Open access peer-reviewed chapter

Prioritizing Human Factors in Emergency Conditions Using AHP Model and FMEA

By Fabio De Felice, Antonella Petrillo and Domenico Falcone

Submitted: November 9th 2015Reviewed: May 30th 2016Published: August 31st 2016

DOI: 10.5772/64411

Downloaded: 1483

Abstract

One of the most critical issues related to safety in industrial plant is to manage accidents that occur in industries. In general, the causes of accidents are twofold: the presence of dangerous equipment and human errors. The aim of this study is to propose a novel approach to ensure safety in emergency conditions in industrial plant considering both of these factors. The proposed idea aims to integrate the human reliability analysis (HRA) and the failure modes and effects analysis (FMEA). The human errors and failure modes are categorized using a multicriteria approach based on analytic hierarchy process (AHP). The final aim is to present a novel methodological approach based on AHP to prioritize actions to carry out in emergency conditions taking into account both qualitative and quantitative factors. A real case study is analyzed. The analysis allowed to identify possible failure modes connected with human error process.

Keywords

  • AHP
  • HRA
  • FMEA
  • emergency
  • human errors

1. Introduction

In recent years, various emergency events have caused loss of lives. The potential for major industrial accidents required a clearly and systematic approach to control the human errors and failure modes to ensure a proper approach in emergency conditions. One of the most important significant sources useful to monitor the emergency events is the database EM-DAT (http://www.emdat.be/). The database is a potential tool to investigate industrial accidents and disaster events worldwide. For instance, Figure 1 shows the total number of industrial accidents, in Europe, between 1900 and 2015. As it is possible to note, during the period 1990–2008, there has been a growth in the number of accidents. This result is not a surprise as a consequence of growth on industrial processes and technology.

Figure 1.

The total number of industrial accidents between 1900 and 2015 (source "http://www.emdat.be/").

Of course, the growth of industrial accidents, highlighted above, behave in accordance with an increase of the total number of economic damage as it is shown in Figure 2.

Figure 2.

The total economic damage between 1900 and 2015 (source http://www.emdat.be/).

Among the major disasters that occur worldwide are the Seveso disaster in Italy(July 1976), caused by the release of dioxins into the atmosphere; the Chernobyl disasterin Ukraine (April 1986) happened after a test on a reactor; the Shell Oil refineryexplosion in the USA (May 1988), caused by hydrocarbon gas leak; and the Fukushima Inuclear accidents in Japan (March 2011).

These undesirable events increased the importance of risk assessment and a good operational planning decisions [1, 2]. In this context, the critical elements to take into account are human factors and failure modes to design a systematic approach to improve the role of human element in a system performance. Thus, what is the point?In our opinion, it is necessary to provide a structured framework to prioritize risks taking into account the failure mode and the human errors.

For this reason in the present research, an integrated approach is proposed. In particular, the approach is based on three fundamental aspects that, in our opinion, help to analyze the problem globally: the human reliability analysis(HRA), the failure modes and effects analysis(FMEA), and the multicriteria approach using analytic hierarchy process(AHP). Here below, a brief overview of the above approaches is provided.

The HRA is a critical issue because the human errors involve the use of qualitative and quantitative aspects [3]. In literature, a great number of tools are proposed for human reliability analysis [4]. The limit of the HRA techniques is related to the uncertainty to consider organizational factors and errors of commission. One of the most important issues related to human errors is the performance shaping factors (PSFs). PSFs allow to assess all the environmental and behavioral factors that influence the decision and actions of man. In particular, the use of PSFs allows to simulate different scenarios. HRA tools include failure modes and effects analysis (FMEA), developed in the late 1950s. FMEA is one of the most popular methods used to perform the risk assessment [5]. The traditional FMEA approach has been extensively criticized due to the limitations in calculating the risk priority number (RPN) [6, 7].

Several methods have been suggested to improve the traditional HRA and FMEA. Among these methods, the multicriteria methods such as analytic hierarchy process (AHP) are very promising. AHP, developed by Saaty [8], is a useful approach to manage the complexity of decision problems. AHP technique is very suitable to determine the weights for each risk factor. With AHP, it is possible to evaluate the ambiguity of human perception and to transform it into a mathematical formula.

