The Value of Demand Postponement under Demand Uncertainty

Similar documents
Online Appendix. t=1 (p t w)q t. Then the first order condition shows that

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan

Amiri s Supply Chain Model. System Engineering b Department of Mathematics and Statistics c Odette School of Business

Price competition with capacity constraints. Consumers are rationed at the low-price firm. But who are the rationed ones?

Credit Card Pricing and Impact of Adverse Selection

k t+1 + c t A t k t, t=0

3.2. Cournot Model Cournot Model

The Minimum Universal Cost Flow in an Infeasible Flow Network

Multi-product budget-constrained acquistion and pricing with uncertain demand and supplier quantity discounts

A NOTE ON A PERIODIC REVIEW INVENTORY MODEL WITH UNCERTAIN DEMAND IN A RANDOM ENVIRONMENT. Hirotaka Matsumoto and Yoshio Tabata

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

Endogenous timing in a mixed oligopoly consisting of a single public firm and foreign competitors. Abstract

EFFECTS OF JOINT REPLENISHMENT POLICY ON COMPANY COST UNDER PERMISSIBLE DELAY IN PAYMENTS

Problem Set 9 Solutions

Market structure and Innovation

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

Hila Etzion. Min-Seok Pang

e - c o m p a n i o n

Simultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals

A LINEAR PROGRAM TO COMPARE MULTIPLE GROSS CREDIT LOSS FORECASTS. Dr. Derald E. Wentzien, Wesley College, (302) ,

Interactive Bi-Level Multi-Objective Integer. Non-linear Programming Problem

Inventory Model with Backorder Price Discount

The (Q, r) Inventory Policy in Production- Inventory Systems

Cournot Equilibrium with N firms

On the Multicriteria Integer Network Flow Problem

f(x,y) = (4(x 2 4)x,2y) = 0 H(x,y) =

Equilibrium with Complete Markets. Instructor: Dmytro Hryshko

University of California, Davis Date: June 22, 2009 Department of Agricultural and Resource Economics. PRELIMINARY EXAMINATION FOR THE Ph.D.

8. Modelling Uncertainty

How Strong Are Weak Patents? Joseph Farrell and Carl Shapiro. Supplementary Material Licensing Probabilistic Patents to Cournot Oligopolists *

(1 ) (1 ) 0 (1 ) (1 ) 0

Supporting Information for: Two Monetary Models with Alternating Markets

A Simple Inventory System

Supporting Materials for: Two Monetary Models with Alternating Markets

In the figure below, the point d indicates the location of the consumer that is under competition. Transportation costs are given by td.

ON A DETERMINATION OF THE INITIAL FUNCTIONS FROM THE OBSERVED VALUES OF THE BOUNDARY FUNCTIONS FOR THE SECOND-ORDER HYPERBOLIC EQUATION

Absorbing Markov Chain Models to Determine Optimum Process Target Levels in Production Systems with Rework and Scrapping

Assortment Optimization under MNL

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

Welfare Properties of General Equilibrium. What can be said about optimality properties of resource allocation implied by general equilibrium?

The Order Relation and Trace Inequalities for. Hermitian Operators

Solutions to exam in SF1811 Optimization, Jan 14, 2015

On the correction of the h-index for career length

Ryan (2009)- regulating a concentrated industry (cement) Firms play Cournot in the stage. Make lumpy investment decisions

PROBLEM SET 7 GENERAL EQUILIBRIUM

A Hybrid Variational Iteration Method for Blasius Equation

Fuzzy Approaches for Multiobjective Fuzzy Random Linear Programming Problems Through a Probability Maximization Model

Some modelling aspects for the Matlab implementation of MMA

A NOTE ON CES FUNCTIONS Drago Bergholt, BI Norwegian Business School 2011

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016

FIRST AND SECOND ORDER NECESSARY OPTIMALITY CONDITIONS FOR DISCRETE OPTIMAL CONTROL PROBLEMS

Chapter 11: Simple Linear Regression and Correlation

Portfolios with Trading Constraints and Payout Restrictions

The Second Anti-Mathima on Game Theory

Convexity preserving interpolation by splines of arbitrary degree

Pricing and Resource Allocation Game Theoretic Models

STOCHASTIC INVENTORY MODELS INVOLVING VARIABLE LEAD TIME WITH A SERVICE LEVEL CONSTRAINT * Liang-Yuh OUYANG, Bor-Ren CHUANG 1.

