1 FIFTH International Conference on ``Analysis of Manufacturing Systems -- Production Management'‘ Zakynthos, Greece, 2005 QUEUEING MODELS FOR MANAGING.

Slides:



Advertisements
Similar presentations
Independent Demand Inventory Systems
Advertisements

Statistical Inventory control models I
Lesson 08 Linear Programming
Unità di Perugia e di Roma “Tor Vergata” "Uncertain production systems: optimal feedback control of the single site and extension to the multi-site case"
Dr. A. K. Dey1 Inventory Management, Supply Contracts and Risk Pooling Dr. A. K. Dey.
“Make to order or Make to Stock Model: and Application” S.Rajagopalan By: ÖNCÜ HAZIR.
Inventory Control Chapter 17 2.
Introduction to Management Science
Inventory Management Chapter 16.
Chapter 11, Part A Inventory Models: Deterministic Demand
HW #7 ANSWER
Lecture 13 – Continuous-Time Markov Chains
1 HEURISTICS FOR DYNAMIC SCHEDULING OF MULTI-CLASS BASE-STOCK CONTROLLED SYSTEMS Bora KAT and Zeynep Müge AVŞAR Department of Industrial Engineering Middle.
Markov Decision Models for Order Acceptance/Rejection Problems Florian Defregger and Heinrich Kuhn Florian Defregger and Heinrich Kuhn Catholic University.
Dynamic lot sizing and tool management in automated manufacturing systems M. Selim Aktürk, Siraceddin Önen presented by Zümbül Bulut.
“Batching Policy in Kanban Systems”
An overview of design and operational issues of kanban systems M. S. AKTÜRK and F. ERHUN Presented by: Y. Levent KOÇAĞA.
Chapter 5 Inventory Control Subject to Uncertain Demand
Planning operation start times for the manufacture of capital products with uncertain processing times and resource constraints D.P. Song, Dr. C.Hicks.
Analysis of a Yield Management Model for On Demand IT Services Parijat Dube IBM Watson Research Center with Laura Wynter and Yezekael Hayel.
A Simple Inventory System
Inventory Management for Independent Demand
MNG221- Management Science –
WINTER 2012IE 368. FACILITY DESIGN AND OPERATIONS MANAGEMENT 1 IE 368: FACILITY DESIGN AND OPERATIONS MANAGEMENT Lecture Notes #3 Production System Design.
Stevenson and Ozgur First Edition Introduction to Management Science with Spreadsheets McGraw-Hill/Irwin Copyright © 2007 by The McGraw-Hill Companies,
Dimitrios Konstantas, Evangelos Grigoroudis, Vassilis S. Kouikoglou and Stratos Ioannidis Department of Production Engineering and Management Technical.
Chapter 12 Inventory Models
Independent Demand Inventory Management
1 Chapter 5 Flow Lines Types Issues in Design and Operation Models of Asynchronous Lines –Infinite or Finite Buffers Models of Synchronous (Indexing) Lines.
Inventory/Purchasing Questions
Linear Programming Topics General optimization model LP model and assumptions Manufacturing example Characteristics of solutions Sensitivity analysis Excel.
5-1 ISE 315 – Production Planning, Design and Control Chapter 5 – Inventory Control Subject to Unknown Demand McGraw-Hill/Irwin Copyright © 2005 by The.
1 Inventory Control with Stochastic Demand. 2  Week 1Introduction to Production Planning and Inventory Control  Week 2Inventory Control – Deterministic.
Lecture 14 – Queuing Networks Topics Description of Jackson networks Equations for computing internal arrival rates Examples: computation center, job shop.
Managing Uncertainty in Supply Chain: Safety Inventory Spring, 2014 Supply Chain Management: Strategy, Planning, and Operation Chapter 11 Byung-Hyun Ha.
Markov Decision Processes1 Definitions; Stationary policies; Value improvement algorithm, Policy improvement algorithm, and linear programming for discounted.
Decision Making in Robots and Autonomous Agents Decision Making in Robots and Autonomous Agents The Markov Decision Process (MDP) model Subramanian Ramamoorthy.
ECES 741: Stochastic Decision & Control Processes – Chapter 1: The DP Algorithm 49 DP can give complete quantitative solution Example 1: Discrete, finite.
ECES 741: Stochastic Decision & Control Processes – Chapter 1: The DP Algorithm 31 Alternative System Description If all w k are given initially as Then,
On a Production/Inventory System with Strategic Customers and Unobservable Inventory Levels Can Oz Fikri Karaesmen SMMSO June 2015.
Networks of Queues Plan for today (lecture 6): Last time / Questions? Product form preserving blocking Interpretation traffic equations Kelly / Whittle.
Practical Dynamic Programming in Ljungqvist – Sargent (2004) Presented by Edson Silveira Sobrinho for Dynamic Macro class University of Houston Economics.
CHAPTER 5 Inventory Control Subject to Uncertain Demand McGraw-Hill/Irwin Copyright © 2009 by The McGraw-Hill Companies, Inc. All rights reserved.
1 The Base Stock Model. 2 Assumptions  Demand occurs continuously over time  Times between consecutive orders are stochastic but independent and identically.
1 Managing Flow Variability: Safety Inventory Operations Management Session 23: Newsvendor Model.
WIN WIN SITUATIONS IN SUPPLY CHAIN MANAGEMENT Logistics Systems 2005 Spring Jaekyung Yang, Ph.D. Dept. of Industrial and Information Systems Eng. Chonbuk.
PROBABILITY AND STATISTICS FOR ENGINEERING Hossein Sameti Department of Computer Engineering Sharif University of Technology Principles of Parameter Estimation.
An Optimal Design of the M/M/C/K Queue for Call Centers
The (Q, r) Model.
MBA 8452 Systems and Operations Management
Operations Research II Course,, September Part 3: Inventory Models Operations Research II Dr. Aref Rashad.
1 Inventory Control with Time-Varying Demand. 2  Week 1Introduction to Production Planning and Inventory Control  Week 2Inventory Control – Deterministic.
IT Applications for Decision Making. Operations Research Initiated in England during the world war II Make scientifically based decisions regarding the.
Inventory Management for Independent Demand Chapter 12.
1 Chapter 4 Single Stage Produce-to-Stock Better customer service Lower manufacturing costs (setups) More uniform utilization.
Queuing Theory.  Queuing Theory deals with systems of the following type:  Typically we are interested in how much queuing occurs or in the delays at.
1 1 © 2003 Thomson  /South-Western Slide Slides Prepared by JOHN S. LOUCKS St. Edward’s University.
1 Optimization Techniques Constrained Optimization by Linear Programming updated NTU SY-521-N SMU EMIS 5300/7300 Systems Analysis Methods Dr.
Data Consolidation: A Task Scheduling and Data Migration Technique for Grid Networks Author: P. Kokkinos, K. Christodoulopoulos, A. Kretsis, and E. Varvarigos.
Energy Optimal Control for Time Varying Wireless Networks Michael J. Neely University of Southern California
Decentralized Multi-Echelon Supply Chains: Incentives and Information Hau Lee Seungjin Whang presented by Evren Sarınalbant.
Flows and Networks Plan for today (lecture 6): Last time / Questions? Kelly / Whittle network Optimal design of a Kelly / Whittle network: optimisation.
Recoverable Service Parts Inventory Problems -Ibrahim Mohammed IE 2079.
Inventory Control. Meaning Of Inventory Control Inventory control is a system devise and adopted for controlling investment in inventory. It involve inventory.
Inventory Control Models 6 To accompany Quantitative Analysis for Management, Twelfth Edition, by Render, Stair, Hanna and Hale Power Point slides created.
Lecture 14 – Queuing Networks
Task: It is necessary to choose the most suitable variant from some set of objects by those or other criteria.
IV-2 Manufacturing Systems modeling
Dr. Arslan Ornek MATHEMATICAL MODELS
Presentation transcript:

1 FIFTH International Conference on ``Analysis of Manufacturing Systems -- Production Management'‘ Zakynthos, Greece, 2005 QUEUEING MODELS FOR MANAGING INVENTORIES, BACKORDERS, AND QUALITY JOINTLY IN STOCHASTIC MANUFACTURING SYSTEMS Vassilis S. Kouikoglou, Technical Univ. of Crete, Greece Stratos Ioannidis, University of the Aegean, Greece Georgios Saharidis, Ecole Centrale de Paris, France

2 Basic components of a production system:  Production facilities: processing units (machines) intermediate storage and transferring of parts (buffers) quality control (inspect, rework, scrap)  Sales department

3 Production control objective Maximize profit from sales less quality, inventory, backlog, etc., costs. Subproblems a)Production control: when to produce and when to stop producing? Stock is costly, but so are stockouts. b)Quality control: accept, rework, or reject a finished item, based on deviations of product characteristics from target values. Rework and scrapping of parts are costly and cause delays in production. c)Admission control: During a stockout period should we backorder incoming orders or reject them? Any better practices other than Lost Sales or Complete Backordering? Remarks Criteria (a)-(c) are in conflict. Space of admissible control policies is vast. Analytical models are not always accurate and simulation is often time consuming. We examine a restricted set of controls which are optimal only for simple systems (base stock, Kanban).

4 Example 1: Base stock levels in a two-stage supply chain Assumptions Processing times in factory M i are exponential rv’s with rates  i Demand is Poisson with rate Demand during stockout periods in buffer B 1 or B 2 is satisfied immediately by purchasing from subcontractors Factory M i produces until stock n i reaches base stock level b i, i = 1, 2 11 22

5 Parameters p 1 price at which factory M 1 sells a component to M 2 p 2 selling price of the final product c i unit production cost at factory M i s i cost of purchasing one unit from subcontractor i h i unit holding cost rate in buffer B i

6 Equilibrium probabilities P(n 1, n 2 ) The system is Markovian. State: number of components and products in stock: (n 1, n 2 ). Define P(n 2 ) = [P(0, n 2 ) P(1, n 2 ) … P(b 1, n 2 )]. Chapman-Kolmogorov equations: P(0)A 0  P(1)C 0 P(n 2 )A  P(n 2  1)B + P(n 2 +1)C, n 2  1, …, b 2  1 P(b 2 )A 1  P(b 2  1)B 1 where A 0, A 1, A, C 0, C, B 1, and B are matrices that describe the transition rates among the various states (n 1, n 2 ). We solve these equations recursively expressing P(1), P(2), … as functions of P(0). The latter is computed from the last equation and the normalization equation  P (n 1, n 2 ) = 1.

