MANAGING SUPPLY CHAIN RISKS: A FUZZY-FAILURE MODE AND EVALUATION APPROACH FOR RANKING THREATS

On the backdrop of lower transportation cost, outsourcing paved the way for borderless production activities and ushered in the era of Supply Chain Management (SCM). For many organizations, achieving the goals of their Supply Chain (SC) is constantly threatened by increased competition and disruption. In this study, the aim is to identify, and rank, SC threats in a developing country using Failure Mode and Effects Analysis (FMEA) with Fuzzy Logic (FL). FMEA parameters were derived for 44 supply chain threats (SCT1 – SCT44) and their Risk Priority Number (RPN) determined. Subsequently, the Mamdani Fuzzy Inference system was utilized to arrive at a Fuzzy-RPN with 125 rules using severity as a determining factor. The rules were ranked to prioritize SC threats. From the conventional FMEA, demand variation (SCT42) and long-distance sourcing (SCT27) had the highest and lowest RPN, respectively. After fuzzification and defuzzification, Fuzzy-RPN identified raw material delay (SCT1), government policy (SCT11), poor transport infrastructure (SCT18) and political instability (SCT19) as threats with the highest Fuzzy-RPN (210) and product recalls (SCT28) with the lowest Fuzzy-RPN (99). Based on these results, it is concluded that a Fuzzy-FMEA approach can identify and rank SC threats with the use of an RPN devoid of sentiments and inaccuracies.


INTRODUCTION
Before the 1990s, the approach of many organizations to customer fulfilment was vertical integration with little or no emphasis on core competencies. On the backdrop of lower transportation cost, outsourcing paved the way for borderless production activities and ushered in the era of Supply Chain Management (SCM). SCM coordinates decisions on production, inventory, location and transportation among the firms in a Supply Chain (SC) to reduce operating and inventory costs [1].
SC as a network of relationships aligns firms and the required chain drivers to deliver products and services [2][3][4]. From a system view, SC drivers include production facilities, material suppliers, distribution and retailing services, location, transporters, and customers connected to exchange materials and share information [1]. While one of the goals of an SC is to ensure virtual integration among participating companies, a robust SC system should possess the ability to maintain a balance between responsiveness and efficiency. Such a system can increase the efficiency of different business process through integration and proper coordination. On responsiveness, an excellent decision-making structure will enable an organisation respond to disruptions from increased competition, globalisation, global pandemic, and technological advancement. Invariably, failure events characterized by the termination of the ability to exchange resources and information in the chain will occur if disruptions and risks are poorly managed.
Supply chain risks are events (or conditions) at macro and micro levels with the potential to negatively influence core drivers in the chain leading to failures [5]. Micro-risk considers recurrent events which have their root cause from components within the network. Macro-risk includes external and man-made events. Harington [6] commented that 90% of firms do not formally measure risk and about 47% do not possess any backup plan in the event of unexpected disruptions. Based on the multiplicity and diverse risky events among chain drivers, Supply Chain Risk Management (SCRM) should be a collaboration effort. SCRM should consider environmental, social, and economic risks at macro and micro levels, and propose a contingency approach to increase SC resilience. The resilience of an SC describes the internal and external capability of the system elements to manage inevitable disruptions and revert to pre-disruption status. This ability will mitigate (or ameliorate) against the ripple effect of disruptions; for example, Peck [7] sighted SC disruptions of two Finnish phone makers Nokia and Ericsson in 2000. The disruption caused Ericsson about $400 million loss in new product sales. In addition, UPF-Thompson, a major supply of chassis to Land Rover became insolvent in 2001. These events and others opened a new era in SCRM research as interests in creating a resilient supply chain increased.
Singhal et al. [8] reviewed research articles on SCRM and concluded that a large proportion of the methods appeared disjointed despite growing diverse techniques and applications. Similarly, lack of understanding of SCRM and an all-inclusive SCRM technique are barriers to effectively manage SC risks [5,9]. Tang and Musa [10] commented that quantitative models in SCRM are lacking despite a significant rise in intellectual understanding of SCRM. Ghadge et al. [11] used Failure Mode and Effects Analysis (FMEA) to identify and access SC risks. Cankabis et al. [12] utilized FMEA with Analytic Hierarchy Process (AHP) to model risk assessment for SC system. Their study indicated that a prioritized sub-risks in SC can be obtained using FMEA despite its limitations. However, the limitations of the conventional FMEA [13][14][15] can be enhanced when integrated with other techniques e.g. Analytic Hierarchy Process (AHP), Fuzzy Logic (FL), Value Analysis, etc. [16]. In this study, the aim is to develop a quantitative knowledge repository to rank, prioritize and manage supply chain threats in a developing country using Failure Mode and Effects Analysis (FMEA) with Fuzzy Logic (FL).

Supply chain risk management
The quest to identify, manage and minimize the vulnerability of the elements of a supply chain via a structured approach is the goal of SCRM [8]. As a background for this research work, previous methods used in SCRM are presented in Table 1.

Failure mode and evaluation analysis and Fuzzy logic
FMEA is a subjective human thinking used to identify and analyse failure modes and their causes in components and systems [24,25]. As a bottom-up framework, each potential failure mode in FMEA; severity (S), occurrence (O) and detectability (D) are identified and assigned values between 1 and 10. Thereafter, a risk priority number (RPN) between 1 and 1000 will be obtained for each failure mode.
FL possess a high tolerance for inaccurate data and allows the modelling of intricate non-linear situations within little time [26]. An FL system has the following components: fuzzifier rules, inference engine, and defuzzifier. The FL process combines a set of crisp input, transformed into a fuzzy set through fuzzy linguistic variables, fuzzy linguistic terms and membership functions. Fuzzification allows an inference to be made based on set of rules.
Using membership functions, the result can be presented as a crisp output in what is known as defuzzification [27,28].

Fuzzy-FMEA
In Fuzzy-FMEA, the failure modes in FMEA are fuzzified with their respective membership function to obtain a degree of importance for each parameter. The technique of Fuzzy-FMEA as described by Balaraju et al. [28] is described in Figure 1.

Primary and secondary data collection
Primary data obtained through interviews and questionnaire distributed among 7 organisations are presented in Table 2. From Rwakira [29], 44 SC threats were adopted to capture relevant aspects threats within the context of developing countries. The threats with their representative codes are shown in Table 3.  Figure 2. In Table 4, the linguistic variables, term sets and membership functions required to convert FMEA into equivalent FIS are highlighted. Membership functions map non-fuzzy inputs to fuzzy linguistic variable and vice-versa in the Fuzzification and Defuzzification process. In essence membership functions are used to quantify linguistic terms. A total of 125 rules combination were obtained from the membership functions.  In Figure 3, a sample screenshot of the developed FIS based fuzzy rules is presented. Using the Fuzzy Logic toolbox in MATLAB software, Fuzzy RPN (FRPN) value was obtained for each threat.

CONCLUSIONS
In this study, the integration of FMEA with Fuzzy Logic was utilized to rank supply chain threats. A Mamdani-FIS was developed with 3 inputs of severity, occurrence, and detectability. The output was a FuzzyRPN with a total of 125 rules. These rules were formed with severity as the most important factor to adequately rank the threats.
With the use of fuzzy logic, subjectivity, vagueness, and incompleteness associated with the supply chain threats