Methodologies for performance enhancement in decentralized supply chains

137 7.1 Integrated Decentralized Distribution Network and Collection Centers164 7.2 Schematic representation of Overall Supply Chain Network.. ai,p,bi,p,ci,p replenishment policy paramet

METHODOLOGIES FOR PERFORMANCE ENHANCEMENT IN

DECENTRALIZED SUPPLY CHAINS

SUNDAR RAJ THANGAVELU

NATIONAL UNIVERSITY OF SINGAPORE

2009

METHODOLOGIES FOR PERFORMANCE ENHANCEMENT IN

DECENTRALIZED SUPPLY CHAINS

SUNDAR RAJ THANGAVELU(M.Tech., I.I.T Kharagpur, India)(B.Tech., University of Madras, India)

A THESIS SUBMITTEDFOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF CHEMICAL AND BIOMOLECULAR ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

2009

Dedicated to My Dear Parents

My deepest thanks to my parents and family members for their continuous support,interest and encouragement They would be my all time assets to pursue mywishes

Next to family, I would like to thank my supervisor Dr Lakshminarayanan ham (Laksh) for his kind guidance, support and timely feedback His suggestionshelped me a lot to advance further in my research I admire his enthusiasm inmotivating others to excel in diverse research areas I personally benefited byhaving Dr Laksh as my supervisor in aspects such as of improving communicationskills, coordinating team work with my colleagues and interactions with visitorsand final year students His encouragement allowed me to spare time for depart-mental activities (teaching activities and tutoring) and ChBE Graduate StudentAssociation Thanks Sir for all the advice and help

Samaved-I was very fortunate to have Prof Karimi and Prof Karthik Natarajan as thepanel members for examining my research proposal and providing their valuablesuggestions to improve my research focus

I am grateful to Prof Fraser for facilitating and funding my visit to Department

of Chemical and Materials Engineering, University of Alberta, Canada duringSeptember 2008 to November 2008 In spite of his busy schedule, the way hemanaged his time and organized himself for discussions with me is tremendous

During my tenure as a PhD student, I got a chance to assist Prof Rangaiah, ProfM.P Srinivasan, Prof Raj Srinivasan and Prof Krishnaswamy in running courses

They shared their views on different topics in a friendly manner; my sincere thanks

to all of them

Nothing comes for free; especially computational facility which is very much nificant requirement challenge for my research Mr Mao Ning and Mr Boey(ChBE, NUS) and Bob Barton (UofA) ensured the required computational re-sources throughout my candidature Their timely contribution and the help fromthe Chemical and Biomolecular Department are much appreciated

sig-My stay at National University of Singapore has been a memorable one because

of my department friends, labmates and roommates They made my stay livelyand enjoyable I am thankful to all of them and to my present and past IPCUcolleagues for their excellent assistance and discussions

Finally, a big thanks to National University of Singapore and University of berta for giving me the opportunity to pursue my research with advanced researchfacilities and financial support

