
ment for operating procedures of the emergency fire pump at ship. Ships Offshore
Struct. 13, 208–216. https://doi.org/10.1080/17445302.2017.1354658.
RBM: Lack of well documented and poorly designed procedures Al-Tarawneh, H.A., 2012. The main factors beyond decision making. J. Manag. Res. 4,
may act as a stimulus for this error. Therefore, well documented and 1–23.
readily available standard operating procedure is an effective control Anderson, J.R., Bothell, D., Byrne, M.D., Douglass, S., Lebiere, C., Qin, Y., 2004. An in-
tegrated theory of the mind. Psychol. Rev. 111, 1036.
measure. Other typical control measures are: Baziuk, P.A., Rivera, S.S., Leod, Núñez Mc, J, 2016. Fuzzy human reliability analysis:
applications and contributions review. Adv Fuzzy Syst 2016.
• Although self-monitoring of activity is needed for all categories of Boring, R.L., Rasmussen, M., Ulrich, T.A., Blackman, H.S., 2017. Retrospective
Application of Human Reliability Analysis for Oil and Gas Incidents: A Case Study
errors, self-monitoring of performance is the most useful technique
Using the Petro-HRA Method. SAGE Publications Sage, CA, Los Angeles, CA, pp.
for RBM. 1653–1657.
• Some error forcing conditions can also be implemented to avoid Burns, K., Bonaceto, C., 2018. An empirically benchmarked human reliability analysis of
general aviation. Reliab. Eng. Syst. Saf. https://doi.org/10.1016/j.ress.2018.07.028.
opting the wrong process.
•
Chang, Y.H.J., Mosleh, A., 2007. Cognitive modeling and dynamic probabilistic simula-
The implementing process should be appropriately designed ac- tion of operating crew response to complex system accidents: part 1: overview of the
cording to the scenario of an industry. IDAC Model. Reliab. Eng. Syst. Saf. 92, 997–1013. https://doi.org/10.1016/j.ress.
2006.05.014.
De Felice, F., Petrillo, A., 2019. An Overview on Human Error Analysis and Reliability
Violation: There are a long list of aggravating factors of violations, Assessment. pp. 19–41. https://doi.org/10.1007/978-3-319-62319-1_2.
e.g. fatigue both mentally and physically, night shifts work, production Di Pasquale, V., Iannone, R., Miranda, S., Riemma, S., 2013. An overview of human re-
pressure, etc. Suggested control measures are: liability analysis techniques in manufacturing operations. Oper. Manag. 221–240.
Di Pasquale, V., Miranda, S., Iannone, R., Riemma, S., 2015. A Simulator for human error

• Give decision power to the personnel as per the need of task pro-
probability analysis (SHERPA). Reliab. Eng. Syst. Saf. 139, 17–32. https://doi.org/
10.1016/j.ress.2015.02.003.
cedures. Directorate General of Mines Safety, 2015. STATISTICS OF MINES IN INDIA, VOLUME – I

• Maintain a congenial working environment and avoid harsh condi- (COAL). Dhanbad.
Dsouza, N., Lu, L., 2017. A Literature Review on Human Reliability Analysis Techniques
tions, e.g. darkness, glare, and excessive noise, etc. Applied for Probabilistic Risk Assessment in the Nuclear Industry. In: Advances in
• Control fatigue due to production pressure and night shift work by Human Factors in Energy: Oil, Gas, Nuclear and Electric Power Industries. Springer,
pp. 41–54.
eliminating unrealistic workload and production pressure.
• Maintain well-designed equipment and tools availability.
Ekanem, N.J., Mosleh, A., Shen, S.-H., 2016. Phoenix – a model-based human reliability
analysis methodology: qualitative analysis procedure. In: Reliab. Eng. Syst. Saf. 145.
• Clarify the issues of personnel, although sometimes it may be per- pp. 301–315. https://doi.org/10.1016/j.ress.2015.07.009.
Embrey, D., Humphreys, P., Rosa, E.A., Kirwan, B., Rea, K., 1984. SLIM-MAUD: an ap-
sonal.
proach to assessing human error probabilities using structured expert judgment.
Volume I. Overview of SLIM-MAUD.
8. Conclusion and limitation Feng, Y., Chen, H., Zhang, Y., Jing, L., 2019. The hybrid systems method integrating
human factors analysis and classification system and grey relational analysis for the
analysis of major coal mining accidents. Syst. Res. Behav. Sci. 36, 564–579.
Present work demonstrates a methodology to calculate the human Feyer, A.-M., Williamson, A.M., 1998. Human Factors in Accident Modelling. In:
error rate in various system activities from the analysis of past accident Stellman, J.M. (Ed.), Encyclopaedia of Occupational Health and Safety,
reports and therefore resolves the limitations of the opinion based Encyclopaedia of Occupational Health and Safety. International Labour Office, p.
56.8.
prevalent methods for error rate or error probability estimation. This Forester, J., Bley, D., Cooper, S., Lois, E., Siu, N., Kolaczkowski, A., Wreathall, J., 2004.
paper mainly focuses on dealing the accident description where the Expert elicitation approach for performing ATHEANA quantification. Reliab. Eng.
embedded uncertainty in accident data classification and human error Syst. Saf. 83, 207–220.
Gertman, D., Blackman, H., Marble, J., Byers, J., Smith, C., et al., 2005. The SPAR-H
categorisation is captured using fuzzy mathematical concepts for error human reliability analysis method. US Nucl. Regul, Comm.
accountability. The results of the present analysis specifically conclude Griffith, C.D., Mahadevan, S., 2011. Inclusion of fatigue effects in human reliability
that slip and KBM are the most prevalent in the mining industry and analysis. In: Reliab. Eng. Syst. Saf. 96. pp. 1437–1447. https://doi.org/10.1016/j.
ress.2011.06.005.
also identified that material handling and transportation are the most Gupta, S., Kumar, P., 2014. Assessment of human error rate in different underground coal
unsafe activities. Thus, safety enhancement program should be directed mining activities using rough set theory 16(2), 51–64. https://doi.org/10.6273/JES.
primarily towards the control and management of slips and KBM in 2014.16(2).01.
Gupta, S., Kumar, P., Karmakar, N.C., Palei, S.K., 2013. Quantification of human error
material handling and transportation activities.
rate in underground coal mines — A fuzzy mapping and rough set based approach. In:
As this work shows a human error criticality-based approach for 2013 IEEE International Conference on Industrial Engineering and Engineering
managing and control of accidents, the result of this analysis will serve Management, pp. 140–144. https://doi.org/10.1109/IEEM.2013.6962391.
as a guiding tool for the safety improvement in the Indian mining in- Gupta, S., Kumar, P., Yuga Raju, G., 2019. A fuzzy causal relational mapping and rough
set-based model for context-specific human error rate estimation. Int. J. Occup. Saf.
dustry. This work concludes that system and process design should Ergon. 1–16. https://doi.org/10.1080/10803548.2019.1578114.
focus on changing the work environment, not the worker and aim to

