Production, Manufacturing and Logistics
A quantitative model for disruption mitigation in a supply chain

https://doi.org/10.1016/j.ejor.2016.08.035Get rights and content

Highlights

  • We develop mitigation planning approaches for a supply chain.

  • We develop a tool for predictive mitigation plan.

  • We develop a quantitative approach for reactive mitigation.

  • An efficient heuristic is developed to deal with sudden disruptions.

  • A good number of random experiment is performed.

Abstract

In this paper, a three-stage supply chain network, with multiple manufacturing plants, distribution centers and retailers, is considered. For this supply chain system we develop three different approaches, (i) an ideal plan for an infinite planning horizon and an updated plan if there are any changes in the data, (ii) a predictive mitigation planning approach for managing predictive demand changes, which can be predicted in advance by using an appropriate tool, and (iii) a reactive mitigation plan, on a real-time basis, for managing sudden production disruptions, which cannot be predicted in advance. In predictive mitigation planning, we develop a fuzzy inference system (FIS) tool to predict the changes in future demand over the base forecast and the supply chain plan is revised accordingly well in advance. In reactive mitigation planning, we formulate a quantitative model for revising production and distribution plans, over a finite future planning period, while minimizing the total supply chain cost. We also consider a series of sudden disruptions, where a new disruption may or may not affect the recovery plans of earlier disruptions and which consequently require plans to be revised after the occurrence of each disruption on a real-time basis. An efficient heuristic, capable of dealing with sudden production disruptions on a real-time basis, is developed. We compare the heuristic results with those obtained from the LINGO optimization software for a good number of randomly generated test problems. Also, some numerical examples are presented to explain both the usefulness and advantages of the proposed approaches.

Introduction

A supply chain is a network that receives inputs or raw materials from suppliers, produces final products at its manufacturing facilities and delivers those products to customers through a distribution network. Every manufacturing and service industry is part of a supply chain network which can have multiple manufacturing plants, multiple distribution centers (DCs) and multiple retailers. There are numerous industries, such as the pharmaceutical, textile and manufacturing, that supply, produce and distribute products using a supply chain network. Depending on the number of entities in each tier of a network, it can be very complex, and in a real-life one, any information can be changed at any time. Therefore, an ideal plan should be updated to incorporate changes in order to generate a better plan. Although some changes in data may be known well in advance, others may not, but can instead be detected using appropriate prediction tools. Such predictions will help to generate a better supply chain plan than the one designed for ideal conditions. In real-world supply chain system, the plan should be revised if there are any changes in data and/or if any future changes that can be predicted in advance. In this paper, we aim to develop both updated and predictive mitigation plans in a three-stage supply chain system to incorporate any known changes in data and predicted changes in demand respectively.

Supply chain entities can also face many sudden uncontrollable problems, which cannot be predicted in advance, such as a production disruption in a manufacturing plant, which can be defined as any form of interruption in the manufacturing system, including a material shortage, machine breakdown, or any other form of accidental or man-made disturbance (Paul, Sarker, & Essam, 2016a). Disruption management is an important research topic in supply chain, as is obvious in the following examples. A recent study conducted in 2015 by the Business Continuity Institute (Supply Chain Resilience Report, 2015) reports that, although the awareness of supply chain risks is increasing, many companies remain exposed to high levels of risk. It states that 74% of survey respondents from 426 organizations had experienced at least one disruption in their supply chain, with 6–20 disruptions per year for 50% of the companies, and the financial losses varied from 50 thousand to 500 million euros. More than 23% of the companies reported that the loss due to a disruption is at least one million euros. Supply chain disruptions not only cause financial loss but can also damage a company's brand or reputation as a result of third-party failures. It has been reported that 27% companies have suffered damage to their reputations, 58% lost productivity and 38% lost revenue. According to Sodhi and Chopra (2004), a disruption at the Royal Phillips Electronics plant in New Mexico on March 17, 2000 was caused by lightning strikes which led to a massive surge in the surrounding electrical grid, and later, a resultant fire damaged millions of microchips. Nokia Corporation and Ericsson were two major customers of the Phillips plant. To obtain a backup supply, immediately after this fire disaster, Nokia took proactive measures by redesigning its products and switching its chip orders to other Phillips plants. In contrast, as Ericsson employed a single sourcing policy and a slow disruption recovery plan, its production was disrupted for months, which caused 400 million US dollars in lost sales. From the above two examples, it is clear that supply, production and distribution systems can be unbalanced due to a disruption, and consequently, organizations can face enormous financial losses as well as loss of customer goodwill. Though disruption mitigation is an important research topic in supply chain, in the literature a very few papers developed quantitative reactive mitigation approaches in production-inventory system but no study extended the concept for multi-stage supply chain system. Therefore, it is essential to extend the concept of disruption recovery to develop an appropriate reactive mitigation model for a supply chain system for minimizing the effect of a sudden disruption. To fulfill this gap in the literature, this paper also aims to develop a new quantitative reactive mitigation model for managing both single and a series of sudden production disruptions in a three-stage supply chain system.

