A Note on the Linear Programming Sensitivity. Analysis of Specification Constraints. in Blending Problems

Similar documents
A New Method for Solving Integer Linear. Programming Problems with Fuzzy Variables

SIMULATION OF THREE PHASE THREE LEG TRANSFORMER BEHAVIOR UNDER DIFFERENT VOLTAGE SAG TYPES

Lucas Imperfect Information Model

Wp/Lmin. Wn/Lmin 2.5V

COLUMN GENERATION HEURISTICS FOR SPLIT PICKUP AND DELIVERY VEHICLE ROUTING PROBLEM FOR INTERNATIONAL CRUDE OIL TRANSPORTATION

Section 3: Detailed Solutions of Word Problems Unit 1: Solving Word Problems by Modeling with Formulas

Physic 231 Lecture 33

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

Circuits Op-Amp. Interaction of Circuit Elements. Quick Check How does closing the switch affect V o and I o?

_J _J J J J J J J J _. 7 particles in the blue state; 3 particles in the red state: 720 configurations _J J J _J J J J J J J J _

Thermodynamics of Materials

Chapter 7. Systems 7.1 INTRODUCTION 7.2 MATHEMATICAL MODELING OF LIQUID LEVEL SYSTEMS. Steady State Flow. A. Bazoune

V. Electrostatics Lecture 27a: Diffuse charge at electrodes

Chapter 6 : Gibbs Free Energy

Linear Plus Linear Fractional Capacitated Transportation Problem with Restricted Flow

Math1110 (Spring 2009) Prelim 3 - Solutions

Approach: (Equilibrium) TD analysis, i.e., conservation eqns., state equations Issues: how to deal with

element k Using FEM to Solve Truss Problems

55:041 Electronic Circuits

Phys 344 Ch 5 Lect 4 Feb 28 th,

Feedback Principle :-

Chapter 2 GAUSS LAW Recommended Problems:

Thermodynamics Partial Outline of Topics

A Note on Equivalences in Measuring Returns to Scale

Chapter 3, Solution 1C.

A method of constructing rock-analysis diagrams a statistical basks.

Water vapour balance in a building moisture exposure for timber structures

Lecture 12. Heat Exchangers. Heat Exchangers Chee 318 1

Design of Analog Integrated Circuits

CHAPTER 3 ANALYSIS OF KY BOOST CONVERTER

CONVEX COMBINATIONS OF ANALYTIC FUNCTIONS

Inference in Simple Regression

Determining the Accuracy of Modal Parameter Estimation Methods

The internal structure of natural numbers and one method for the definition of large prime numbers

CIRCLE YOUR DIVISION: Div. 1 (9:30 am) Div. 2 (11:30 am) Div. 3 (2:30 pm) Prof. Ruan Prof. Naik Mr. Singh

Conduction Heat Transfer

Big Data Analytics! Special Topics for Computer Science CSE CSE Mar 31

Physics 2B Chapter 23 Notes - Faraday s Law & Inductors Spring 2018

Online Classification: Perceptron and Winnow

Reproducing kernel Hilbert spaces. Nuno Vasconcelos ECE Department, UCSD

Chem 204A, Fall 2004, Mid-term (II)

Naïve Bayes Classifier

Physics 107 HOMEWORK ASSIGNMENT #20

Conservation of Energy

Problem Set 5 Solutions - McQuarrie Problems 3.20 MIT Dr. Anton Van Der Ven

Module B3. VLoad = = V S V LN

Fundamentals of Finite Elements. Mehrdad Negahban. W311 Nebraska Hall Department of Engineering Mechanics University of Nebraska-Lincoln

Optimization of frequency quantization. VN Tibabishev. Keywords: optimization, sampling frequency, the substitution frequencies.

Transient Conduction: Spatial Effects and the Role of Analytical Solutions

Exercises H /OOA> f Wo AJoTHS l^»-l S. m^ttrt /A/ ?C,0&L6M5 INFERENCE FOR DISTRIBUTIONS OF CATEGORICAL DATA. tts^e&n tai-ns 5 2%-cas-hews^, 27%

CTN 2/23/16. EE 247B/ME 218: Introduction to MEMS Design Lecture 11m2: Mechanics of Materials. Copyright 2016 Regents of the University of California

Exploiting vector space properties for the global optimization of process networks

Interference is when two (or more) sets of waves meet and combine to produce a new pattern.