9
P. Kumar, et al. Safety Science 123 (2020) 104555

Hobbs, A., Williamson, A., 2003. Associations between errors and contributing factors in Rasmussen, J., 1997. Risk management in a dynamic society: a modelling problem. Saf.
aircraft maintenance. Hum. Factors J. Hum. Factors Ergon. Soc. 45, 186–201. https:// Sci. 27, 183–213. https://doi.org/10.1016/S0925-7535(97)00052-0.
doi.org/10.1518/hfes.45.2.186.27244. Reason, J., 2000. Human error: models and management. BMJ 320, 768 LP – 770.
Hollnagel, E., 1998. Cognitive Reliability and Error Analysis Method (CREAM). Elsevier https://doi.org/10.1136/bmj.320.7237.768.
Science. Reason, J., 1990. Human error. Cambridge University Press. https://doi.org/10.1017/
Kirwan, B., 2017. A Guide To Practical Human Reliability Assessment. CRC Press. CBO9781139062367.
Konstandinidou, M., Nivolianitou, Z., Kiranoudis, C., Markatos, N., 2006. A fuzzy mod- Rushworth, A., Talbot, C., Von Glehn, F., Lomas, R., 1999. Investigate the causes of
eling application of CREAM methodology for human reliability analysis. Reliab. Eng. transport and tramming accidents on coal mines.
Syst. Saf. 91, 706–716. https://doi.org/10.1016/j.ress.2005.06.002. Sanders, M.S., McCormick, E.J. (Ernest J., 1993. Human factors in engineering and de-
Kotek, L., Mukhametzianova, L., 2012. Validation of human error probabilities with sign. McGraw-Hill.
statistical analysis of misbehaviours. Procedia Eng. 42, 1955–1959. https://doi.org/ Sgobba, T., Kanki, B.G., Clervoy, J.F., Sandal, G., 2017. Space Safety and Human
10.1016/j.proeng.2012.07.591. Performance. Elsevier Science.
Kumar, P., Gupta, S., Agarwal, M., Singh, U., 2016. Categorization and standardization of Simpson, G., Joy, P.J., Horberry, P.T., 2012. Understanding Human Error in Mine Safety.
accidental risk-criticality levels of human error to develop risk and safety manage- Ashgate Publishing Limited.
ment policy. Saf. Sci. 85, 88–98. https://doi.org/10.1016/j.ssci.2016.01.007. Swain, A., 1987. Accident Sequence Evaluation Program: Human Reliability Analysis
Lewis, C., Norman, D.A., 1986. Designing for Error. In: Norman, D.A., Draper, S.W. (Eds.), Procedure. Nureg/Cr-4772. https://doi.org/10.2172/6370593.
User Centered System Design; New Perspectives on Human-Computer Interaction. L. Swain, A.D., Altman, J.W., Rook, L.W., Society., H.F., Meeting, Collection., H.F. and E.,
Erlbaum Associates Inc., Hillsdale, NJ, USA, pp. 411–432. Corporation., S., Technical Information Division, Applied Psychology Corporation,
Li, P Cheng, Chen, G.Hua, Dai, L Cao, Li, Z., 2010. Fuzzy logic-based approach for 1963. Human error quantification : a symposium, in: Human Error Quantification: A
identifying the risk importance of human error. Saf. Sci. 48, 902–913. https://doi. Symposium.
org/10.1016/j.ssci.2010.03.012. Swain, A.D., Guttmann, H.E., 1983. Handbook of human reliability analysis with em-
Marseguerra, M., Zio, E., Librizzi, M., 2007. Human reliability analysis by fuzzy phasis on nuclear power plant applications: Final report, Nureg/Cr-1278, Usnrc.
“CREAM”. Risk Anal. 27, 137–154. https://doi.org/10.1111/j.1539-6924.2006. Albuquerque, NM, and Livermore, CA (United States). https://doi.org/10.2172/
00865.x. 5752058.
Ministry of Statistics and Programme Implementation, 2017. Statistical Year Book [WWW Ung, S.T., Williams, V., Chen, H.S., Bonsall, S., Wang, J., 2006. Human error assessment