Production-inventory system is considered as a sub-set of a practical supply chain system. For this reason, most researchers concentrated their study in production-inventory system and some others extended their work to supply chain environments. At first, the researchers focused on developing models under ideal conditions, for example, based on distribution systems with a single product, single warehouse and multiple retailers (Petrovic, Xie, Burnham, & Petrovic, 2008), a single manufacturer and single retailer, with the demand and manufacturing cost fuzzy variables (Zhou, Zhao, & Tang, 2008), a single period and two-stage supply chain coordination problem (Xu & Zhai, 2010) and a three-stage system consisting of supplier, manufacturer and retailer producing a combination of perfect and/or defective items (Sana, 2011, Sana, 2012). Recently, Pal, Sana, and Chaudhuri (2012) developed an inventory model for multiple items production with multiple suppliers, one manufacturer and multiple retailers with deterministic demand. A few more studies on supply chain models under ideal conditions, can be found in Paul et al., 2014, Purnomo et al., 2012, Reza Nasiri et al., 2014, and Heydari (2014).

The above studies, and many others, used perfect supply chain environments. In real-life supply chain systems, there may be some known changes in data, such as changes in cost and capacity. Although some changes in data may be known well in advance, others may not, but rather need to be predicted using an appropriate tool. Most of the papers considered traditional forecasting techniques to predict supply change events (Syntetos, Babai, Boylan, Kolassa, & Nikolopoulos, 2016). But an appropriate prediction tool, which can be used to predict any future changes over the base forecast, can be helpful for developing an efficient quantitative predictive mitigation plan. In this paper, one of our objectives is to develop an appropriate prediction tool to predict the changes in demand and to develop a quantitative predictive mitigation plan based on the prediction.

Over the last decade, developing reactive mitigation plan for managing sudden disruption has become an important research topic in supply chain system. If a system is disrupted for a given period of time (known as a disruption period/duration), it is necessary to revise, after a disruption, the production schedule (known as a recovery plan) for some periods in the future (known as a recovery time window) until the system returns to its normal schedule (Hishamuddin, Sarker, & Essam, 2012). In modeling, the recovery time window can be either user specified or a variable that must be determined. In recent years, Fahimnia et al., 2015, Paul et al., 2016a, Snyder et al., 2016 provided an extensive review of supply chain risk and disruption management models. However, in this paper, we review mainly recent researches on sudden disruption recovery models in production-inventory and supply chain systems.

