1 Introduction

World is facing lots of challenges in front of environmental pollution and its protection. It has initiated strict norms and regulation for gas emission due to burning of fuel. This has necessitated use of fuels that have low carbon dioxide (CO2), nitrogen oxide (NOx), sulphur dioxide (SO2) and particulate matter (PM) emission. Compressed natural gas (CNG), is one such alternative fuel that has low emissions of the above-mentioned gases and PM. The use of CNG instead of gasoline reduces environmental pollution, as there is reduction of 26% CO2 emission thereby reducing greenhouse effect considerably. The use of CNG as alternative automotive fuel has increased in recent years. There are different standards of CNG for different purpose. IS:15958 applies to CNG as fuel for automotive purpose and ISO: 16923:2016 for CNG dispensing system. CNG is a form of natural gas (mainly methane CH4) compressed at high pressure. CNG is considered a safe automotive fuel alternative as it doesn’t burn below 5% and above 15% composition respectively, when mixed with air. Though, the hazards at dispensing station are high. Being lighter than air, there is always a possibility of leakage due to faulty sub-systems of dispensing station. As CNG is inflammable, its leakage can lead to fire and explosion hazard at dispensing station. This can affect components of system, process, monitory functions, personals and people repectively. Therefore, proper maintenance of CNG dispensing station is essential. For proper maintenance planning, identification of causes of failure and their prioritization is essential. Risk can be defined situation of exposure to danger that can be initiated by failure event. Risk analysis is essential:

  1. 1.

    To improve safety level of various operating system.

  2. 2.

    To identify component whose probability of failure is high.

  3. 3.

    To optimize maintenance planning.

  4. 4.

    To support emergency planning.

Identification of risk using FMEA is a proactive approach that can reduce facility, monetary and human life loss in case of future catastrophic failure. The FMEA approach is subjected to intrinsic limitations. It is a group activity that depends on experience, knowledge and preference of cross-functional team of experts. There is always a chance of error or biasness in judgement that can lead to uncertainty. Fuzzy methodology can be used to overcome this limitation. The present research expounds the application of fuzzy integrated approach of FMEA and GRA. A case from CNG dispensing station has been selected for exemplifying the proposed framework, as it is a complex system comprising of various sub-components. These sub components are connected in combined series and parallel configurations, which makes its risk analysis pertinent for proper maintenance planning. The paper is structured as, Sect. 2 discuss previous research work in area of risk analysis of system, Sect. 3 which expounds flowchart/sequential steps followed for analysis, Sect. 4 discuss different techniques used for risk prioritization, Sect. 5 discuss the process flow of CNG dispensing station and subsequent analysis according to research methodology, Sect. 6 discuss analysis of result obtained, Sect. 7 discuss outcome of study and managerial implication of the present work.

2 Literature review

In an industrial system, risk of a system failure is always there. Therefore, for maintaining high availability of a system it is necessary to eliminate or minimize various types of risk. FMEA is a systematic technique that is used for risk analysis of complex systems by identifying possible future failures and its prioritization. The risk mainly depends on three factors, occurrence, severity and detection. Occurrence (Oc) indicates frequency of failure over span of time. Severity (Sr) indicates adverse effect of a failure. Non-detection (Dn) indicates the likelihood of the failure traceability. Different mathematical models have been developed for failure analysis of the system. Amuthakkannan et al. (2008) applied FMEA and other quality control tools for upgrading the software quality used for improving the performance of an internal combustion engine. Arabian-Hoseynabadi and Tavner (2010) applied FMEA approach for improving quality and reliability of wind turbine system. Hekmatpanah et al. (2011) implemented FMEA to improve quality of products in an Iranian Oil Company. Gaula and Sharma (2015) implemented FMEA to analyse the effect of maintenance strategies on throughput of a typical flexible manufacturing cells (FMC’s). Feili et al. (2013) demonstrated the application of FMEA approach for risk analysis in geothermal power plant. The conventional Failure Mode and Effect Analysis is susceptible to intrinsic faults and these drawbacks are as follow.

  1. a.

    It is a group decision activity and is performed by a cross functional team. There is every chance of a biased opinion.