Perfect Competition and the Nash Bargaining Solution

Environmental taxation: Privatization with Different Public Firm s Objective Functions

Lecture 3: Probability Distributions

Revenue comparison when the length of horizon is known in advance. The standard errors of all numbers are less than 0.1%.

Dual-Channel Warehouse and Inventory Management with Stochastic Demand

Power law and dimension of the maximum value for belief distribution with the max Deng entropy

The optimal delay of the second test is therefore approximately 210 hours earlier than =2.

Lecture 14: Bandits with Budget Constraints

Economics 8105 Macroeconomic Theory Recitation 1

8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS

Statistics for Economics & Business

The Study of Teaching-learning-based Optimization Algorithm

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming

Predictive Analytics : QM901.1x Prof U Dinesh Kumar, IIMB. All Rights Reserved, Indian Institute of Management Bangalore

Maximizing the number of nonnegative subsets

Uncertainty and auto-correlation in. Measurement

(1, T) policy for a Two-echelon Inventory System with Perishableon-the-Shelf

Structure and Drive Paul A. Jensen Copyright July 20, 2003

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011

Lecture 4: November 17, Part 1 Single Buffer Management

ECE559VV Project Report

A NEWSVENDOR MODEL WITH UNRELIABLE SUPPLIERS

2016 Wiley. Study Session 2: Ethical and Professional Standards Application

Conjectures in Cournot Duopoly under Cost Uncertainty

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD

Quantity Precommitment and Cournot and Bertrand Models with Complementary Goods

The lower and upper bounds on Perron root of nonnegative irreducible matrices

Norm Bounds for a Transformed Activity Level. Vector in Sraffian Systems: A Dual Exercise

The oligopolistic markets

EEE 241: Linear Systems

find (x): given element x, return the canonical element of the set containing x;

Managing Capacity Through Reward Programs. on-line companion page. Byung-Do Kim Seoul National University College of Business Administration

Heuristic Algorithm for Finding Sensitivity Analysis in Interval Solid Transportation Problems

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

THE ARIMOTO-BLAHUT ALGORITHM FOR COMPUTATION OF CHANNEL CAPACITY. William A. Pearlman. References: S. Arimoto - IEEE Trans. Inform. Thy., Jan.

MULTI-ITEM PRODUCTION PLANNING AND MANAGEMENT SYSTEM BASED ON UNFULFILLED ORDER RATE IN SUPPLY CHAIN

Measuring the Impact of Increased Product Substitution on Pricing and Capacity Decisions under Linear Demand Models

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Comparison of Regression Lines

Transcription:

Recent Researches n Appled Mathematcs, Smulaton and Modellng The Value of emand Postponement under emand Uncertanty Rawee Suwandechocha Abstract Resource or capacty nvestment has a hgh mpact on the frm proftablty. However, ths decson must be made earler when demand s uncertan and t s very dffcult to change later on. Many frms search for other strateges to deal wth uncertanty n order to gan more proft and stay n a busness. Postponement strategy s one of the strateges that many companes use to hedge aganst the uncertanty. A frm can ncrease ts proftablty when t can postpone some decsons or actvtes to the tme later untl the more nformaton s obtaned. In ths paper, we consder a frm producng two substtutable products. The frm needs to make two decsons: the capacty nvestment and producton quantty. The objectve of ths work s to study how the demand postponement affects the frm s capacty nvestment and ts proftablty under demand uncertanty and how degrees of substtuton mpact our fndngs. Based on ths framework, we model the problem as a two-stage stochastc programmng. We characterze the optmal nvestment capacty and producton quantty under dfferent postponement strateges. In addton, the necessary and suffcent condtons for nvestment n flexble capacty are obtaned. eywords Capacty nvestment, emand postponement, emand uncertanty, Stochastc programmng, Substtutable Products. I. ITROUCTIO OSTPOEMET strategy s one of the strateges that P companes can use to hedge aganst uncertanty. Many type of postponement have been studed as stated n [] and [] such as tme postponement, place postponement, and form postponement, etc. There are applcatons of postponement n ndustry and results from [3] ndcated that there s a postve relatonshp between postponement and company performance. The postpone strateges we consder n ths work are the quantty postponement and demand postponement strateges. The producton quantty postponement refers to a frm s ablty to postpone ts producton quantty untl after demand nformaton s obtaned. Consderng the quantty postponement along wth the flexble resources ncreases the Ths work was supported n part by the Thaland Research Fund (TRF under Grant MRG58 and Faculty of Graduate Studes, Mahdol Unvesty. R. S. Author s wth the epartment of Mathematcs, Faculty of Scence, Mahdol Unversty, Bangkok, 4 Thaland (correspondng author to provde phone: 66--5346; fax: 66--5343; e-mal: scrsw@ mahdol.ac.th. frm proftablty. The frm s flexble resources allow the frm to swtch capacty amount of the multple products and provdes a rsk poolng effect. Many researchers have focus on the flexblty provded by the producton postponement strateges as shown n [], [4], and [5]. The mpact of quantty postponement on the optmal capacty has been studed by [6] for a sngle-product frm. On the other hand, there s a few works studyng on demand postponement. [] formulates and solves the capacty plannng problem for one product. In ths paper, we extend the model to the substtutable products. In addton, we are also nterested n the effect of the demand postponement strategy when the frm has mplemented the producton quantty postponement. Therefore, the objectve of ths paper s to consder the mpact of the demand postponement on the optmal capacty nvestment under dfferent postponement strateges. Specfcally, we consder the no postponement and the producton quantty postponement as base cases. Then, we examne the effect of demand postponement on the optmal capacty nvestment and the frm s proft under both strateges as well as study how the degrees of substtuton between products affect the fndngs. II. OTATIOS, MOELS, A ASSUMPTIOS In ths work, we consder a frm that produces two substtutable products n a monopolstc settng. Ths frm needs to determne capacty, prcng, and producton quantty. Let denote the capacty of the flexble resources that can produce both products, c be a unt nvestment cost, c h be a unt holdng cost, c q be a unt product cost, c be a dscount prce (or rembursement cost for the demand n postponement perod. The producton cost for each tem s ncluded here, d be a demand of product, p be a sellng prce of product and, wthout loss of generalty, let (p p, s be an amount of sellng product, q and q p be producton quanttes of product for regular and postponement perod, respectvely. Tmelne s dvded nto two stages shown n Fg. Stage s a plannng perod and Stage ncludes the regular and postponement perods. ISB: 98--684-6-9 6