Al-Table of Contents

1.1 Supply Chain System - An Overview 2

1.2 Supply Chain Decision Levels 3

1.3 Supply Chain Structure 5

1.4 Supply Chain Cost 6

1.5 Centralized and Decentralized Supply Chains 7

1.6 Importance of Decentralized Distribution Networks 8

1.7 Research Scope 10

1.8 Thesis Overview 11

2 Distribution Network and its Management 14 2.1 Inventory Management Methods 15

2.1.1 Push and Pull Inventory Systems 15

2.1.2 Just-in-Time and Vendor Managed Inventory 16

2.2 Bullwhip Effect 18

2.2.1 Sources of Bullwhip Effect 19

2.2.2 Consequence of Bullwhip 21

2.2.3 Bullwhip Quantification and Impact 21

2.3 Distribution System 22

2.3.1 Modeling & Control of the Distribution Node 24

2.3.2 Distribution Network 25

2.4 Performance Metrics and their Quantification 28

2.4.1 Performance Metrics 28

2.4.2 Performance Benchmarking 31

2.5 Product nature and supply chain type 32

Table of Contents

2.6 Demand and Responsiveness 34

2.6.1 Market Demand 34

2.6.2 Relation between demand type and inventory requirement 34 2.6.3 Influence of shift in demand patterns 35

2.6.4 Demand Forecasting 36

2.6.5 Responsiveness 38

2.7 Supply Chain Diagnosis 38

2.8 Optimization Methodology 40

3 Decentralized Distribution Systems 45 3.1 Material and Information Balances 46

3.2 Market Demand 51

3.3 Replenishment Strategies 51

3.4 Performance Indicators 54

4 Performance Assessment Framework for Decentralized Distribu-tion Network Systems 57 4.1 Background 59

4.2 Motivation of this study 60

4.3 Objectives 61

4.4 Proposed Methodology 61

4.5 Problem Description 65

4.5.1 Market Demand 68

4.5.2 Performance Indicators 68

4.5.3 Performance index of the Distribution node and network 69

4.6 Solution Approach 70

4.6.1 Identification of inefficient distribution nodes 71

4.6.2 Identification of the potential opportunities for performance improvements 73

4.6.3 Lead time information of all nodes in the network 73

4.6.4 Responsiveness 75

4.7 Results and Analysis 77

4.7.1 Case Study (1): Stationary Demand 77

4.7.2 Case Study (2): Non-Stationary Demand 83

4.8 Conclusions 89

5 Multi-Objective Optimization in Multi-Echelon Decentralized Dis-tribution Networks 90 5.1 Background 92

5.2 Motivation of this study 93

5.3 Objectives 93

5.4 Multi-objective Optimization and Pareto Analysis 94

5.5 Problem Description 98

5.6 Proposed Methodology 99

5.7 Results and Analysis 104

Table of Contents

5.7.1 Customer Focused Approach 107

5.7.2 Cost Effective Approach 108

5.7.3 Optimal Cost Tradeoff Approach 110

5.7.4 Performance tradeoff Strategy (with reference to Utopia Per-formance) 112

5.8 Conclusions 116

6 Entropy Based Optimization of Decentralized Supply Chain Net-works 118 6.1 Background 120

6.1.1 Complexity and Consequences - Overview 121

6.1.2 Uncertainty Sources and Quantification 124

6.2 Motivation 126

6.3 Objectives 127

6.4 Problem Description 127

6.5 Complexity Modelling 130

6.6 Proposed Complexity Management Methodologies 136

6.6.1 Strategy I (S-I) 138

6.6.2 Strategy II (S-II) 139

6.6.3 Strategy III (S-III) 140

6.6.4 Strategy IV (S-IV) 140

6.7 Results and Discussions 142

6.7.1 Scenario 1: Complexity Reduction 143

6.7.2 Scenario 2: Complexity Optimization with desired CS 145

6.7.3 Scenario 3: Complexity Reduction (using similar replenish-ment rules) 147

6.7.4 Scenario 4: Complexity Minimization with desired CS (using similar replenishment rules) 148

6.8 Conclusions 151

7 Divide and Conquer Optimization for Closed Loop Supply Chains152 7.1 Background 154

7.1.1 Closed Loop Supply Chains 156

7.1.2 Decomposition based Optimization 159

7.2 Motivation: 161

7.3 Model Assumptions & Supply Chain Description 162

7.3.1 Assumptions 162

7.3.2 Problem Description 163

7.4 Proposed Methodology 167

7.4.1 Forward Channel 167

7.4.2 Reverse Channel 169

7.4.3 Production Facility 170

7.4.4 Optimization Agent 171

7.5 Results and Discussions 172

7.5.1 Comparison with Single/Multi Objective Optimization 173

Table of Contents

7.5.2 Robustness of the derived solution (obtained by three

op-timization methods) with respect to uncertain demand and

used product returns 176

7.5.3 Desirability of the solution with respect to the customer sat-isfaction constraint 178

7.6 Conclusions 180

8 Conclusions and Recommendations 181 8.1 Conclusions 181

8.2 Recommendations for further work 186

8.2.1 Diagnosis of Oscillations in Supply Chains 186

8.2.2 Demand (opportunity) Forecast 187

8.2.3 Sensitivity and Robust of Supply Chain Decisions 187

8.2.4 Optimization Efficiency 187

8.2.5 Advanced Control Strategies 188

A Reverse and Production system Model 200 A.1 Reverse Channel Formulation 200

A.2 Multi-purpose Production facility Formulation 202

B Publications and Presentations 206

List of Figures

1.1 Schematic representation of Supply Chain Architecture 3

1.2 Schematic representation of Supply Chain Decisions 4

1.3 Supply Chain Cost Structure 7

2.1 The Bullwhip Effect 19

2.2 Performance Metrics 29

3.1 Schematic representation of the decentralized distribution system 46 3.2 Schematic representation of the internal strategy of a distribution node 47

4.1 Performance Assessment Framework 63

4.2 Schematic representation of the Decentralized Distribution System 66 4.3 Performance of the Existing Distribution Strategy and after Damp-ening Aggressive Nodes 79

4.4 Performance Enrichment after Improving Weak Nodes and after Restructuring 80

4.5 Performance of similar replenishment policy at all nodes in the net-work 81

4.6 Total Cost of Decentralized Network facing Stationary Stochastic Demand 82

4.7 Excess Inventory and Backorder of heuristics rules (PI and SOP2) 83 4.8 Performance of the Existing Distribution Strategy and after Damp-ening Aggressive Nodes 85

4.9 Performance Enrichment after Improving Weak Nodes and after Restructuring 86

4.10 Performance of similar replenishment policy at all nodes in the net-work 87

4.11 Total Cost of Decentralized Network facing Non-Stationary Stochas-tic Demand 88

4.12 Excess Inventory and Backorder of Heuristic rules (PI and SOP2) 88 5.1 Schematic representation of the decentralized distribution system 92 5.2 Multi-echelon Decentralized distribution system 98