  2. b.

    It is dependent on raw data from various sources that are often vague and imprecise.

To overcome limitation of conventional FMEA, fuzzy methodology can be integrated with it and be applied in diverse fields Srivastava et al. (2017). Xu et al. (2002) applied fuzzy logic based FMEA in the diesel engine system. Fuzzy integrated FMEA and GRA approach was used for risk analysis of water treatment plant (Panchal and Kumar 2016a, b) and further for real industrial system (Panchal and Kumar 2017a, b). An approach was used for criticality analysis of components of sub-system of nuclear power plant by Guimaraes and Lapa (2004). Das et al. (2011) implemented an aggregative fuzzy risk analysis for the flood incident management. There have been studies of failure analysis on complex system using integrated approach. Chen et al. (2014) applied integrated approach of fuzzy FMEA and fault tree for prioritization of risk in oxygen enhance combustor. Du et al. (2014) proposed fuzzy FMEA approach using evidential reasoning (ER) and the technique for order preference by similarity to ideal solution (TOPSIS). Ilangkumaran et al. (2014) developed an integrated model for assigning the risk priorities/ranking to the risky components of a system in a paper industry. Tang (2015) proposed a framework for decision making based on GRA and Dempster AND Shaffer theory. Zhou and Thai (2016) expounded application of fuzzy FMEA and GRA for relative ranking failure causes for tanker equipment. Ighravwe and Oke (2017) implemented fuzzy-grey-weighted aggregate sum product assessment methodical approach for multi-criteria analysis of maintenance performance systems. Panchal et al. (2018a, b) expounded the application of GRA for validating results of fuzzy RPN of transmission system in heavy motor vehicle.

It is clear from the preceding section that FMEA is a powerful tool for risk analysis. The uncertainty can be removed by using fuzzy methodology. This prioritization results from fuzzy GRA can be used for checking, comparing and validating the results as obtained by FMEA and fuzzy FMEA. From the reviewed literature, author observes that there is either no or less inference of use of this methodology for risk analysis of CNG dispensing system. The paper seeks to fill this gap by applying this methodology for risk analysis of above mentioned system located in northern part of India.

3 Research methodology

The research methodology for the present study is shown in Fig. 1.

Fig. 1
figure 1

Research methodology

Full size image

Steps Involved:

  • With personal interaction with operators, maintenance analysts and engineer, cause and effect diagram for system under study is made.

  • A proper scale is selected and information of Oc, Sr and Dn for FMEA is collected from cross-functional team of experts and log book.

  • The RPN values are calculated for conventional FMEA.

  • For fuzzy FMEA, input and out-put variables are defined by linguistic terms, approximated by triangular or trapezoidal fuzzy membership function. The if–then rules are made in fuzzy inference system (FIS).

  • Using these rules, output is processed in FIS and converted again into crisp value by using some suitable method. This crisp output will be fuzzy RPN for respective cause of failure. The if–then rule in rule base of FIS will become fuzzy decision support system for any type of input–output variables used in risk analysis.

  • The RPN and fuzzy RPN values are prioritized and compared for each cause of failure.

  • The prioritization is compared from the prioritization as obtained by fuzzy GRA approach. The degree of relationship is measure of prioritization. It is evaluated by establishing comparative, standard and difference series and evaluating grey relation coefficient.

4 Methods and approach

This section discusses various approach used for risk analysis and prioritization.

4.1 Cause and effect diagram

Cause and effect diagram, also called as fish-bone diagram, is an effective technique for assessing cause of failure associated with system under study. The system under study is placed on extreme right end of a horizontal line. The sub-systems or any identified source, are placed at branches protruding out of horizontal line. The failure causes for each sub-system are directed towards it (Fig. 2).