In this paper, the AHP technique is proposed to weigh and to prioritize failure modes and human factors that characterize actions in emergency conditions. AHP is used to assess the relative importance among human factors and failure modes. As a result, a new methodological approach to calculate the risk priority number using weighting method integrating with human factors is presented. To validate the application of the model and to examine its effectiveness, the proposed methodology is used for analyzing an emergency scenario in a petrochemical plant.

The rest of the paper is organized as follows. The literature review on emergency management is presented briefly in Section 2. The methods are introduced in Section 3. Section 4 is about the proposed evaluation methodology. Furthermore, a numerical example is presented. Conclusions are analyzed in Section 5.

Advertisement

2. Literature review on emergency management and human errors

In emergency management, it is necessary to collect many and different types of information to deal with emergency tasks in different phases and to deal with emergency situation such as stress, lack of time, and adequate training [9].

A systematic literature review was carried out to search if a similar study has already been published in the literature. We looked for studies in the emergency managementarea in Scopus database (www.scopus.elsevier.com). In particular, we investigated the “engineering field,” in which 8,445 documents came out during the period of 1925–2016. Figure 3 shows the number of documents by year. Results point out that there is a growing interest on emergency management issue.

Figure 3.

Documents by year.

A depth analysis shows that the majority of the researches are developed in the USA (Figure 4).

Figure 4.

Documents by country/territory.

In recent literature, several approaches and models have been proposed to manage emergency conditions. For instance, Pérez et al. [10] in their study proposed a fleet assignment model for the Santiago Fire Department to maximize the number of incidents successfully attended. Omidvari et al. [11] developed a model based on analytical hierarchy process and failure modes and effects analysis logic to determine the factors influencing the fire risk of an education center. Bariha et al. [12] proposed the analysis of hazards associated with accidental release of high pressure from gas-pipeline transportation system. A probabilistic risk assessment from potential exposures to the public applied for innovative nuclear installations has been analyzed by Dvorzhak et al. [13]. While Shi et al. [14] used a technique plan repository and evaluation system based on AHP group decision-making for emergency treatment and disposal in chemical pollution accidents, Ergu et al. [15] applied the analytic network process, a generalization of the AHP, in risk assessment a decision analysis. Liu et al. [16] present a novel approach for FMEA based on AHP and fuzzy VIKOR method.

The above works relate to emergency management in the sense that they analyze multiple approaches to manage emergency conditions. However, no specific systematic approach is presented to take into account both qualitative and quantitative factors in the emergency management. Our work focuses in understanding the failure modes and human errors in emergency conditions, following a strict and standardized protocol to identify and to prioritize risks.

3. Methods

To design the model, a novel approach including HRA, FMEA, and AHP is defined. In this section, a short description of these methods is given.

3.1. The failure modes and effects analysis (FMEA): the traditional approach

According to the American Society for Quality [17], FMEA is a procedure that is performed after a failure modes and effects analysis to classify each potential failure effect according to its severity and probability of occurrence. Traditional FMEA method is based on risk priority number (RPN) for analyzing the risk associated with potential problems. RPN is calculated by multiplying the scores of severity (S), occurrence (O), and detection (D). In details:

  1.  Occurrence (O) represents the probability that a particular cause for the occurrence of a failure mode occurs (1–10). The range of occurrence scale is from score 1 (failure is unlikely) to score 10 (failure is very high and inevitable).

  2.  Severity (S) represents the severity of the effect on the final process outcome resulting from the failure mode if it is not detected or corrected. The severity scale is from score 1 (none effect) to score 10 (hazardous without warning).

  3.  Lack of detectability (D) represents the probability that the failure will not be detected. The detectability scale is from score 1 (detection almost certain) to score 10 (absolute uncertainty).

Generally, it will give more importance to the failure modes with higher RPNs. The RPN method has been criticized due to its limitations, among which are:

  1.  The relative weights of risk factors are not taken into account.

  2.  S, O, and D indexes have the same relevance.

  3.  Different sets of O, S, and D scores produce the same value of RPN.

  4.  They are ordinal type of scale (not cardinal). Thus, no arithmetic operation among them is permitted.

To avoid some of the drawbacks mentioned above, some different approaches have been proposed in literature. For instance, Narayanagounder and Gurusami [18] used analysis of variance (ANOVA) to prioritize failure modes or using additional characteristic indexes to define the order to analyze (from a design point of view) the failure mode. However, this proposal are very complex, and it is not simple to apply it in industrial context. For this reason, our study aims to propose a simple application using AHP technique according to human reliability analysis that is explained in the next sections.