5.3 Multi-objective Performance Enhancement Framework 100

5.4 A Pareto analysis representation for various business strategies (sym-bol Θ - utopia point) 106

List of Figures

5.5 Pareto frontiers for the distribution network under various replen-ishment scenarios EI (empty circles) and BO (filled circles) are

plotted on the y-axis 109

5.6 Cost Pareto fronts for the distribution network under various re-plenishment scenarios BO is shown on the x-axis and EI is shown on the y-axis 111

5.7 Optimal performance tradeoffs of the distribution network under various replenishment scenarios ECS EI is shown on the x-axis while ECS BO is shown on the y-axis 115

6.1 Decentralized Distribution Network 128

6.2 Schematic representation of Distribution Node 131

6.3 Entropy measure for the desired state (pi) 135

6.4 Complexity Management Framework 137

7.1 Integrated Decentralized Distribution Network and Collection Centers164 7.2 Schematic representation of Overall Supply Chain Network 166

7.3 Closed Loop Supply Chain (Divide and Conquer) Optimization Framework 168

7.4 Non-dominated solution obtained by NSGA-II 175

7.5 Influence of uncertain inputs (demand/used product return) on sup-ply chain cost 176

7.6 Sensitivity of supply chain cost for -15% to +15% uncertainties in market demand and used-product returns (a) Decomposition Method (b) Conventional Method 177

7.7 Influence of uncertain inputs (demand/used product return) on cus-tomer satisfaction 178

7.8 Desirability of Customer Satisfaction constraint for -15% to +15% uncertainties in market demand and used-product returns (a) De-composition Method (b) Conventional Method 179

A.1 Scheduling sequence of Multi-purpose production facility 204

List of Tables

4.1 Internal Strategies of the Distribution Nodes 67

4.2 Current Performance Measure of the Distribution Network facing Stationary demand 72

4.3 Current Performance Measure of the Distribution Network facing Non-Stationary demand 72

4.4 Derived Lead Time information from the time-series data 74

4.5 Performance Improvement by dampening the Aggressive Nodes (stage 1) 77

4.6 Performance Enhancement by optimizing the Weak Nodes (stage 2) 77 4.7 Performance Enrichment by retrofitting the Conflicting Nodes (stage 3) 78

4.8 Performance Improvement and Achievable Performance Benchmark from Heuristics rules: Stationary Demand 82

4.9 Performance Improvement and Achievable Performance Benchmark from Heuristic rules: Non-Stationary Demand 84

5.1 Internal Strategies of the Distribution Nodes 99

5.2 Multi-Objective Optimization Parameters 103

5.3 Customer Focused Approach - Maximum Output 108

5.4 Cost Effective Approach 110

5.5 Optimal Cost Trade-off approach (BO and EI) 113

5.6 Optimal Performance Tradeoff (with reference to utopia point) 114

6.1 Internal Strategies of the Distribution Nodes 129

6.2 Existing (Base Case) Complexity 138

6.3 Scenario 1: Complexity Optimization 145

6.4 Scenario 2: Complexity Optimization with desired CS 147

6.5 Scenario 3: Complexity optimization using similar replenishment rules 149

6.6 Scenario 4: Complexity optimization using similar replenishment rules and with CS constraint 150

7.1 Summary of Decomposition Approaches 172

7.2 Results of Single and Multiobjective Decomposition Approaches 174

A.1 Associated cost in Closed Loop Supply Chains 205

Acronym What it Stands For

3PLs Third-party Logistics Providers

AHP Analytic Hierarchy Process

APIOBPCS Automatic Pipeline, Inventory and Order

Based Production Control SystemARIMA Auto-Regressive Integrated Moving Average

LSSVM Least Square Support Vector Machine

MILP Mixed Integer Linear Programming

MINLP Mixed Integer Nonlinear Programming

MIP Mixed Integer Programming

MOO Multi-Objective Optimization

MOSA Multi-Objective Simulated Annealing

MPC Model Predictive Control

MVC Minimum Variance Control

NSGA-II Non-dominated Sorting Genetic Algorithm-II

SCI Supply Chain Intelligence

SCN Supply Chain Networks

SOP Smoothing ordering policy

SPCA Spectral Principal Components Analysis

SPSA Simultaneous Perturbation Stochastic ApproximationVMI Vendor-Managed inventory

ai,p,bi,p,ci,p replenishment policy parameters

BOi,p Average Backorder

dc,p(t) Demand for product ‘p’ at time ‘t’ from customer ‘c’

ei,p(t) Inventory Position discrepancy for product ‘p’ at node ‘i’ at time ‘t’

EIi,p Average Excess Inventory

IHi,p(t) Inventory at-hand of product ‘p’ at node ‘i’ at time ‘t’

IHUi,p(t) Inventory at-hand of used-product ‘p’ at node ‘i’ at time ‘t’

IPi,p(t) Inventory Position of product ‘p’ at node ‘i’ at time ‘t’

IRi,p(t) Inventory on-road of product ‘p’ at node ‘i’ at time ‘t’