Section 5.8 Notes Page Exponential Growth and Decay Models; Newton s Law

Mingqing Xing 1 School of Economics and Management, Weifang University, Weifang ,

CHAPTER 3 INEQUALITIES. Copyright -The Institute of Chartered Accountants of India

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

Section 10 Regression with Stochastic Regressors

Analysis The characteristic length of the junction and the Biot number are

ME2142/ME2142E Feedback Control Systems. Modelling of Physical Systems The Transfer Function

Regression with Stochastic Regressors

Global Sensitivity. Tuesday 20 th February, 2018

Problem 1. Refracting Surface (Modified from Pedrotti 2-2)

Spring 2002 Lecture #17

Part One: Heat Changes and Thermochemistry. This aspect of Thermodynamics was dealt with in Chapter 6. (Review)

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

One-sided finite-difference approximations suitable for use with Richardson extrapolation

READING STATECHART DIAGRAMS

Department of Economics, University of California, Davis Ecn 200C Micro Theory Professor Giacomo Bonanno. Insurance Markets

WYSE Academic Challenge 2004 Sectional Physics Solution Set

The Minimum Universal Cost Flow in an Infeasible Flow Network

CAUSAL INFERENCE. Technical Track Session I. Phillippe Leite. The World Bank

A Simple Set of Test Matrices for Eigenvalue Programs*

Logistic regression with one predictor. STK4900/ Lecture 7. Program

BME 5742 Biosystems Modeling and Control

Computational modeling techniques

Chemistry 20 Lesson 11 Electronegativity, Polarity and Shapes

Computational modeling techniques

Shell Stiffness for Diffe ent Modes

Edexcel GCSE Physics

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Chapter 5: Force and Motion I-a

General Chemistry II, Unit I: Study Guide (part I)

4DVAR, according to the name, is a four-dimensional variational method.

State-Space Model Based Generalized Predictive Control for Networked Control Systems

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

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

Statistics Chapter 4

EE 204 Lecture 25 More Examples on Power Factor and the Reactive Power

Unit 11 Solutions- Guided Notes. What are alloys? What is the difference between heterogeneous and homogeneous mixtures?

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

BASIC DIRECT-CURRENT MEASUREMENTS

HANSEN SOLUBILITY PARAMETERS IN CHROMATOGRAPHIC SCIENCES ADAM VOELKEL, K. ADAMSKA POZNAŃ UNIVERSITY OF TECHNOLOGY, POLAND

Solutions to Homework 7, Mathematics 1. 1 x. (arccos x) (arccos x) 1

Dead-beat controller design

Lecture 17: Free Energy of Multi-phase Solutions at Equilibrium

Fall 2010 Analysis of Experimental Measurements B. Eisenstein/rev. S. Errede. (n.b. for now, we do not require that k. vectors as a k 1 matrix: ( )

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

The Order Relation and Trace Inequalities for. Hermitian Operators

Transcription:

Aled Mathematcal Scences, Vl. 2, 2008, n. 5, 241-248 A Nte n the Lnear Prgrammng Senstvty Analyss f Secfcatn Cnstrants n Blendng Prblems Umt Anc Callway Schl f Busness and Accuntancy Wae Frest Unversty, Wnstn-Salem NC, 27109, USA Anc@wfu.edu Abstract In blendng rblems there are tycally secfcatn cnstrants that lmt the cntent f varus rertes f the blend that t acqures frm the ngredents t certan maxmum r mnmum ercentages f the ttal blend. Fr sae f lnearty, these cnstrants are cmmnly ncluded n the rblem n a way that recludes drect senstvty analyss wth resect t changes n these ercentages. Ths nte shws that the senstvty analyss wth resect t changes n the target secfcatn ercentages can be derved frm the rdnary senstvty analyss wth mnmum addtnal effrt; and that cmmn LP cdes can be slghtly mdfed t facltate senstvty analyss f blendng ercentages.

242 Umt Anc Mathematcs Subect Classfcatn: 90C31 Keywrds: Lnear Prgrammng, Blendng Prblems, Senstvty Analyss. Thrugh ts hstry, ne f the mst frequently cted alcatn examles f lnear rgrammng has been the s-called blendng, rblem n whch varus ngredents (nuts) are mxed nt ne r mre blends (ututs) t satsfy sme bectve [1]. Blendng rblems, amng ther tyes f cnstrants, may nclude the secfcatn cnstrants. These cnstrants lmt certan rertes (such as msture) f the blend, whch t nherts frm the ngredents, t certan mnmum r maxmum ercentages f the ttal. Mst cmmn LP sftware utut des nt allw drect senstvty analyss f the target ercentages. The urse f ths nte s t shw hw smlex-based LP sftware may be slghtly mdfed t enable drect and full senstvty analyss f these target ercentages. A tycal such cnstrant, say the th cnstrant f the rblem, may be wrtten generally as: sx / x ( 1) where s s the ercentage f the rerty n questn cntaned n ngredent, and S s the subset f all ngredents whch may be used n blend. By selectng arrate zers and nes fr s, Eq (1) can als mdel a cmmn tye f

Lnear rgrammng senstvty analyss 243 secfcatn cnstrant ne whch lmts the ercentage f a certan subset f ngredents n the blend. Snce ths frm s nt lnear and thus unsutable t lnear rgrammng t s equvalently ncluded n the LP as: s x x 0 (2) Hwever ths transfrmatn recasts the senstvty wth resect t the RHS f Eq (1) nt the senstvty wth resect t several technlgcal cnstrants n Eq (2). Ths mre dffcult frm f senstvty has been studed extensvely as art f LP thery. See fr nstance Smnnard [2,. 145]. Hwever, the theretcal results cncernng the technlgcal cnstrants have nt been aled t the senstvty analyss f target ercentages; nr they have been mlemented n the mst cmmnly used LP sftware. Yet n many blendng and mxng rblems ths arameter, may be set as a matter f management lcy and thus effect f adtng dfferent lces n the tmal LP value may be qute a valuable gude. The tmal dual rce fr the mdfed cnstrant, say d, gves the rate f change n the bectve functn as the rght hand sde (RHS) f (2) s erturbed wthn the range n whch the tmal bass and thus d des nt change (allwable range). Let z / r dente ths rate, where z s the bectve functn value and r s the RHS f (2) (currently zer); and r and r dente the allwable ncrease