3.2. The human reliability analysis (HRA)

The development of human reliability methods occurred over time in three stages [19]: (1) the first stage (1970–1990), known as the first human reliability method generation; (2) the second phase (1990–2005), known as the second human reliability method generation; and (3) finally, the third generation, started in 2005 and still in progress [20, 21].

Among the several approaches that have been proposed and developed for error classification, we remember the Systematic Human Error Reduction and Prediction Approach (SHERPA) [22], the Cognitive Reliability Error Analysis Method (CREAM) [23], the Human Error Identification in Systems Tool (HEIST) and Human Error Assessment and Reduction Technique (HEART) [24], and the Human Factors Analysis and Classification System (HFACS) [25], among others.

Figure 5.

The Weibull function.

The common basis for all techniques is the assessment of the human error probability (HEP) representing the index of human error and its variation according to PSFs. The analysis of human error starts with a definition of HEP. Obviously, the probability of error is a growing function of the time. The HEP value can be calculated as:

HEPnom=1-e-αtβE1

The above formula can be changed in relation to the working hours. Since the operator’s reliability is highest in the first hour of work and descending gradually, it gets closer to the eighth hour, and the unreliability decreases with the time (Figure 5).

HEPnomt=1-k*e-α1-tβt01HEPnomt=1-k*e-αt-1βt]1;[E2

where α is the scale parameter and β is the shape parameter. The β parameter is defined as 1.5 by the scientific literature of the HEART method.

The α parameter is calculated by the following formula:

α=-lnkt=8kt=1t-1βE3

Using (Eq. 2), it is possible to define the trend of reliability associated to specific generic task or human errors, as it is shown in Table 1.

# Human errors/generic taskLimitations of unreliability for operation (%)k (t = 1) k (t = 8)
1Total unfamiliar0.35–0.970.650.03
2Shift or restore system to a new or original state0.14–0.420.860.58
3Complex task0.12–0.280.880.72
4Fairly simple task0.06–0.130.940.87
5Routine highly practiced0.007–0.0450.9930.955
6Restore or shift a system to original or new state0.008–0.0070.9920.993
7Completely familiar0.00008–0.0090.999920.991
8Respond correctly0.00000–0.000910.9991

Table 1.

Human errors scale.

One of the most important aspects is the study of factor interactions that increase the probability of error and interdependencies of performance shaping factors (PSFs). Specifically, the context in which humans make errors is analyzed. The PSFs allow the inclusion, in the model, of all the environmental and behavioral factors that influence the decision and actions of man [26]. In particular, the use of PSFs allows to simulate different scenarios. Analytically, the PSFs are modifying the value of the error probability because they introduce external factors that strain and distract the decision maker [27]. The PSF and their values are obtained from the literature, as is shown in Table 2. The PSFs considered are (1) available time, (2) stress/stressor, (3) complexity, (4) experience and training, (5) procedures, (6) ergonomics, (7) fitness for duty, and (8) work processes.

#PSFsLevelsValues
1Available timeInadequate time1
Time available = time required adequate time10
Nominal time1
Time available > 5 time required (extra time)0.1
Time available > 50 time required0.01
2StressExtreme5
High2
Nominal1
3ComplexityHighly complex5
Moderately complex2
Nominal1
4Experience/trainingLow3
Nominal1
High0.5
5ProceduresNot available50
Incomplete20
Available, but poor5
Nominal1
6ErgonomicsMissing50
Poor10
Nominal1
good0.5
7Fitness for dutyUnfit1
Degraded fitness5
Nominal1
8Work processesNominal1
Good0.5

Table 2.

PSFs scale.

It is important to note that four main sources of deficiencies can be identified in current HRA methods:

  1.  No empirical data for model development and validation

  2.  Lack of inclusion of human cognition

  3.  Large variability in implementation

  4.  Heavy reliance on expert judgment in selecting PSFs and the use of these PSFs to obtain the HEP in human reliability analysis

The limitations characterizing the FMEA and HRA methods are the motivation under our study and provide the reason to integrate them with AHP technique.