Fig. 2
figure 2

Cause and effect diagram

Full size image

4.2 FMEA

This technique was proposed by US Military for assessment of weapon system reliability in the year 1949. Further this technique was used by NASA for risk assessment of space program (Apollo Mission) in the year 1960. This technique is extensively used for risk analysis in aviation, automotive, manufacturing, medical, power plants (nuclear, thermal, hydraulics, wind energy, solar energy etc.), paper plant and food process industries respectively. It is a systematic and knowledge-based approach (Amuthakkannan et al. 2008; Panchal and Kumar 2016a, b), which is used to assess possible causes of failure, its frequency, severity on system and detection probability (for systems and sub-systems), so that effective and timely maintenance planning leads to avoidance of failure and improve availability of system. The product of these variables is called risk priority number (RPN), used for risk analysis and prioritization (Eq. 1).

$$RPN = O_{c} \times S_{r} \times D_{n}$$
(1)

4.3 Fuzzy FMEA

Fuzzy set theory has been used in research work to take care of imprecise and vague judgment resulting in uncertainty. The linguistic variables are used to by expert to represent events that are imprecisely and vaguely defined and can be used for transforming subjective knowledge into quantitative terms. This transformation requires a well-defined scale (Table 1).

Table 1 Fuzzy linguistic scale for variable
Full size table

These linguistic variables are defined by fuzzy set membership function. There are different types of membership function e.g. triangular, trapezoidal, normal etc. Mostly, triangular and trapezoidal membership functions are used for the sake of their simplicity and ease of computations. Triangular membership function (TMF) is defined by Eq. 1.

$$\mu_{{\tilde{T}}} \left( x \right) = \left\{ {\begin{array}{*{20}l} {\frac{x - a}{b - a}} \hfill & { a \le x \le b} \hfill \\ {\frac{x - c}{{\begin{array}{*{20}c} {b - c} \\ \end{array} }}} \hfill & { b \le x \le c} \hfill \\ {0 } \hfill & {otherwise} \hfill \\ \end{array} } \right\}$$
(2)

where a, b, c is the lower, mean and upper bound respectively and \(\tilde{T}\) is TFN represented by (a, b, c).

Similarly, Trapezoidal membership function is defined by Eq. 3.

$$\mu_{{\widetilde{Tr }}} \left( x \right) = \left\{ {\begin{array}{*{20}l} {0,} \hfill & {\left( {x > d\;or\; x < a} \right)} \hfill \\ {\frac{x - a}{b - a} ,} \hfill & { a \le x \le b} \hfill \\ {1,} \hfill & { b \le x \le c} \hfill \\ {\frac{d - x}{d - c},} \hfill & { c \le x \le d} \hfill \\ \end{array} } \right\}$$
(3)

The first step in process is fuzzification of input and output variables by defining membership functions. Then if–then rules are made in rule base (Panchal and Kumar 2016a, b; Panchal et al. 2018a, b).

These rules are processed in FIS to get an fuzzified output. Due to simplicity and ease of computation, center of area method is used in present research work (Eq. 4).

$$\tilde{x} = \frac{{\int_{{x_{1} }}^{{x_{2} }} {x \mu_{{\tilde{T}}} \left( x \right)dx} }}{{\mu_{{\tilde{T}}} \left( x \right)dx}}$$
(4)

where \(\mu_{{\tilde{T}}} \left( x \right)\) membership function and \(\tilde{T}\) is output fuzzy set.

The output after de-fuzzification will be fuzzy-RPN value as evaluated by Eq. 5.

$${\text{Crisp}}\;{\text{value}}\;{\text{of}}\;{\text{TFN}} = \frac{{\left( {a + 4b + c} \right)}}{6}$$
(5)

The process flow for fuzzy decision support system is shown in Fig. 3.

Fig. 3
figure 3

Fuzzy decision support system

Full size image

4.4 Grey relation analysis

Grey theory was proposed by Deng (1989). For most of the real-world system, capturing perfect information for model development is impossible. The grey system theory is one of the powerful tool for solving problems related to prioritization of decision making units (DMU), alternatives, strategies etc. The grey relation analysis (GRA) is part of the grey system. MCDM problems have been solved using GRA technique, like, human resource process (Olson and Wu 2006), facility layout and dispatching rules (Kuo et al. 2008), supplier selection for vehicle components based on green supply chain (Li and Zhao 2009) warehouse location problem (Özcan et al. 2011), selection of advanced manufacturing system (Goyal and Grover 2012), evaluation of innovation competency of aviation (Zhu et al. 2012), resilient supplier selection (Rajesh and Ravi 2015). The comparative and standard series in GRA for n decision units is given by Eqs. 6 and 7 respectively (Deng 1989; Panchal et al. 2018a, b):