Recent Researches n Appled Mathematcs, Smulaton and Modellng Stage Stage Plannng Perod, q,q emand s realzed Fgure : Tmelne Regular Perod, q,q q p,q p q,q elvery Tme q,q, q p,q p In Fg, the nvestment capacty ( s determned n the plannng perod when demand s hghly uncertan. There are four cases that are consdered, o postponement (, emand postponement (, Producton postponement (P, and producton and demand postponement (P. All the decson varables are determned n the plannng perod under strategy (. For strategy (P, the producton quantty, q, s determned after the demand curves are realzed. When the demand postponement strategy s mplemented, the postponed producton quantty, q p, s determned after the demand curves are realzed and ths products wll be delvered n the postponement perod. We assume that the fracton of unsatsfed demand, -, s retaned n the postponement perod. That s, when the demand exceeds the producton quantty, there s only (-(d -q remanng demand n the postponement perod. Aggregate lnear demand functon for two substtutable products s consdered. It s a functon of the prces of both products: d = ξ αp βp3, =, ξ s an ntercept of product, β s the degree of substtuton ( β α. Let E[.] be an expectaton operator wth respect to random varables, ξ, and Pr(. be a probablty. Let π (. be the proft functon n Stage. III. MATHEMATICAL FORMULATIOS Postponement Perod tmelne Strategy Strategy Strategy P Strategy P From the above, we can dvde the problem nto two stages. Stage corresponds to the tme when the demand s uncertan and Stage s the tme after the demand nformaton s obtaned. The mathematcal model for the problem under demand strategy ( can be formulated as (stage max V E[ π (, q] c ( c c q qp,, = q h = s.t. q ( q, for =, (stage max π ( q, = ps ( p cq sq, p p = = ( s.t. s q for =, (3 s ( ε αp βp3 for =, (4 qp q for =, (5 qp ( ( ε αp βp3 s for =,(6 s, q for =, ( p The objectve of the model n Stage s to maxmze the expected proft functon. The objectve functon n Stage s to maxmze the frm proft by determnng the amount of sell, s, and the postponed producton quantty q p. Constrant ( states that the total producton quantty should not exceed the frm s capacty. Constrants (3 and (4 mply that the amount of sell cannot exceed producton quantty and nonnegatve demand of each product. In addton, (5 and (6 ensures that the postponed producton quantty, q p, does not exceed the producton quantty and t does not exceed retaned demand. Constrants ( and ( are non-negatvty constrants for nvestment capacty, producton quantty, sellng amount, and postponed producton quantty, respectvely. ote that the model for the Problem s smlar except that q p s set to zero. The mathematcal model for the problem under producton quantty and demand postponement strategy (P can be formulated as (stage max V E[ π ( ] c s.t. (8 qq, p (stage max π ( = ( p c q ( p c q q p = = s.t. q q (9 q ( ε αp βp3 for =, ( qp qp ( qp ( ( ε αp βp3 s for =, ( q, q for =, (3 p The objectve of the model n Stage s to maxmze the expected proft functon. In ths case, the frm determnes the producton quantty after the demand s realzed. Therefore, the producton and the holdng costs do not occur n the frst ISB: 98--684-6-9 68

Recent Researches n Appled Mathematcs, Smulaton and Modellng stage. In Stage, the objectve functon s to maxmze the frm proft by determnng the producton quantty and the postponed producton quantty gven that the nvestment capacty and prces. Constrants (9 and ( state that the total producton quantty and the total postponed producton quantty should not exceed the frm s capacty, respectvely. Constrant ( mples that the producton quantty should not exceed the nonnegatve demand of each product. In addton, ( the postponed producton quantty (q p does not exceed the retaned demand. (8 and (3 are non-negatvty constrants for nvestment capacty, producton quantty, and postponed producton quantty. The model for the producton quantty postponement (P s smlar to Problem (P except that q p s set to zero. IV. RESULTS The optmal soluton to Problems,, and P can be characterzed n the smlar procedures. Hence, n what follows, we focus on Problem P. To analyze the problem, we solve the problem backward. Frst, for a gven set of p and p, we derve the optmal soluton for the problem n Stage and show that the expected proft functon n Stage s strctly jontly concave n q, and q. Consequently, the optmal soluton s unque and the arush-uhn-tucker (T frstorder condtons are necessary and suffcent for optmalty to the problem. Ths leads to the followng results. Theorem The producton quanttes q and q are the unque optmal soluton to Problem P f and only f there exsts v that satsfes the followng condtons: [ c cp ( pc]pr( 5 [ c cp ( p c]pr( 6 9 [ p cp ( ( p c]pr( [ pcp c]pr( [ p cp c]pr( 8 = c v v = 5 = { αp βp < ξ < αpβp, ξ < αp βp} 6 = { ξ < αpβp, αp βp < ξ < αp βp}, = { ξ > αpβp}, 8 = { ξ < αp βp, ξ > αp βp}, 9 = { αp βp < ξ < αpβp, ( α β( p p < ξ ξ < ( α β ( p p} = { αp βp < ξ < αp βp, ξ ξ > ( α β( p p} = { ξ > αp βp, ξ > αp βp, ξ ξ < ( α β( p p} = αp βp < ξ < αp βp, ξ ξ > ( α β( p p Proof: See Appendx. Theorem The producton quanttes q and q are the unque optmal soluton to Problem f and only f there exsts v and v that satsfes the followng condtons: ppr( ξ > q αp βp = c ch cq v ppr( ξ > q αp βp = c ch cq v q v =, for =, Theorem 3 The producton quanttes q and q are the unque optmal soluton to Problem f and only f there exsts v and v that satsfes the followng condtons: ( p cpr( ξ > q αp β p ( p ( cpr( = c ch cq v ( p cpr( ξ > q αp β p ( p ( cpr( = c ch cq v q v =, for =, = { q αp βp < ξ < q αpβp}, = { q αp βp < ξ < q αp βp}. Theorem 4 The producton quanttes q and q are the unque optmal soluton to Problem P f and only f there exsts v that satsfes the followng condtons: ppr( ξ > αp β p, ξ > αp β p ppr( ξ > αp β p ppr( = c v v = = { αp β p < ξ < αp β p, ξ ξ > ( α β( p p } The optmal capacty nvestment under each postponement strategy follows a threshold polcy. Corollary The optmal nvestment polcy under postponement strategy s one of the followng cost threshold polces: If c ch cq < c, then >. Otherwse, =. * = q q the threshold values s gven as ISB: 98--684-6-9 69