3.3. The analytic hierarchy process (AHP)

AHP uses mathematical objectives to process the inescapable, subjective, and personal preferences of an individual or a group making a decision [28]. In the AHP process, firstly, the hierarchy is defined. Secondly, judgments on pairs of elements with respect to a controlling element are expressed to derive ratio scales that are then synthesized throughout the structure used to select the best alternative [29]. The modeling process can be divided into different phases; to provide a better understanding of the main phases, they are described as follows:

  1. 1. Pairwise comparison and relative weight estimation.Pairwise comparisons of the elements in each level are conducted with respect to their relative importance toward their control criteria. Saaty suggested a scale of 1–9 when comparing two components. For example, a score of 9 represents an extreme importance over another element, while a score of 8 represents an intermediate importance between “very strong importance” and “extreme importance” over another element. The pairwise comparisons can be represented in the form of a matrix. Score 1 represents equal importance of two components, and score 9 represents extreme importance of the component i over the component j.

  2. 2. Priority vector.After all pairwise comparisons are completed, the priority weight vector (w) is obtained.

  3. 3. Consistency index estimation. The consistency index (CI) of the derived weights could then be calculated by CI = (λmax − n)/n − 1., where λmax is the largest eigenvalue of the judgment matrix A and n is the rank of the matrix. In general, if CI is less than 0.10, the satisfaction of the judgments may be derived.

4. Research framework: Human reliability model based on AHP

In this section, a human reliability model is proposed to analyze the human reliability in emergency conditions, based on failure modes and effects analysis. Here, below is a brief description of the phases and steps characterizing the model.

4.1. The rationale

As reported in the previous sections, a quantitative method was developed to prioritize failure modes and human errors using AHP technique. The logic model is depicted in Figure 6. Each model phase is discussed in brief, as follows:

Figure 6.

Methodological approach.

Phase #1: Preliminary analysis.The aim of the first phase was to identify scenario under study defining failure modes and human errors characterizing the accident. This phase represents the most critical phase because an incorrect assessment could determine the inappropriateness of the overall model. Three main steps were identified, as follows:

  1.  Step #1: Identify of risk factors

  2.  Step #2: Identify failure modes (FMs)

  3.  Step #3: Identify human errors (HRs)

Phase #2: Multicriteria analysis.One of the crucial elements in measuring the effectiveness of a decision model is to obtain consistent results according to common standards. Thus, the aim of the second phase was to define the AHP model according to prioritize failure modes (FMs) and human errors (HRs), when there is a disagreement in ranking scale for severity, occurrence, detection, human errors, and PSFs. The final purpose of this phase was to define the final score for each criteria and subcriteria according to the AHP model and according to the expert team.

Four main steps were identified, as follows:

  1.  Step #4: Construction of the AHP model

  2.  Step #5: Pairwise comparison for each expert

  3.  Step #6: Analysis of the consistency

  4.  Step #7: Aggregation of expert judgments

  5.  Step #8: Ranking

Phase #3: Synthesis.The aim of the last phase was to define the final assessment for RPN according to the expert team. One step was identified to carry out the phase, as follows:

  1.  Step #9: Final assessment

4.2. Case study: the model validation

To demonstrate the proposed approach, a real-world application is employed in this section. The example is related to an emergency shutdown valve (ESDV) in a petrochemical plant. Of course, reliability is crucial to safety (Figure 7). An emergency shutdown valve is an actuated valve designed to stop the flow of a hazardous fluid upon the detection of a dangerous condition, protecting people, equipment, or the environment.

Figure 7.

Example of emergency shutdown valve.

4.2.1. Phase #1: preliminary analysis

According to Step #1, the identification of risk factors were carried out. In detail, an expert team, composed by five engineers, was selected to develop the model.

According to Step #2, the main failure modes characterizing the malfunction of an ESDV were identified: valve open (FM1), partially open (FM2), closed (FM3), partially closed (FM4), and wobble (FM5). The main failure causes were identified in broken and corrosion.

Scales for S, O, and D were defined by each expert, as is shown in Table 3.