Li,p Lead time faced by the node ‘i’ for the product ‘p’

mi,p(t) percentage of order satisfied by the node ‘i’ at time ‘t’

pi probability of desired state at node ‘i’

SIPi,p(t) Inventory Position set point for product ‘p’ at node ‘i’ at time ‘t’SIHi,p(t) Inventory at-hand set point for product ‘p’ at node ‘i’ at time ‘t’SIRi,p(t) Inventory on-road set point for product ‘p’ at node ‘i’ at time ‘t’

Uji,p(t) Demand faced by node ‘i’ from downstream node ‘j’ at time ‘t’

Yij,p(t) Material ‘p’ delivered from node ‘i’ to downstream node ‘j’

YUig,p(t) Used-product ‘p’ pushed from node ‘i’ to upstream ‘g’

ψi,p Performance index of node ‘i’ for product ‘p’

ψN Overall performance index

φi,p weight parameter for product ‘p’ at node ‘i’

Today’s aggressive business environment and profound competition necessitatebusinesses to develop strong and robust operating strategies and continually re-vise/improve them in order to remain competitive In business logistics, companieswith slack coordination cannot compete or cope up with a rival system that is well-coordinated A well-organized coordination between the business entities enablesthe business chain to tackle competitors by providing a high level of service at lowproduct cost Inappropriate management strategies, improper integration amongthe supply chain entities and/or poor coordination with external participants de-teriorate supply chain performance leading to poor sales and reputation amongmarket customers Identifying the opportunities to enrich performance and to ex-pand the market share will pave way for future business evolution and will be thefocus of this thesis

Supply chain management emerged decades ago to address issues related to diversebusiness objectives such as efficient resource utilization, high customer satisfactionand to reduce supply chain cost In product distribution part, operations researchtechniques have enabled the replenishment of right amounts of products from theright suppliers at minimal (product and transportation) cost Advances in mod-eling have encouraged the practitioners to model existing systems and study theirbehavior under different scenarios and determine ways for improvement Processcontrol concepts are also being applied to supply chain systems with a view toachieve a required level of performance The tradeoff between performance metrics

Summary(customer satisfaction and cost) is an attractive concept that brings optimizationtools into play Flexibility in transportation further encourages the distributionsystem to maintain lesser inventory because frequent product replenishment atshort intervals is possible Managing fewer inventories reduces the inventory cost

at the cost of more backorder and unsatisfied customers To get rid of this trouble,supply chain managers need to focus on achieving improved coordination betweenthe supply chain entities

In this thesis, we play the role of a third-party supply chain consultant to developmethodologies to measure and enhance the performance of supply chain with refer-ence to diverse business goals The recommendations which are restricted to tacti-cal decisions may either lead to changes in control parameters, control structures orre-design of the internal strategy and the network configuration A multi-productmulti-echelon large scale decentralized supply chain is considered to investigatethe applicability of the proposed performance enhancement methodologies andsupport the real world supply chains with right decisions Several scenarios areused to illustrate the proposed performance enhancement methodologies

Chapter 1

Introduction

Tough competition and globalization are the two drivers for supply chain agement Supply chain management deals with information, material and cashflows and the relationships among channel intermediaries from the point of origin

man-of raw materials supplier through to the final consumer Uncertainty is one man-ofthe main issues that make supply chain operation intricate and ineffective Un-certain demand & material supply, limited production facility and the lack ofco-ordination among supply chain entities affect the smooth flow of material fromplant ware-house to market customers thereby affecting supply chain performance.Therefore, a well-oiled supply chain network is a prerequisite to compete success-fully in today’s market place This research work is focused on developing efficientperformance assessment and enhancement (i.e supply chain decision revision al-gorithm) methodologies that lead to efficient supply chain decisions or strategieswith reference to the relevant yet diverse business goals

1.2 Supply Chain Decision Levels

Supply chain aims to achieve well coordinated information, material and cashflows between raw material supplier and market customers i.e it aims to achieve aperfectly functioning supply chain Raw materials flow downstream from suppliers

to the production plants where they are transformed to value added products Thefinished products flow from plant warehouses to customers The cash flows in theupstream direction i.e from customers to the production plant and suppliers(figure 1.1) In a supply chain system, business logistics is categorized into threestages as inbound logistics, material management and product distribution Eachstage is a single or multi-echelon depending on the supply chain network and thelocations of manufacturing plant, raw material suppliers and market customers.Inbound logistics is the upstream stage of the system that mainly concerns withraw material procurement [1], supply contracts [2, 3] and selecting combination

of suppliers [4, 5] Material management is concerned about the optimum plantscheduling through Enterprise Resources Planning (ERP) and Utilization [6, 7].The product distribution part is a dominant aspect of the supply chain system anddeals with movement of products from warehouse to distribution centers, retailersand finally customers This portion of the supply chain is exposed to the problem

of uncertain customer demand The revenue generated by the supply chain systemdepends on customer satisfaction with respect to product quality, availability andservice level provided by the distribution network