Firstly, we discuss the disruption recovery models in production-inventory systems. Xia, Yang, Golany, Gilbert, and Yu (2004) proposed a general disruption management approach for a two-stage production and inventory control system that incorporates a penalty cost for deviations of the new plan from the original. They introduced the concept of a disruption recovery time window that was considered in most recent models. Eisenstein (2005) introduced a flexible dynamic produce-up-to policy that is able to respond to a disruption by adjusting the amount of idle time during recovery and re-establishes the target idle time as a schedule is recovered. A production disruption recovery model, for a single disruption within a single-stage single-item production system, for obtaining a recovery plan within a user-defined time window was developed by Hishamuddin et al. (2012), which was basically an extension of the model of Xia et al. (2004). This study considered back orders and lost sales options as recovery strategies. This concept was further extended to develop a real-time disruption recovery model for managing both single and multiple disruptions in a single-stage production-inventory system (Paul, Sarker, & Essam, 2015a), a two-stage imperfect production-inventory system (Paul, Sarker, & Essam, 2014a), and a three-stage mixed production environment (Paul, Sarker, & Essam, 2015b). In the same direction of research, an approach for managing demand fluctuation on a real-time basis in a supplier-retailer coordinated system was developed by Paul, Sarker, and Essam (2014b). In the same research area, some other recovery models for deterministic demand in a production-inventory system can be found in Gallego, 1994, Paul et al., 2013, Qi et al., 2004, Tang and Lee, 2005, Yang et al., 2005 and Paul, Sarker, and Essam (2014c).

In the supply chain context, the delivery disruptions have been studied by a few researchers. Giunipero and Eltantawy (2004) discussed a transportation disruption in general in their study, but did not specify strategies for facing it. Wilson (2007) investigated the effects of a transportation disruption on the performance of a supply chain by using a system dynamics simulation in a 5-echelon supply chain system. Unnikrishnan and Figliozzi (2011) formulated a mathematical model for a new type of freight network assignment problem in a dynamic environment in the presence of probable network disruptions or significant delays. Recently, Hishamuddin, Sarker, and Essam (2013) developed a transportation disruption recovery model in a two-stage, single supplier and single retailer, supply chain system. Recently, the recovery modeling concept was further extended for managing a supply disruption in a two-stage supply chain consisting of a single supplier and single retailer (Hishamuddin, Sarker, & Essam, 2014) and for managing both single and multiple sudden supply disruptions in a three-stage supply chain with multiple suppliers and retailers (Paul, Sarker, & Essam, 2016b). Over the last few years, some other supply disruption management approaches have been studied by Chopra et al., 2007, Craighead et al., 2007, Li et al., 2004, Mohebbi and Hao, 2008, Qi et al., 2009, Ross et al., 2008, Tomlin, 2006, Wu et al., 2007, and Hult, Craighead, and Ketchen (2010).

There are several gaps in the literature. It is clear that most researchers focused on supply chain coordination and optimization problems under ideal conditions, although a number of studies developed recovery and reactive mitigation models after the occurrence of a sudden disruption. However, no study has been found that predicts the possible changes in future demand that is used as input to the mitigation planning model in supply chain environment. Moreover, most past studies focused mainly on a single disruption in production and a very few focused on a series of disruptions on a real-time basis, but they are again for a single supplier and single retailer, which limits their applicability in real-life situations (Paul et al., 2016a). Interestingly, no study has been found which developed a quantitative recovery for managing sudden production disruption in a supply chain with multiple entities in each stage of the system.