$$Z_{0} = \left[ {z_{0} \left( 1 \right),z_{0} \left( 2 \right), \ldots \ldots \ldots \ldots z_{0} \left( k \right) \ldots \ldots . z_{0} \left( n \right)} \right]$$
(6)
$$Z_{i} = \left[ {z_{i} \left( 1 \right),\;z_{i} \left( 2 \right), \ldots \ldots \ldots \ldots z_{i} \left( k \right) \ldots \ldots . z_{i} \left( n \right)} \right]$$
(7)

where i = 1,2,3………m and k = 1,2,3………n.

\(z_{0 } \left( k \right)\) and \(z_{i} \left( k \right)\) numbers in standard and comparative series respectively.

$${\text{Comparative}}\;{\text{Series}} = \left[ {\begin{array}{*{20}l} {z_{1} \left( 1 \right)} \hfill & {z_{1} \left( 2 \right)} \hfill & {z_{1} \left( 3 \right)} \hfill \\ \vdots \hfill & \vdots \hfill & \vdots \hfill \\ {z_{m} \left( 1 \right)} \hfill & {z_{m} \left( 2 \right)} \hfill & {z_{m} \left( 3 \right)} \hfill \\ \end{array} } \right]$$
(8)
$${\text{Standard}}\;{\text{Series}} = \left[ {\begin{array}{*{20}l} {z_{0} \left( 1 \right)} \hfill & {z_{0} \left( 2 \right)} \hfill & {z_{0} \left( 3 \right)} \hfill \\ \vdots \hfill & \vdots \hfill & \vdots \hfill \\ {z_{m} \left( 1 \right)} \hfill & {z_{m} \left( 2 \right)} \hfill & {z_{m} \left( 3 \right)} \hfill \\ \end{array} } \right]$$
(9)
$$\varPhi = \left[ {\begin{array}{*{20}l} {\Delta_{01} \left( 1 \right)} \hfill & {\Delta_{01} \left( 2 \right)} \hfill & {\Delta_{01} \left( 3 \right)} \hfill \\ \vdots \hfill & \vdots \hfill & \vdots \hfill \\ {\Delta_{m} \left( 1 \right)} \hfill & {\Delta_{m} \left( 2 \right)} \hfill & {\Delta_{m} \left( 3 \right)} \hfill \\ \end{array} } \right]$$
(10)

where \(\varPhi\) is difference of Eqs. 8 and 9 respectively.

The values in these series are defuzzied values of the fuzzy number. The grey relation coefficient (\(\gamma\)) for input variable is calculated by using Eq. 11.

$$\begin{gathered} \gamma \;\left\{ {z_{0} \left( k \right),z_{i} \left( k \right)} \right\} \hfill \\ = {{\left\{ {\underbrace {{min}}_{i}\underbrace {{min}}_{k}\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right| + \zeta \underbrace {{max}}_{i}\underbrace {{max}}_{k}\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right|} \right\}} \mathord{\left/ {\vphantom {{\left\{ {\underbrace {{min}}_{i}\underbrace {{min}}_{k}\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right| + \zeta \underbrace {{max}}_{i}\underbrace {{max}}_{k}\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right|} \right\}} {\left\{ {\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right| + \zeta \underbrace {{max}}_{i}\underbrace {{max}}_{k}\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right|} \right\}}}} \right. \kern-\nulldelimiterspace} {\left\{ {\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right| + \zeta \underbrace {{max}}_{i}\underbrace {{max}}_{k}\left| {z_{0} \left( k \right) - z_{i} \left( k \right)} \right|} \right\}}} \hfill \\ \end{gathered}$$
(11)

The degree of grey relation (\(\Upgamma\)) for each risk is calculated by using Eq. 12.

$$\Gamma \left( {z_{0} ,\;z_{i} } \right) = \sum\limits_{{k = 1}}^{n} {\xi _{k} } \;\gamma \left\{ {z_{0} \left( k \right),z_{i} \left( k \right)} \right\}$$
(12)

where, \(\xi_{k}\) is weighting coefficient of for the input variables. It is decided by the expert’s judgement. In present research work, weighting coefficients are calculated using Wang scale (Table 2), and weights are calculated using fuzzy extent analysis (Panchal and Kumar 2017a, b; Srivastava et al. 2018).