1.2 Supply Chain Decision Levels

Figure 1.1: Schematic representation of Supply Chain Architecture

1.2 Supply Chain Decision Levels

The supply chain strategy represents various decision levels to attain overall ness goals The four important decision areas in supply chain management arelocation, production, inventory and distribution The strategy varies with the de-cision level, present and future focus of the companies and the objectives Thedecisions are categorized into three hierarchical levels (figure1.2) as strategic, tac-tical and operational - the difference is in the time scale of the revision periodsand the level of implementation The strategic goal is the decision made overlong time frames (yearly basis), focused on future business opportunities, compe-tition evolution and business economics Strategic supply chain network design[8], redesigning the supply chain architecture [9, 10] the number and location ofproduction sites, warehouses and distribution nodes and connectivity between the

busi-1.2 Supply Chain Decision Levelsentities, facility allocation and outsourcing are some examples of strategic deci-sions Facility allocation is a strategical approach which creates an environmentthat strongly supports the primary objectives of an organization [11] The out-sourcing decision is choosing contract manufacturer and component (raw material)supplier [12].

Figure 1.2: Schematic representation of Supply Chain Decisions

Tactical decisions are made over the time frame of a few to several months and dealswith internal strategies like supply contract, outsourcing, production- distributionplanning [13–15], capacity allocation, inventory allocation (including replenish-ment rule, parameters) and transportation strategy (lead time [16] and logistics[17]) The capacity allocation problem assigns production targets in relation tothe forecasted demand Inventory allocation aims at maintaining a buffer stock

in order to attain high customer satisfaction, low excess inventory and backorder

at retail and distribution facilities [18, 19] Operational strategies focus on day

to day issues associated with managing the product flow effectively and efficiently

1.3 Supply Chain Structurewithin the strategically planned supply chain It includes production scheduling,order processing, demand forecasting, distribution planning, product replenish-ment, transshipment and emergency shipment The Transshipment [20] is theredistribution of the stock within the same tier (e.g retailers), Emergency ship-ment bypasses the distributor and moves material from supplier to retailer, whenthe retailer’s inventory falls to a certain level Both Transshipment and Emer-gency shipment are the alternative options to facilitate necessary products whenimmediate suppliers fall short to supply the required products In general, alldecisions should coordinate with the higher decision levels so as to attain overallbusiness goals.

1.3 Supply Chain Structure

The five perspectives of supply chain structure are identified as dyadic, serial,divergent, convergent and network [21] The dyadic structure consists of two busi-ness entities (e.g buyer and vendor) The serial structure is obtained by cascadingseveral dyadic structures A typical serial supply chain studied in the literatureusually consists of one retailer, one distributor, one manufacturer and one supplier

A divergent structure is a modified serial structure [22] It is used to represent

a more realistic distribution supply chain in which one supplier (e.g turer) distributes stock to several downstream entities (e.g retailer, distributor)

manufac-A convergent structure, which is another modification of a serial supply chain,basically represents the manufacturing supply chain in which several components

1.4 Supply Chain Costand materials provided by suppliers are assembled by a manufacturer The con-figuration (e.g number of tiers, parts and number of suppliers and manufacturers

in each tier) of the convergent structure depends on the bill-of-materials of theend product Finally, a network structure is a combination of the convergent anddivergent structures It represents a complex supply chain The dyadic structure

is usually studied by an analytical model, because the simplicity of the model lows a complete mathematical analysis It is observed that there is less literature

al-in which mathematical analyses are employed to study convergent and divergentstructures This is because the complexity of such structures does not allow astraightforward extension of the results obtained in a dyadic structure Thesecomplex structures are usually investigated by using a simulation approach It

is clear that the selection of supply chain modeling approach and supply chainanalysis is highly dependent on the supply chain structure

The total supply chain cost consists of material acquisition cost, production cost,inventory carrying cost, transportation cost and order management cost (figure

1.3) Procurement cost is the cost of raw material procured during the turer’s time of interest Production cost is the value added (manufacturing cost)during the window of manufacturing interest Inventory cost is the holding cost

manufac-of finished product at hand and on-road It includes the cost manufac-of storage for space,insurance against fire, flood and theft, inventory shrinkage and obsolescence Theprofit gained is described as the difference between the revenues and the supply

1.5 Centralized and Decentralized Supply Chainschain cost [23] The revenue is the monetary value of product sales at the market.