In this paper, firstly, we have formulated a model to generate an ideal supply chain plan. If any variation in data in any period is observed, this plan will be updated for a finite period on a rolling horizon basis with the new data. In real-life situations, some changes may not be known in advance, but can be predicted using an appropriate prediction tool. Therefore, in this paper, a predictive mitigation planning approach is developed, with the predicted data used to generate a revised plan in advance on a rolling planning horizon basis. A fuzzy inference system (FIS) based tool is used to predict future changes in demand, and then such predicted demand data are used to develop a revised plan in advance, which is said to be a predictive mitigation plan. Finally, we have developed a disruption recovery (reactive mitigation) model for a supply chain network consisting of multiple manufacturing plants, DCs and retailers. We consider disruptions due to technical and internal problems, which take place more frequently (repetitive type) and are for short durations. In real-life, a system can face a series of production disruptions (known as multiple disruptions), one after another, at any plant. If a new disruption occurs at any plan during the revised planning window of a previous production disruption, known as a dependent disruption, the production and distribution plan must be revised again while considering the effects of both disruptions. Therefore, this can be a continuous process that must be dealt with on a real-time basis. A real-time disruption management scenario in a three-stage supply chain network, where the disruptions are not known a priori, is considered in this study. This means that the current plan is revised immediately after a disruption occurs, as this disruption is impossible to predict. For experimentation, as we assume that any disruption event is random, we generate disruption scenarios using a uniformly random distribution (Paul et al., 2015b) to determine characteristics such as the disruptions’ start times and durations. However, it is possible to generate disruption scenarios by using other probability distributions. To achieve this objective, we develop a new mathematical and heuristic approach for obtaining a recovery plan after the occurrence of a single disruption or a series of disruptions on a real-time basis. The results for both the predictive and reactive mitigation approaches are discussed and some numerical examples are presented to demonstrate their usefulness.

The main contributions of this paper can be summarized as follows.

  • i.

    Developing an updated supply chain plan for a finite period on a rolling horizon basis to incorporate any changes in data.

  • ii.

    Developing a predictive mitigation planning approach for obtaining a better supply chain plan. A fuzzy inference system based tool is designed to predict any changes in base forecasted demand, based on information about fluctuation, unexpected incident and natural incident, and the supply chain plan is accordingly revised well in advance.

  • iii.

    Developing a mathematical model for managing sudden production disruptions which cannot be predicted in advance. The supply chain plan is revised, after the occurrence of a disruption, for a finite period into the future, on a real-time basis, to minimize loss due to the disruption.

  • iv.

    Developing a new heuristic for generating recovery plan for the sudden production disruption problem in iii. The heuristic results are compared with those from another established solution technique for a good number of randomly generated test problems.

  • v.

    Extending the heuristic to deal with multiple disruptions, one after another as a series, on a real-time basis. This heuristic is capable of determining a recovery plan, after the occurrence of each disruption, for as long as disruptions take place in the system.

The paper is organized as follows. In Section 2, we describe the problem, along with the notations and assumptions used in this study. In Sections 3 and 4, respectively, we present the model formulation and solution approach. In Section 5, we discuss the experimentation and analysis of results. Applicability of the model and value in practice are presented in Section 6. Finally, we provide a summary of, and future guidelines for, our research in Section 7.

Section snippets

Problem description

Firstly, this paper develops a supply chain model under ideal conditions for an infinite planning horizon. Basically, this ideal plan is used to determine the cycle length, which we consider the length of a period required for planning and analysis. A three-tier supply chain network with multiple entities in each tier (such as manufacturing plants, DCs and retailers) is considered, as presented in Fig. 1. In the ideal system, products are produced in the manufacturing plants and then are moved

Model formulation

In this section, we formulate the mathematical model for both an ideal and a disrupted supply chain system. The ideal plan is updated if there are any changes in the data and is also revised according to any prediction of future changes for a finite planning period. In the case of managing a disruption, the model is re-formulated to incorporate the effect of a disruption and the production and distribution plan is revised for a finite planning period. After the recovery window, the production

Solution approaches

In this section, solution approaches for both ideal and disrupted systems are developed. A standard solution technique for solving the ideal supply chain system is proposed and applied to obtain updated and predictive mitigation plans for changes in the data and future predictions respectively. An efficient heuristic for managing a single disruption in the system is developed and then extended to be implemented for managing multiple disruptions on a real-time basis.

Experimentation and analysis of results

In this section, we discuss the experiments and results for both the ideal and disrupted systems, and the updated and predictive mitigation plans for a good number of randomly generated test problems. For the disrupted system, the results for both a single disruption and multiple disruptions are analyzed. The test problems are solved using both the heuristic, and the branch and bound algorithm of the LINGO optimization software. To judge the quality of the heuristic solutions, we also compare

Applicability and value in practice

This paper developed three different approaches for supply chain mitigation:

  • i.