7 Mean profit rate of the system J(b 1, b 2 ) J(b 1, b 2 )  p 2  [production costs in M 1 and M 2 ]  [costs of purchasing from subcontractors in M 1 and M 2 ]  [inventory costs in M 1 and M 2 ]  a function of the equilibrium probabilities Coordination FULL: Perform exhaustive search to track down values for b 1 and b 2 that jointly maximize the mean profit rate of the system. PARTIAL: 1)Factory M 2 determines a base stock b 2 which maximizes the mean profit rate by considering its own costs and profits. Factory M 2 is an M/M/1/b 2 queue. 2)Factory M 1 uses the individually optimal value b 2 to estimate its demand rate and to compute a base stock b 1 which, again, maximizes its own profit rate.

8 Standard parameters:  5,  1   2  6.25, p 1  70, p 2  100, c 1  50, c 2  10, s 1  60, s 2  90, h 1  3, h 2  8 Numerical comparisons

9 Example 2: Single-product system with a base stock s, a base backlog c, and quality control Equivalent closed queueing network: #jobs is m = s + c n 0 = n F + (c - n B )

10 Relationship between the original and closed systems When n F is and n B is Then the total number n H of parts in the original system is and n 0 in the equivalent closed system is 00sc 10sc + 1 ………… s0ss + c = m 01s + 1c - 1 ………… 0cs + c = m0

11 punit profit hunit holding cost rate bunit backlog cost rate i C inspection cost per outgoing item r C rework/rejection cost per nonconforming item Y value of quality characteristic of each outgoing item; random variable t target value of Y q probability that Y is in an acceptable region [t  , t +  ] kquality loss coefficient; we assume a quadratic loss function k(Y  t) 2 Parameters Mean quality cost per outgoing item Q = i C + r C (1  q) + k q E[(Y  t) 2, given that Y is acceptable] Mean profit rate J( , m, s) = pTH  hH  bB  Q TH/q TH = throughput, H = average inventory, B = average backlog m  s + c  base stock + base backlog

12 H  E [n H ]  sP(n 0  c) + (s+1)P(n 0  c  1) +…+ mP(n 0  0) B  E [n B ]  1P(n 0  c  1) + 2P(n 0  c  2) +…+ cP(n 0  0) m  s + c  base stock + base backlog U  [U 0 U 1 … U N ], U  UΠ, Π  [p ij ]  matrix of part-routing probabilities Then: Assumption: The equivalent system is of the Jackson type Let (n 0, n 1, …, n N ) be the vector whose entries are the items in each machine. Then

13 Theorem 1 (a) The function J( , m, s) is concave in s for any fixed ( , m) and assumes its maximum value at the point s which satisfies the following condition (b) If s is optimal for m, then the optimal base stock for m + 1 is either s or s + 1. Theorem 2 For any fixed , the profit rate J is a unimodal function of m for all m  k, where k is the smallest nonnegative integer such that G 0 is the normalizing constant of the closed queueing network with node 0 removed.

14 Optimization For   [  min,  max]; for m = 0, 1, …, m , where m  = local maximizer satisfying condition of Th 2; compute optimal s for ( , m) by applying Th 1, find (  *, m*, s*) which maximize mean profit J We perform exhaustive search for  and m, but m  is finite.z

15 Admission control policies PLS: partly lost sales (proposed policy) CB: complete backlog LS: lost sales Coordination FULL: this strategy seeks values for , m, and s that jointly maximize the mean profit rate of the system (proposed strategy). PARTIAL: the quality control department computes the value , which minimizes the mean quality cost Q per outgoing item. Using this value, the production department computes the probability q of a conforming item. Then, m and s are determined so as to maximize the quantity pTH  hH  bB which is the total profit without quality costs. NO: similar to PARTIAL except that the production department assumes that q = 1, ignoring the possibility of rework or scrap. Test case: A six-machine production line. Numerical comparisons

16

17  Managing inventory levels, sales, and quality tolerances jointly achieves higher profit than independently determined control policies.  Key to the computational efficiency of the optimization algorithms is the adoption of simple control policies and the use of analytical models.  In all the numerical experiments we have performed, the objective functions appear to be quasiconcave (unimodal). We have supported this by a few theoretical results.  Establishing this property is important in order to speed up search for the optimal control parameters, as it will be safe to stop when a locally optimal solution is found. Conclusions