Figure 1.3: Supply Chain Cost Structure

1.5 Centralized and Decentralized Supply Chains

In a supply chain, entities such as suppliers, manufacturers, distributors, and tailers, can belong to a single organization or independent organizations However,the distinction between centralized and decentralized systems is more properly re-lated to the incentive structures within the chain At the most basic level, in

re-a centrre-alized supply chre-ain, there is re-a centrre-al plre-anner who mre-akes decisions forthe entire system, whereas each entity in a decentralized system functions as anautonomous unit Decentralized control policies can be easily implemented andanalyzed at the local level (department, firm, etc) However, coordinated planning

of the individual entities in a way that optimizes the value of the overall supplychain (system) is a difficult undertaking Research tools that are used for plan-ning such systems include network flow models and Mixed Integer Programming(MIP) models The performance of both the centralized and decentralized system

1.6 Importance of Decentralized Distribution Networks

is influenced by delays in information and material transfer and inaccurate mand forecasting [24] The support model developed for HP Company confirmedthat partly centralized system manages better material flow across the organiza-tional barrier than centralized and decentralized system [25] An investigation of

de-a multi-product multi-echelon supply chde-ain system using de-advde-anced control strde-at-egy [26] concludes that centralized control of the overall network provides betterperformance than decentralized management of individual nodes in the supplychain network Although the centralized management provides better benefitsthan decentralized management, decentralized management is unavoidable in thereal world where the entities of the distribution network belong to different com-panies and prefer to focus only on their individual performances As a result,significant research efforts are required to develop decentralized supply chain net-work models to study and capture the complexity of the interactions among variousdecision-makers influencing the overall performance of the supply chain [27]

strat-1.6 Importance of Decentralized Distribution Networks

The prevailing challenges in the supply chain arise from the large number of theinbound and outbound material and information flows that converge in and di-verge from the plants [28] Inventory management and logistics become increas-ingly more important in the distribution network because of the large number ofmaterial flows and the uncertainty associated with the demands and the inappro-priate distribution nodes In real circumstances, the individual distribution nodes

1.6 Importance of Decentralized Distribution Networksbelonging to different organizations prefers to optimize their performance with-out considering the adverse effects caused to the other parts of the network Alldistribution nodes prefer to achieve maximum output (sales-profit and customersatisfaction) at minimum resources (excess inventory and backorder) by aggres-sive replenishment of products from the suppliers Such aggressive replenishment

is the source of demand distortion which amplifies the demand information whilepropagating through non-optimum distribution nodes up to the production site.The production site facing distorted demand (and not the true market demand),produces product according to the distorted demand information Production ac-cording to amplified demand results in the purchase of excess raw material, highoperating cost and excess inventory in the distribution nodes This lead to anincrease in supply chain cost and/or product cost and ultimately affect productsales and supply chain reputation This problem can be eliminated by integrat-ing the activities within the entities and across other entities in the supply chain[29] The risk of collapse in the business (or to avoid profit loss) can be rectified

by the effective synchronization between inbound logistics, material managementand outbound logistics Complexity of the chain varies from firm to firm For ex-ample, new government legislation encourages major manufacturing companies torecycle their used products for environmental protection and associated economi-cal benefits [30] There is plenty of evidence to suggest that major manufacturingcompanies are shifting from traditional to closed loop supply chains [31] Now,distribution network not only manages the new product inventory but is also in-volved in used-product collection, storage and transportation The complexity ofdistribution network is increased by the additional uncertain used-material flow

1.7 Research ScopeFor an efficient distribution, both internal and external integration between dis-tribution nodes is required to provide smooth material flow between productionplant warehouse and market customers and to handle sudden fluctuations in thedemand from uncertain customers.

As discussed in the previous sections, this thesis concentrates on improving theperformance of decentralized supply chains with reference to their diverse busi-ness goals In particular, this work emphasizes on the methodologies to identifyright supply chain decisions for performance enhancement The methodologiesdeveloped takes advantage of supply chain model, forecasted uncertain inputs andoptimization algorithms to synchronize the internal entities of large scale supplychains Computational studies were carried out to investigate the feasibility of theproposed performance enhancement strategies on realistically sized supply chains.Based on a thorough understanding of the supply chain characteristics and busi-ness goal(s), well-established optimization algorithms are utilized in this work toenhance supply chain performance

Control theory concepts help to model, analyze and understand the dynamics ofthe supply chains [32] Advanced control strategies such as model predictive con-trol [26, 33], minimum variance control [19] and model reference control [34, 35]have the advantage of being able to provide superior and precise control All ad-vanced strategies either require additional information and/or have implementa-tion difficulty, while extending to the large scale networks Heuristic based control

1.8 Thesis Overviewpolicies are preferred over advanced control strategies owing to their simplicityand ease of implementation in decentralized supply chains.

Our primary interest is to identify the potential opportunities to improve supplychain behavior by revising the tactical decisions of the supply chain entities Wehave taken the role of a third party supply chain consultant and exercised thefollowing options: (a) enhance the overall performance with minimal modifica-tions in the supply chain decisions (b) achieve superior performance at all internalentities of the network in multi-objective fashion (c) improve the predictability

of the supply chain nodes by reducing the uncertainty transfer (import/export)for the benefit of individual entities as well as overall system and (d) tackle largescale closed supply chains (integrated manufacturing, distribution & used productrecycle system) where the revision in supply chain decision is very challenging andextremely expensive The viabilities of the proposed concepts are complemented

by realistic simulation examples

(inven-1.8 Thesis Overview

4, a performance assessment and enhancement framework is developed for a centralized supply chain The proposed methodology aims to focus on identifyingthe problematic supply chain entities to resolve them through revised supply chaindecisions which are easy to implement The workability of this framework is in-vestigated using two case study problems involving stationary and non-stationarydemand patterns

de-Real world supply chains aim for multiple performance criteria’s such as mal supply chain cost and maximum customer satisfaction Chapter 5 aims tosupport the supply chains interested in multi-objective performance metrics Amulti-objective performance optimization methodology is developed to identify theright supply chain decisions to the improved supply chain performance depending

mini-on their business objective such as customer focused, cost effective and optimalperformance tradeoff strategies