Recent Researches n Appled Mathematcs, Smulaton and Modellng c = max{ p Pr( ξ > αp β p, p Pr( ξ > αp βp } c ch cq c If <, then >. Otherwse, =. * = q q the threshold values s gven as c = max{ ( p cpr( ξ > αp βp,( p cpr( ξ > αp βp } 3 c c If <, then >. Otherwse, =. the threshold values s gven as c = p Pr( ξ > αp βp p Pr( < ξ < αp βp, ξ > αp βp 4 c c If <, then >. Otherwse, =. the threshold values s gven as c = ( pc cppr( ξ > αp βp ( p c cppr( < ξ < αp β p, ξ > αp βp Proof: It follows drectly from Theorems -4. The above corollary suggests that f the total unt cost c ch cp s less than the cost threshold, the frm always nvests n the capacty under Strateges and. On the other hand, f the total unt cost s hgher than the cost threshold, t s better for the frm not to nvest. Smlarly, under the Strateges P and P, f the unt nvestment cost s less than the cost threshold value ndcated n Corollary, then the frm should nvest n the capacty. Corollary c c and c c Proof: It follows drectly from Proposton. Corollary ndcates that there are some regons that the frm should nvest n ts capacty to gan more proft when the demand postponement s consdered. ext, we perform a numercal study to nvestgate the mpact of the optmal nvestment capacty changes n demand substtuton parameter β. Results from the numercal studes suggest that the optmal * nvestment capacty s ncreasng n β for all strateges (,, P, P as shown n Fg.. The results show that the frm wll nvest more when the products are more substtutable. Ths s because the frm can take advantage from the substtuton. However, the rate of change of the optmal nvestment capacty s less when producton postponement s mplemented. Ths s because of the flexble resources whch s also used to hedge aganst demand varablty. It s not surprsng that when the products are ndependent or not substtutable the frm nvests more under strategy P. Ths s because the frm has ablty to deal wth the demand varablty by usng the producton quantty postponement from the flexble resource. The results also show that the demand postponement strategy leads to nvest less n the frm s capacty and t leads to lower nvestment when addton producton quantty s mplemented along wth. Moreover, our fndng ndcates that the frm nvests less when the demand postponement s addtonally mplemented. ue to the ablty of the frm to postpone part of the demand to tme later, t may cause the frm nvests less. 8 6 4 Proft...3.4.5.6..8.9 Fgure 3: The frm's proft versus β/ν. Smlar results are obtaned for the frm s proft. Fg 3 shows that the proft s ncreasng n substtuton parameter for all strateges. The demand postponement generates the hgher proft for the frm. P P Optmal nvestment capacty 35 3 5 5 5...3.4.5.6..8.9 Fgure : Optmal nvestment capacty * versus β/ν. P P V. COCLUSIO In ths paper, we study the mpact of the demand postponement on the optmal capacty under demand uncertanty for a two-product frm n a monopolstc settng. The frm makes decsons nvestment capacty and product quantty to maxmze the proft functon. We formulate mathematcal models for dfferent postponement strateges. The necessary and suffcent condton for the optmal capacty nvestment s obtaned for each strategy. Our fndngs show that mplementng the demand postponement leads to nvest less n optmal capacty and ncrease n the frm s proftablty. Our results come wth lmtatons. We consder the lnear demand functon whch depends on the prce of the products. ISB: 98--684-6-9