Table 2 Fuzzy
Full size table

The selection and estimation of weights is important for legitimate results and ranking.

5 Case description and analysis

The process flow diagram is shown in Fig. 4.

Fig. 4
figure 4

Process flow chart

Full size image

The description of the components in Fig. 4 is tabulated in Table 3.

Table 3 Component description
Full size table

5.1 Analysis

The cause and effect diagram has been shown in Fig. 5. Due to limitation of space only eighteen (18) causes failure have been shown.

Fig. 5
figure 5

Fish-bone/cause and effect diagram

Full size image

After identification of failure causes, expert rating for risk for three input variables was recorded and subsequently RPN calculated according to the steps as discussed in methodology. For evaluation of fuzzy RPN, input and output variables were approximated by triangular and trapezoidal fuzzy numbers respectively (Fig. 6 and 7), and subsequently decision support system was prepared using Matlab R-13a (Eq. 4).

Fig. 6
figure 6

Membership function for Oc, Sr and Dn

Full size image
Fig. 7
figure 7

Membership function for FRPN

Full size image

A total of 125 if–then rules were prepared and reduced to eliminate chances of duplicity. The output of decision support system is fuzzy RPN. The output of FMEA and fuzzy FMEA has been shown in Table 4.

Table 4 RPN, FRPN
Full size table

The degree of relationship for fuzzy GRA has been evaluated using Eq. from 8–12. The standard, comparative and \(\varPhi\) series as evaluated by Eqs. 8, 9 and 10 respectively is as follow:

$${\text{Comparative}}\;{\text{Series }} = \left[ {\begin{array}{*{20}c} {5.015} & \ldots & {0.9983} \\ \cdots & \cdots & \cdots \\ {9.324} & \ldots & {9.324} \\ \end{array} } \right]$$
$${\text{Standard}}\;{\text{Series}} = \left[ {\begin{array}{*{20}c} 0 & \ldots & 0 \\ \cdots & \cdots & \cdots \\ 0 & \ldots & 0 \\ \end{array} } \right]$$
$$\varPhi = \left[ {\begin{array}{*{20}c} {5.015} & \ldots & {0.9983} \\ \cdots & \cdots & \cdots \\ {9.324} & \ldots & {9.324} \\ \end{array} } \right]$$

The grey relation coefficient (\(\gamma\)) and degree of relation (\(\Gamma\)) (Table 5) is as follow:

Table 5 Grey relation coefficient and degree of relationship
Full size table

The prioritization from all the approaches has been shown in Table 6.

Table 6 Combined ranking
Full size table

6 Result and discussion

Qualitative analysis for risk shows a crucial part to bridge up the space amongst the process and operational conditions by providing pre-emptive maintenance strategy and suggesting precaution. Currently, the aim is to reduce or prevent the risk conditions in CNG filling stations where the subcomponents are metering skid, compressor, dispenser, priority panel and cascade. The results are as follow:

  1. 1.

    The results elucidate average RPN values for metering skids by traditional FMEA is 27.2, by fuzzy is 3.975 and by GRA is 0.761. It is found that at least four and at most six failure causes have values more than the average value as evaluated by conventional FMEA, fuzzy FMEA and fuzzy GRA approach respectively.

  2. 2.

    The average RPN values for compressor by traditional FMEA is 147, by fuzzy is 4.479 and by GRA is 0.653466. It is found that at least three and at most eight failure causes have values more than the average value as evaluated by conventional FMEA, fuzzy FMEA and GRA approach respectively.

  3. 3.