Uncertainty is a major issue that affects the predictability of the supply chain andleads to deviation of the operation and performance from optimal values Uncer-tainty caused by market customers due to rapid changes in business environmentand competition is exogenous, whereas uncertainties arising due to inefficient sup-ply chain entities practicing inappropriate internal strategies are endogenous type

An uncertainty prone supply chain requires more investment than a similar butuncertainty free supply chain to satisfy similar market customers In other words,the return on investment is less in uncertainty prone supply chains Exogenousuncertainty such as customer demand cannot be controlled due to complex in-teraction with other uncertain variables such as customer willingness to buy the

1.8 Thesis Overviewproduct, product availability, seasonal changes and competitive products Endoge-nous uncertainty can be reduced by synchronizing all internal entities to improvethe predictability of all supply chain entities Chapter6investigates the benefit ofreducing endogenous uncertainty (i.e increasing the predictability) of the supplychain entities on the performance of the overall network.

Supply chain cost and customer satisfaction are not the only performance sures in real world supply chains Environmental factors also play a significantrole Government regulations necessitate supply chain practitioners to take backused products for recycling to minimize environmental impact Reusing the usedproducts to produce new products depends on the product type and customer will-ingness to use, recycle & reuse refurbished products The products produced fromrecycled material have advantages both in cost and time Synchronizing the pro-duction system with used material recycle system and new product distributionsystem is the exigent problem in closed loop supply chains Chapter 7 consid-ers the issues involved in closed loop supply chain decision making and outline anovel optimization methodology to derive feasible decisions in reasonably cheaptime and cost Chapter 8 consolidates the conclusions derived from this researchstudy and provides suggestions for future works

2.1 Inventory Management Methods

Various inventory management strategies are practiced in real world supply chainsdepending on the product type, demand pattern, product flow strategy and busi-ness goals Some well established strategies are push driven strategy, pull (de-mand) driven strategy, just-in-time and vendor managed inventory In this the-sis, our focus is mainly on demand-driven supply chains Demand driven supplychain handles functional products which are expensive and faces highly uncertaindemand Inventory management is the crucial decision necessary for this kind ofsupply chain where (a) insufficient inventory creates unsatisfied customer and ruinthe business or (b) managing excess inventory than the required level diminishesthe profit margin

2.1.1 Push and Pull Inventory Systems

In push-inventory system, products are produced based on long term demand casts using historical demand patterns Push strategy takes longer time to respond

fore-to changes in demand, and can result in oversfore-tocking or unacceptable service els At overstock situation, to prevent product obsolescence, finished product ispushed to the market customer through distribution network by various promo-tional schemes like cost cut-off to create or increase the sales opportunity of theproduct in the market Push strategy is beneficial for commodity products (water,electricity) that have steady demand and lower unit price In this case, productsare produced at maximum capacity to utilize the benefits of economics of scale

lev-In pull (demand) driven system, the products are produced and distributed to the

2.1 Inventory Management Methodsmarket customers according to the forecasted demand This strategy responds tothe market demand through demand information flow from actual customers tothe production plant through various intermediate distributors The pull strategyholds for functional products having uncertain demand (e.g seasonal, innovativeand fashion markets with unpredictable demand and high obsolescence risks) Attimes, pull strategy breaks down and fails to react to the demand fluctuationsbecause of long lead time and poor demand forecasting Long lead times andinefficient forecasting necessitate that the entire distribution system must holdinventory as a safety buffer stock to engage sudden fluctuations in the marketdemand and to maintain smooth product flow between manufacturing firms andmarket customer irrespective of the uncertain demand This also reduces the ef-fect of the mismatch between the production and the demand The dynamics

of the supply chain varies with the product type like fast moving product, slowmoving product, perishable and recyclable product Few supply chain adopt ahybrid push/pull strategy, and implement push at the beginning stages (close toproduction) and pull at the later stages (assembly line) as seen in reference [37]

2.1.2 Just-in-Time and Vendor Managed Inventory

Just-in-Time (JIT) inventory management is preferred for products produced bymake-to-order production systems for the benefit of improving the return on in-vestment [38] A well tuned JIT [39] has competitive advantage with less/noinventory holding cost, frequent delivery, short lead time, and close supplier ties

In JIT, customer order triggers the raw material purchase and production to isfy the customer orders and to maintain the inventory in-process respectively

sat-2.1 Inventory Management MethodsInventory in-process depends on historical demand; sudden fluctuation in demandmay cause adverse effect and cause customer satisfaction issues D’Ouville et al[40] analyzed the trade-off between conventional push and JIT production meth-ods JIT guarantees substantial savings in inventory costs by sacrificing at leastone important degree of freedom (optimizing the rate of intermediate productionprocess), which may lead to substantial inefficiency in production JIT type willsuit systems with reliable suppliers [41] and customers having loyalty towards theproduct type Both customer and manufacturer gain by adopting the JIT strat-egy For example, in an automobile system, customers have the choice to customizetheir requirement at the expense of having to deal with the production lead time.