    Updated plan for any change in data – the plan was revised if there were any changes in cost and demand data.

  • ii.

    Predictive mitigation plan – the changes in demand over the base forecast were predicted in advance by using a rule and logic based FIS prediction tool that is based on information of demand fluctuation, unexpected incidents and natural incidents. The supply chain plan was revised in advance according to that

Conclusions

The main objective of the paper was to develop quantitative approaches for managing any changes in data and for generating both a predictive and reactive mitigation plan. In the case of any changes in data that were known in advance, the supply chain plan was updated according to the changes. For a predictive mitigation plan, a FIS based prediction tool was developed to predict changes in future demand, which are hence not known but predicted and the supply chain plan was revised to obtain a

Acknowledgment

The authors express their sincere gratitude and thanks to the reviewers for their positive, constructive and helpful comments and suggestions that helped to improve the manuscript.

References (50)

  • S.K. Paul et al.

    Managing disruption in an imperfect production-inventory system

    Computers & Industrial Engineering

    (2015)
  • S.K. Paul et al.

    A disruption recovery plan in a three-stage production-inventory system

    Computers & Operations Research

    (2015)
  • D. Petrovic et al.

    Coordinated control of distribution supply chains in the presence of fuzzy customer demand

    European Journal of Operational Research

    (2008)
  • H.D. Purnomo et al.

    Two inventory review policies on supply chain configuration problem

    Computers & Industrial Engineering

    (2012)
  • QiX. et al.

    Supply chain coordination with demand disruptions

    Omega

    (2004)
  • A.M. Ross et al.

    Supply disruptions with time-dependent parameters

    Computers & Operations Research

    (2008)
  • S.S. Sana

    A production-inventory model of imperfect quality products in a three-layer supply chain

    Decision Support Systems

    (2011)
  • S.S. Sana

    A collaborating inventory model in a supply chain

    Economic Modelling

    (2012)
  • A.A. Syntetos et al.

    Supply chain forecasting: Theory, practice, their gap and the future

    European Journal of Operational Research

    (2016)
  • TangL.C. et al.

    A simple recovery strategy for economic lot scheduling problem: A two-product case

    International Journal of Production Economics

    (2005)
  • M.C. Wilson

    The impact of transportation disruptions on supply chain performance

    Transportation Research Part E: Logistics & Transportation Review

    (2007)
  • XuR. et al.

    Manufacturer's coordination mechanism for single-period supply chain problems with fuzzy demand

    Mathematical & Computer Modelling

    (2010)
  • ZhouC. et al.

    Two-echelon supply chain games in a fuzzy environment

    Computers & Industrial Engineering

    (2008)
  • I. Ahmed et al.

    Employee performance evaluation: A fuzzy approach

    International Journal of Productivity and Performance Management

    (2013)
  • A. Cavallo et al.

    Prices and supply disruptions during natural disasters

    Review of Income and Wealth

    (2014)
  • Cited by (82)

    • Future of artificial intelligence and its influence on supply chain risk management – A systematic review

      2022, Computers and Industrial Engineering
      Citation Excerpt :

      FIS has a significant role in reactive mitigation planning models to minimize cost. It allows an organization to update its plan for any changes in data, future prediction of demand changes due to sudden production disruptions (Paul et al., 2017). In addition, the case-based reasoning approach addressed manufacturing-related issues like quality control (Camarillo et al., 2018).

    • Time-preference-based on-spot bundled cloud-service provisioning

      2021, Decision Support Systems
      Citation Excerpt :

      Thus, for example, the Availability experiment has eight (= 4*2) different parameter settings and 80 runs. We compare Algo's performance with that of solving TPOP using LINGO, a commercial production-grade solver (64-bit Lingo 18.0 and 8 GB RAM with 3.2 GHz Intel Core i5 CPU) that has been used in a number of past research works [63–65]. We consider two metrics:

    View all citing articles on Scopus
    View full text