244 Umt Anc and decrease n the RHS resectvely. Althugh d gves sme useful senstvty nfrmatn ertanng t the current tmal slutn, t des nt drectly answer the mre legtmate questn f the behavr f the tmal slutn fr changes n the arameter, the RHS f (1). The lnear frm (2) wth a RHS f r s s x x r. It can be re-wrtten as s x / x ( r / x) S whch, n turn, means that changng the RHS f Eq (2) by r s equvalent t changng the RHS f (1) by r / x. Therefre we have: S = r / x S () 3 = r / x and S = r / x (4) S z / = ( z / r) x. S (5) Suse that the tmal slutn t the blendng rblem, wth Eq. (2) and ts senstvty analyss nfrmatn s avalable frm a standard LP rgram. Althugh z / r = d s cnstant n the allwable nterval; z / may nt be cnstant, that s z() may be nn-lnear n. The behavr f the tmal slutn vectr and the bectve functn value, as changes can be estmated by smly evaluatng the quantty d x at the current tmal slutn x. Hwever, ths

Lnear rgrammng senstvty analyss 245 wuld nly be arxmate, because as the RHS f Eq (2) changes wthn the allwable range, whle the tmal bass and d stay cnstant, the tmal x and thus z / may nt. The qualty f ths arxmatn deends n the magntude f the change n the quantty x. Whle n sme cases ths change mght be S small and can be gnred, n thers t mght be cnsderable enugh t sgnfcantly dstrt the senstvty results wth resect t arameter,. The standard LP slutn wth Eq (2) hwever, can be used t erfrm a full and exact senstvty analyss f the slutn fr changes n the arameter,. Let x ( r) be the vectr f slutn values as a functn f the change n the RHS f (2), vectr b the rgnal rght hand sdes f the LP wth Eq (2), and B the current tmal bass. We have: 1 x ( r) = B ( b u r ) = x B 1 u r, (6) where u s the th unt vectr. In (6), the term, B 1 u traces ut the th clumn f the tmal bass nverse. Let us dente the elements f ths clumn by β. Therefre, we have: x ( ) = r x r β, where S s the subset f S S S crresndng t the currently basc vectrs. Wth ths nfrmatn the exact lmts f, and are btaned as: = r /( x r β ) and = r /( r ). ( 7 x β ) S

246 Umt Anc Als the exact rate f change n the bectve functn value, as the RHS f Eq. (1) changes, may be wrtten as: z / = z / r ( x r β ). T exress ths rate n terms f rather than r, we can slve r = ( x r β ) fr r and substtute n abve t gve: z / = d x β x 1 β, (8) whch s the average rate f change n z, as changes by an amunt. The tmal value f the LP as changes by wthn the allwable range, can be btaned by multlyng (8) thrugh by and smlfyng z( ) = z d x 1 β. (9) In Eq (9), when β = 1, z( ) becmes undefned. Hwever ths trublesme stuatn des nt ccur, because1 β > 0. T rve ths, assume that β < 0. As changes n the negatve drectn frm 0, t reaches ( 0) befre t becmes / β, that s t say >1/ β. Ths s 1 easly shwn by substtutng r /( x r β ) fr and rearrangng terms t r β get < 1. x β r Ths s true whenever x > 0, snce bth β and r

Lnear rgrammng senstvty analyss 247 are nn-stve. Furthermre, snce 1 / β < we have 1 0 β >. A smlar argument hlds fr the case β > 0. It s als straghtfrward t examne the functnal behavr f z( ) changes n the allwable range. The frst dervatve f z( ) as s d x 1 β ( ) 2 and thus has the same sgn as d. Thus z( ) s nn-decreasng f d 0; nn-ncreasng therwse. The secnd dervatve s 2d β x (1 β ) 3 whch mles that f β = 0, then z( ) s lnear n the allwable nterval; als snce x and 1 β are bth nn-negatve, z( ) s cnvex f d and β have the same sgn, cncave therwse. The abve suggests that standard LP sftware acages can be mdfed slghtly t enable users t cnduct exact senstvty analyses f secfcatn-tye cnstrants. All that s requred, ssbly as a user selectable tn, s t rert the β quanttes crresndng t thse cnstrants that the user has cded as secfcatn-tye durng nut. REFERENCES [1] Gerge Dantzg, Lnear Prgrammng and extensns. Prncetn Unversty Press, Prncetn, N.J., 1963

248 Umt Anc [2] Mchel Smmnard, Lnear Prgrammng. Prentce Hall, Englewd Clffs, N.J., 1966. Receved: June 5, 2007