    The average RPN values for dispenser by traditional FMEA is 67.1, by fuzzy is 5.3 and by GRA is 0.659. It is found that at least three and at most four failure causes have values more than the average value as evaluated by conventional FMEA, fuzzy FMEA and GRA approach respectively.

  4. 4.

    The average RPN values for priority panel by traditional FMEA is 113.714, by fuzzy is 4.107 and by GRA is 0.6711. It is found that at least one and at most five failure causes have values more than the average value as evaluated by conventional FMEA, fuzzy FMEA and GRA approach respectively.

  5. 5.

    The average RPN values for cascade by traditional FMEA is 95.25, by fuzzy is 5 and by GRA is 0.5988. It is found that It is found that at least one and at most four failure causes have values more than the average value as evaluated by conventional FMEA, fuzzy FMEA and GRA approach respectively.

The fuzzy method represents an extensive concern while bearing with the failures. The failure ranking comparison elucidates that traditional FMEA has uncertainty and ambiguity in the assessment while the fuzzy approach provides more stable solution. This is justified as follow:

  1. 1.

    It can be noticed from failure mode CS2 (Internal Pipe Leakage), DS3 (Failure of Motherboard), DS5 (Flow Sensor Failure), PP2 (Air Filter Choke), PP7 (Actuator Failure), CA3 (Tank Leakage/Tank Corrosion) and CA4 (Burst Disk Failure) in which the value of RPN are similar. Fuzzy FMEA along with fuzzy GRA approach can convincingly assess the expert’s judgements and accordingly improve the significance of FMEA technique. The rule base analysis is substituted by the fuzzy GRA. The evaluated weights for Oc, Sr and Dn and integrating it with the proposed methodology, makes it realistic, applied and flexible.

  2. 2.

    There is an issue of clustering because of assignment of multiple belief to single RPN, using rule base approach.

  3. 3.

    It has been noticed that MS7 in metering skid has a less importance in fuzzy GRA and a greater importance in rule-based method. Failure modes like CS5, CS4, DS6, DS7, PP4, CA2 and CA3 have greater importance in the fuzzy GRA method and lower importance in rule-based method.

  4. 4.

    By the use of projected methodology failure, it is being evaluated that the following failure needs safety improvement action in order to promote a safer surrounding:

    • MS10 Metering Skid’s Regulator Malfunction results in Pulsating Flow

    • CS2 Compressor’s internal pipe leakage results in loss of pressure and gas leakage

    • CS3 Compressor’s internal valve failure results in valve closing

    • DS3 Failure in Dispenser’s motherboard results in complete breakdown

    • PP1 Internal Pipe Leakage in Priority Panel results in Gas leakage

    • PP2 Priority Panel’s air filter choke results in particulate matter contamination

    • CA1 Cascade’s plug leakage results in low output

7 Conclusion

Fuzzy FMEA and MADM approach is applied to CNG dispensing unit. The possible failure modes have been diagnosed and analysed for their effects, detection methods and their root causes. The conventional and fuzzy RPN are being calculated with the input analysed and discussed from the experts. The use of etymological expressions permits the experts to provide a noteworthy worth for the risk factors under consideration. The fuzzy GRA approach allows the practitioner to fix on the significant factor for the study and provides a weight to the results that provide precious inferences for the prioritization.

The benefits of integrated approach are as follow:

  • Experts experience and knowledge are thoroughly combined for usage in the risk analysis.

  • The judgement of expert is conveyed easily in form of linguistic variable.

  • Fuzzy integrated approach can overcome uncertainties, thereby making analysis more realistic, flexible and practical.

  • Fuzzy GRA approach can alternatively be used without framing ‘if–then’ rules which otherwise requires expertise. It can be easily modified to meet system requirement.

  • Assigning weights to the variables in study, makes risk analysis more realistic.

  • The integrated approach obtains risk priorities in simple and straightforward manner.

  • The integrated approach in form of decision support system assist system analyst in decision making.

The outcomes are useful for maintenance engineer and system analyst for assessing risk and planning for effective maintenance policy. The present methodology can be applied to other processing system like Petrochemical Plant, Chemical Plant, Aerospace Industries, Fertilizer Plant, Power Plant, where there is use of inflammable fuel or working substance.