At the same time, the manufacturer has the advantage of holding no/less raw terial and finished product inventory Vendor-managed inventory (VMI) system[42] is an interesting option in supply chains where the retailers face high risk ofdemand uncertainty In VMI system, retailers share demand information with theupstream nodes; upstream nodes will make decision to maintain inventory in theretailer nodes and owns the inventory until they are sold In this management, anupstream node bears the risk of demand uncertainty Many companies, such asCampbell Soup Company, Procter and Gamble, garment industry, Dell, Walmart(a pioneer and one of the major US-based supply chain) and grocery industryfollow VMI management A slight modification in VMI strategy will help bothvendor and retailer for better pay off, if they have an agreement to share the im-pact of demand uncertainty Compared to traditional pull driven supply chains,the VMI system is found to be superior, and less affected from transportationdisruptions caused between distribution echelons [43] The inventory fluctuations

ma-2.2 Bullwhip Effectare less in VMI system, even the unfilled customer orders remains approximatelythe same for both pull and VMI systems.

2.2 Bullwhip Effect

Various researchers identified bullwhip as a dominant source-causing inefficiencies

in supply chain operations Bullwhip (BW) refers to the phenomenon that ordervariability increases as orders move upstream along the supply chain (figure 2.1,[44]) This phenomenon is so well known that it is sometimes referred to as, “thefirst law of supply chain dynamics” [45] The importance and influence of bullwhiphas been well analyzed by researchers both in theory and practice Forrester [46]

is the pioneer in identifying the oscillations in supply chain due to ineffectivecoordination between internal entities Wikner et al [47] showed the performance

of three-echelon Forrester production-distribution system is far from optimal due

to bullwhip effect Demand forecasting, lead-times, batch orders, supply shortages,and price variations are identified as the major sources causing bullwhip in supplychain systems [24,48–50] Lead time (delay in demand/order information transferand material transportation) has been identified as the primary cause of bullwhip[51, 52] Secondary causes include inaccurate demand forecast, batch ordering,price fluctuations, rationing and shortage gaming Disney et al [53] identified andquantified the bullwhip in multi-echelon system Other than these sources, humanbehavior can also generate bullwhip effect in supply chain networks [54]

2.2 Bullwhip Effect

Figure 2.1: The Bullwhip Effect

2.2.1 Sources of Bullwhip Effect

(1) Planning and behavioral aspects: Production planning and inventory location based on distorted information from succeeding downstream nodes instead

al-of end customer demand causes bullwhip effect

(2) Batch Order: Batch order is practiced by the supply chain entities mainly toreduce set-up costs and fixed order cost by placing orders in batches periodically tosuppliers [55] Because of this ordering technique, the suppliers receive distortedand delayed information about end customer demand

(3) Price fluctuations: Depending on the market reasons, the companies varythe product prices to retailers and end customers either by promotional offers ortemporary price reductions This leads buyers to speculate, buying large quantitieswhen prices are low and avoiding buying when prices are higher This forwardbuying increases the variation in end customer demand and amplifies the demandvariation results in bullwhip [48,49]

(4) Rationing and shortage gaming: Suppliers often ration their products by

2.2 Bullwhip Effectprioritizing the customers or products during insufficient inventory situation Thiscauses delivering only a proportion of the quantity that customers order Buyersanticipating shortages and rationing will often increase the size of their orders

in excess of their actual demand to ensure that they get the amount that theyreally require As soon as delivery bottlenecks are overcome, they cancel theirorders for the unneeded quantity [48, 49] This phenomenon of gaming leavesthe manufacturer with a much distorted picture of consumer demand, and thebullwhip effect sets in

(5) Role of human behavior: Simulation experiments clearly reveal two humanstrategies causing the bullwhip effect [54] In supply chains, some practitioners actaggressively (safe harbour) by ordering more products than necessary and increasetheir safety stock Aggressive nature not only costs high capital employed in stock

at their tier but also force their suppliers to either increase their orders or to payfor out-of-stock situations Thus aggressive strategy practiced by only one tier canhave a negative impact on the whole supply chain The second extreme of humanbehavior is very conservative and aims to prevent inventory built-up and can lead

to out-of-stock situations The conservative strategy depletes the inventory beforethe end customer’s demand increases Initially, cautious ordering does not affectother nodes in the network badly But as soon as end customer’s orders increase,

a node following this strategy will order more than the required level causing anegative impact on the entire supply chain Furthermore, the node following theconservative strategy is not able to deliver for some periods causing out-of-stocksituations for its customers