Failure modeSODRPN
Expert#1
FM124756
FM235690
FM326448
FM4475140
FM547256
Expert#2
FM1357105
FM224864
FM3376126
FM4638144
FM577149
Expert#3
FM118324
FM244696
FM3587280
FM433545
FM524540
Expert#4
FM1387168
FM242648
FM3576210
FM463590
FM524540
Expert#5
FM1387168
FM214416
FM3546120
FM464248
FM524540

Table 3.

Failure mode – for each experts.

The aggregation of values expressed by each experts was made computing the modeor in other words identifying the value that appeared most often in the set of data. For instance, if we consider the set data for FM1 related to the severity scale, the most frequent value or the mode value is 3. The final set of data used is shown in Table 4.

Table 4.

Calculation of the mode values for the five experts

Frequently, human errors are overlooked. Because the FMEA examines individual faults of system elements taken singly, the combined effects of coexisting failures are not considered. Thus, according to Step #3, the human errors (HR) were identified by the expert team. For the scenario under study, three HRs and three PSFs were selected, as is shown in Tables 5 and 6.

#IDHuman errors/generic taskLimitations of unreliability for operation (%)k (t = 1)k (t = 8)
2HR1Shift or restore system to a new or original state0.14–0.420.860.58
5HR2Routine highly practiced0.007–0.0450.9930.955
7HR3Completely familiar0.00008–0.0090.999920.991

Table 5.

Human errors selection.

#IDPSFsLevelsValues
1PSF1Available timeInadequate time
Time available = time required adequate time
Nominal time
Time available > 5 time required (extra time)
Time available > 50 time required
1
10
1
0.1
0.01
2PSF2StressExtreme
High
Nominal
5
2
1
3PSF3ComplexityHighly complex
Moderately complex
Nominal
5
2
1

Table 6.

PSFs scale.

Scales for PSFs were defined by each expert. In Table 7, a synthesis is shown.

Human errorPS1PS2PS3
HR11011
HR2151
HR3512

Table 7.

PSF scale—experts—synthesis.

4.2.2. Phase #2: AHP model

According to Step #4, the AHP model was built to determine the criteria and subcriteria weights. The model consists of eight criteria and six subcriteria. In Figure 8, the AHP model is shown.

Figure 8.

R-AHP model.

FM1FM2FM3FM4FM5Weight
FM1155660.547
FM21/513440.215
FM31/51/31330.121
FM41/61/41/3120.065
FM51/61/41/31/210.050

Table 8.

Example of pairwise comparison for potential failure modes criteria.

Consistency: 0.078 < 0.10.


In this phase, according to Step #5, pairwise comparison matrices were filled out by the expert team to define the weights of criteria and subcriteria. It is worth noting that the same importance was attributed to the main criteria or 50 % potential failure mode and 50 % potential human errors. Tables 8 and 9 show two examples of pairwise comparison. For each pairwise comparison, according to Step #6, consistency analysis was carried out.

HR1HR2HR3Weight
HR111/31/30.139
HR2311/20.332
HR33210.527

Table 9.

Example of pairwise comparison for potential error criteria.

Consistency: 0.051 < 0.10.


Figure 9.

Final weights.

Figure 10.

Example of initial and revised severity scale for Expert #1.

The determination of relative weights in AHP model is based on the pairwise comparison conducted with respect to their relative importance toward their control criterion. In detail, evaluation expressed in Tables 47 was used. Furthermore, Saaty’s semantic scale was used for the comparison.

Failure modeSODRPN revised
Expert #1
FM1573105
FM245480
FM335460
FM425440
FM525220
Expert #2
FM1457140
FM236590
FM336590
FM463472
FM567142
Expert #3
FM1583120
FM2554100
FM3554100
FM444348
FM544348
Expert #4
5831205
454804
572705
432244
244322
Expert #5
3851203
5541005
5451005
552505
244322

Table 10.

Revised failure mode—for each expert.

Furthermore, according to Step #7, the geometric meanwas used to synthesize the set of judgments given by the expert team.

According to Step #8, the ranking was obtained. The aim of the present step was to define the final score for each criteria and subcriteria according to the AHP model and according to the expert team. One step was identified to carry out the phase, as follows. The scores of these each weighted criteria and subcriteria are shown in Figure 9.