Document]. Gov. India. URL http://mospi.nic.in/sites/default/files/statistical_year_ and management in port operations using fuzzy AHP. Mar. Technol. Soc. J. 40,
book_india_2015/Table 38.1_4.xls. 73–86. https://doi.org/10.4031/002533206787353583.
Moore-Ede, M.C., 1993. The Twenty-Four Hour Society: Understanding Human Limits in Whaley, A.M., Hendrickson, S.M.L., Boring, R.L., Joe, J.C., Blanc, K.L. Le, Xing, J., 2012.
a World That Never Stops, A William Patrick Book. Addison-Welsey. Bridging Human Reliability Analysis and Psychology, Part 1: The Psychological
National Academies of Sciences and Medicine, E., 2018. The Human Factors of Process Literature Review for the IDHEAS Method. United States.
Safety and Worker Empowerment in the Offshore Oil Industry: Proceedings of a Wickens, C.D., Hollands, J.G., Banbury, S., Parasuraman, R., 2015. Engineering psy-
Workshop. The National Academies Press, Washington, DC. https://doi.org/10. chology and human performance. Psychology Press.
17226/25047. Wiegmann, D.A., Shappell, S.A., 2017. A human error approach to aviation accident
Pan, X., Lin, Y., He, C., 2017. A review of cognitive models in human reliability analysis. analysis: The human factors analysis and classification system. Routledge.
Qual. Reliab. Eng. Int. 33, 1299–1316. https://doi.org/10.1002/qre.2111. Wu, Z., Pan, X., Zhao, X., Jiang, Y., 2019. The task demands-resources method: a new
Patterson, J.M., Shappell, S.A., 2010. Operator error and system deficiencies: analysis of approach to human reliability analysis from a psychological perspective. Qual.
508 mining incidents and accidents from Queensland, Australia using HFACS. Accid. Reliab. Eng. Int. 35, 1200–1218. https://doi.org/10.1002/qre.2453.
Anal. Prev. 42, 1379–1385. https://doi.org/10.1016/j.aap.2010.02.018. Yin, W., Fu, G., Yang, C., Jiang, Z., Zhu, K., Gao, Y., 2017. Fatal gas explosion accidents on
Paul, P.S., Maiti, J., 2007. The role of behavioral factors on safety management in un- Chinese coal mines and the characteristics of unsafe behaviors: 2000–2014. Saf. Sci.
derground mines. Saf. Sci. 45, 449–471. https://doi.org/10.1016/j.ssci.2006.07.006. 92, 173–179. https://doi.org/10.1016/j.ssci.2016.09.018.
Perrow, C., 1999. Normal Accidents, REV-Revised, ed. Princeton University Press. Zhang, Y., Jing, L., Bai, Q., Liu, T., Feng, Y., 2019. A systems approach to extraordinarily
Philippart, M., 2018. Chapter 12 - Human reliability analysis methods and tools, in: major coal mine accidents in China from 1997 to 2011: an application of the HFACS
Sgobba, T., Kanki, B., Clervoy, J.-F., Sandal, G.M.B.T.-S.S. and H.P. (Eds.), . approach. Int. J. Occup. Saf. Ergon. 25, 181–193. https://doi.org/10.1080/
Butterworth-Heinemann, pp. 501–568. https://doi.org/https://doi.org/10.1016/ 10803548.2017.1415404.
B978-0-08-101869-9.00012-1. Zio, E., Baraldi, P., Librizzi, M., Podofillini, L., Dang, V.N., 2009. A fuzzy set-based ap-
Rashid, H.S.J., Place, C.S., Braithwaite, G.R., 2013. Investigating the investigations: a proach for modeling dependence among human errors. Fuzzy Sets Syst. 160,
retrospective study in the aviation maintenance error causation. Cogn. Technol. Work 1947–1964. https://doi.org/10.1016/j.fss.2009.01.016.
15, 171–188.

10