Recent Researches n Appled Mathematcs, Smulaton and Modellng It would be nterestng to consder other realstc demand functon. We also assume a frm that s n a monopolstc settng. Consderng the competton would be another nterestng extenson to our models. APPEIX Consder problem, we frst solve the Stage Problem. The demand space can be decomposed nto dfferent dsjont sets; see Fg. 4. αp β p α p β p α p β p 8 6 3 9 α p 4 5 β p α pβ p Fgure 4: The demand space for Problem P We can obtan the closed form expressons for the producton quanttes and postponed producton quanttes for each regon. It s easy to verfy that the optmal soluton to the Stage Problem s unque and the optmal Stage proft can be obtaned drectly. Before we solve the Stage problem, let ε be a realzaton of ξ and f (, be a jont probablty densty functon of ξ and ξ, respectvely. To solve the Stage Problem, the Stage proft s consdered. We derve E[ Π( ] = [ c cp ( p c]pr( 5 [ c c ( p c]pr( p 6 9 p p cp c cp c]pr( [ p c ( ( p c]pr( [ ]Pr( [ p 8 α p β p E[ Π( ] = αp β p αp β p αp β p [( p c ( ( p c] p f( αp βp, ε dε ( p p f( αp βp, ε dε [ p ( c cp] f( ε, ( α β( p p dε <. αpβ p ( ( p c f ( ε, ( α β( p p ε dε αpβ p It can be shown that V = E[ π ( ] c s strctly concave n. Therefore, the optmal soluton to the Stage Problem s unque, and snce the constrants are lnear, the T frst-order optmalty condtons, gven n Theorem, are necessary and suffcent for optmalty. Smlar method can be used to obtan the results n Theorems -4. ACOWLEGMET I would lke to thank Thaland Research Fund (TRF for the fnancal support and r. ancht Malavongs for all hs comments and support for ths research. REFERECES [] J. M. Swamnathan and H. L. Lee, esgn for Postponement, Handbook of OR/MS on Supply Chan Management, Ed., Graves, S. and de ok, 3. [] B. Yang,.. Burns, and C. J. Backhouse, Management of Uncertanty through postponement, Int. J. Prod. Res. Vol 4, o. 6, 4, pp. 49-64. [3] B. Yang,.. Burns, and C. J. Backhouse, The Applcaton of Postponement n Industry, IEEE Trans. Eng. Manage., Vol. 5, o., May 5, pp. 38-48. [4] C. H. Fne and R. M. Freund, Optmal Investment n Product-Flexble Manufacturng Capacty, Manage. Sc., Vol. 36, 99, pp. 449-466. [5] J. A. Van Meghem, Investment strateges for flexble resources, Manage. Sc. Vol. 44, 998, pp. -8. [6] J. A. Van Meghem and M. ata, Prce versus Producton Postponement: Capacty and Compeptton, Mange. Sc., Vol 45, 999, pp. 63-649. [] A. V. Iyer, V. eshpande, and Z. Wu, A Postponement Model for emand Management, Manage. Sc., Vol. 49, o. 8, August 3, pp. 983-. Rawee Suwandechocha receved a B.S. n mathematcs from Unversty of Rochester, Y, USA, n 999, and M.S. and Ph.. n Operatons Research from epartment of Industral and System Engneerng, Vrgna Polytechn c Insttute and State Unversty, Blacksburg, VA n 6. She s a lecturer n the epartment of Mathematcs, Faculty of Scence, Mahdol Unversty, Bangkok, Thaland. Her research nterests nclude appled operatons research, supply chan and logstcs, nventory, and capacty plannng. ISB: 98--684-6-9