Results pointed out that, according to expert judgments, the occurrence is the most important parameter with a score of 46.4 % than the other (severity with a score of 32.4 % and detectability with a score of 21.2 %). Furthermore, the most critical failure mode is FM1 with a score of 54.7 %, the most critical human error is HR3 with a score 52.7 %, and the most important performance shape factor is PS3 with a score of 47.9 %.

4.2.3. Phase #3: a new approach for prioritizing human errors and failure modes

Phase #3 is a crucial phase in the proposed methodological approach. In fact, it is important to note that, for instance, for Expert #1, FM1 and FM 5 had the same RPN (56) with different ranking values for occurrence, severity, and detection. For determining the most significant failure mode H-RPN, the weights defined through AHP model were used, as is shown in Table 9.

In detail, according to the final weights obtained with AHP, each expert expressed a reassessment judgments for S, O, and D scales. Figure 10 shows an example of revised severity scale for Expert #1.

In a similar way, revised scales for occurrence and detectability were obtained by each expert, as shown in Table 10.

The aggregation of values expressed by each expert was made computing the mode, as shown in Table 11.

Table 11.

Revised calculation of the mode values of five experts.

Table 12 shows an example of new RPN.

SeverityOccurrenceDetectionNew RPN
Initial387168
Revised583120
Multicriteria RPN = % reduction in RPN = (RPN initial—RPN revised)/RPN initial29 %

Table 12.

Example of calculation of RPN.

Figure 11 shows a comparison between initial and revised RPNs, highlighting the RPN reductionfor each FM. It is significant to point out that in general Figure 11 shows, for the specific case study, a reduction of the RPN because usually the expert team have overestimated one parameter. But an increase of RPNcould happen also.

Figure 11.

Comparison between initial and revised RPN.

Of course, in our opinion, the above assessment allows to evaluate the RPN more precisely and objectively with respect to the traditional approach.

5. Conclusion

This paper proposes a novel approach to prioritize failure modes taking into account human errors. The final aim was to improve the evaluation of risk priority number integrating human errors in the calculation. The goal was achieved through a multicriteria model, based on AHP. The method allowed us to weigh the failure modes integrating with human errors and to prioritize failure modes. The model can be applied when there is a disagreement in ranking scale for severity, occurrence, and detection. The novelty of the method is because there is no evidence in literature of this kind of approach using AHP. In our opinion, the proposed method ensures several benefits, as detailed follows: (1) it is a generic method that can be applied in several industrial processes; (2) it can be used to identify human errors that can become single points of failure; and (3) it can be used to define potential human errors that are the most critical by revealing the severity and probability of occurrence. Future research will investigate a great number of failure modes and human errors. Furthermore, several scenarios will be taken into account to compare results.

© 2016 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

How to cite and reference

Link to this chapter Copy to clipboard

Cite this chapter Copy to clipboard

Fabio De Felice, Antonella Petrillo and Domenico Falcone (August 31st 2016). Prioritizing Human Factors in Emergency Conditions Using AHP Model and FMEA, Applications and Theory of Analytic Hierarchy Process - Decision Making for Strategic Decisions, Fabio De Felice, Thomas L. Saaty and Antonella Petrillo, IntechOpen, DOI: 10.5772/64411. Available from:

chapter statistics

1483total chapter downloads

More statistics for editors and authors

Login to your personal dashboard for more detailed statistics on your publications.

Access personal reporting

Related Content

This Book

Next chapter

Disaster Risk Assessment Developing a Perceived Comprehensive Disaster Risk Index: The Cases of Three Chilean Cities

By Carmen Paz Castro, Juan Pablo Sarmiento and Claudio Garuti

Related Book

First chapter

A Cognitive Model for Emergency Management in Hospitals: Proposal of a Triage Severity Index

By Marco Frascio, Francesca Mandolfino, Federico Zomparelli and Antonella Petrillo

We are IntechOpen, the world's leading publisher of Open Access books. Built by scientists, for scientists. Our readership spans scientists, professors, researchers, librarians, and students, as well as business professionals. We share our knowledge and peer-reveiwed research papers with libraries, scientific and engineering societies, and also work with corporate R&D departments and government entities.

More About Us