0% found this document useful (0 votes)
14 views

Linear Programming 1

Part 1 of Linear Programming Summary

Uploaded by

Faraj Al-Otaibi
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
14 views

Linear Programming 1

Part 1 of Linear Programming Summary

Uploaded by

Faraj Al-Otaibi
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 17
20 Linear Programming 1: A Geometric Approach A geometric technique for maximizing on min mizing a Linear expression in tuo variables cubject to a eet of Linear conetrainte ie de~ seribed. PREREQUISITESE Linear systeas Linear inequalities IntropucTion ‘The study of Linear prograuming theory has expanded greatly since the pioneer work of George Dantzig in the late nineteen-for- ties. Today, linear programming is applied to # wide variety of problens in industry and science. In this chapter we present a geo- etric approach to the solution of simple linear progranzing prob- lens. In Chapters 21 and 22 we develop the algebraic theory re- quired to solve more general problems in this field. Let us begin with some examples: EXAMPLE 20.1 4 candy manufacture® has 130 pounds of chocolate- covered cherries and 170 pounds of chocolate-covered mints in stock. He decides to sell them in the form of two different mixtures. One 285 286/ A Geometric Apsroach mixture will contain half cherries and half mints and will sell for $2.00 per pound. The other mixture will contain one-third cherries and two-thirds mints and will sell for $1.25 per pound. How many pounds of each mixture shovld the candy manufacturer prepare in or- der to maximize his sales revenue? Let us first formate this problen mathematically. Let the mixture of half cherries and half mints be called mix 4, and Let a be the mumber of pounds of this mixture to be prepared.” Let the mixture of one-third cherries and two-thirds mints be called mix B, and let 22 be the nuaber of pounds of this mixture to be prepared. Since mix 4 sells for $2.00 per pound and mix # sells for $1.25 per pound, the total sales # (in dollars) will be 2 = 2.002, + 1.2529, Sine each pound of mix 4 contains 1/2 pound of cherries and each pound of mix # contains 1/3 pound of cherries, the total number of pounds of cherries used in both mixtures is Similarly, since each pound of mix 4 contains 1/2 pound of mints and each pound of mix 2 contains 2/3 pound of mints, the total number of pounds of mints used in both mixtures is we 392" Because the manufacturer can use at most 130 pounds of cherries and 170 pounds of mints, we aust heve z ste, < 130 deyede, cin. Atso, since a and, camot be negative numbers, we mist have 20 and 20 ‘The problen can therefore be formulated mathonatically as follows: Find values of = and 2, ubich maxintze a= 2,00, + 1.2509 Linear Programming 1 / 287 eubjest to ‘ Ferd, < 130 1 ad 4a +e, < 170 2,20 2,29) Im the next section we shall show how to solve this type of mthoma- tical problem geonetrically. EXAMPLE 20.2 4 wonan has up to $10,000 t0 davest. Her broker suggests investing in two bonds, 4 and B. Bond 4 is a rather risky bond with an annual yield of 104, and bond 5 is a rather safe bond with an annual yield of 7%. After some consideration, she decides to invest at most $6,000 in bond A, at least $2,000 in bond 2, and to invest at least as much in bond A as in bond Z. How should she invest her $10,000 in order to maximize her annual yield? In order to formulate this problem mathematically, let 2 be the number of dollars to be invested in bond 4 and let’#2 be the number of dollars to be invested in bond 8. Since each dollar in- vested in bond A earns $.10 por yoar and each dollar invested in bond 8 earns $.07 per year, the total dollar amount # earned cach year by both bonds is B= 1102, + 072). ‘The constraints imposed can be formulated mathematically as follows Invest no more than $10,004 21 ta) < 10,000 Invest at most $6,000 in bond A: on Invest at least $2,000 in bond 3: 2 Invest at Jeast as mich in bond 4-as in bond B: 222 We also have the implicit assumption that 2, and 2, ere nonnegative: a)20 and zy > ‘Thus, the complete mathonatical formulation of the problem is as follows: Pind values of «, and x, which motmize a= 102, +072, 288/ A Geometric Approach subject to +2, < 10,000 2, < 6,000 2 2,000 w EXAMPLE 20.3 4 student desizes to design a breakfast of corn flakes and milk which is as econowical as possible. On the basis of what he eats during his other meals, he decides that his breakfast should supply hin with at least nine grans of protein, at least one- third the recommended daily allowance (RDA) of vitamin D, and at least one-fourth the RDA of csleium. He Finds the following nutri. tion information on the milk and com flakes containers: { Milk Corn Flakes 3s cup. 2 ounce Cost, | my | Buby weet, | tamame | 2 deans vitamin [1/8 of #04 |1/10 of RDA cotciun | 1/6 of 0A] mone In order not to have his mixture too soggy or too dry, the student decides to limit himself to mixtures which contain one to three ounces of corn flakes per cup of milk, inclusive. What quantities of milk and corn flakes should he use to minimize the cost of his breakfast? For the wathenatical formation of this problem, let 2) be the quantity of milk used measured in ‘-cup units and let #2 be the quantity of corn flakes used measured im l-ounce units. Then if = is the cost of the breakfast in cents, wo may write the following: Cost of breakfast: B= T1Sey + 5.00, At least nine grans of protein: fay +22) >9 At least 1/3 of RDA of vitamin D: Linear Programming I / 289 : 2,24 At least 1/4 of ROA of calcium: yt At least one ounce of corn flakes per cup (two Jecups) of milk: L 2 softy 2 $ (or 2, - 25<0) ‘At most three ounces of comm é Flakes per cup (two J-cups)of aitk:

0) 4s before, we also have the implicit assumption that 2) > 0 and 2220. ‘Thus the complete mathenatical formulation of the problem 48 Bs follows: Pind values of 2, and 2, wktch minimize 7.50, + 5.02) eubject to Geometric SoLuTION OF LINEAR PROGRAMMING PROBLEMS Each of the three examples in the introduction is a special case of the following problen: 290/ A Geometric Approach PROBLEM 20.4 Find values of a, and 2, whieh nasimise ov mintmtze 2 ety $092) subject to ayy aes (CEC) Py apy + agg, (£)(2)(") By Sys *Syate (C29) By In each of the m equations of (20.2), any one of the symbols <, >, = may be used. Problem 17.1 is the gensval Linear programing problem in two variables. The linear function x in (20.1) is called the objective fwetion. Equations (20.2) and (20.3) are called the constraints; in particular, Eqs. (20.3) are called the nonnegativity conetrainte on the variables «1 and 22. We shall now show how to solve a linear programming problem in vwo variables graphically. A pair of values (=, 22) which satisfy all of the constraints is called a feaatble eolution. The set of all feasible solutions determines # subset of the 2yzy-plane called the feaetble region. Our desire is to find a feasible solution which maximizes the objective function, Such a solution is called an optinal solution. ‘To examine the feasible region of a Linear programing problan, let us note that each constraint of the form Serta * S522 Pe defines @ line in the 2j2-plane, while each constraint of the form gy) + p9%y 5 Bg P1* 8092 2b defines » half-plane which includes its boundaty Line Bers * 822 * bes ‘Thus, the feasible region is el- ways an intersection of finitely many Lines and half-planes. For example, the four constraints of Example 20.1 define the half- planes illustrated in Fig. 20.1(2), @), (6), and (a). The feasible region of this problem is thus the intersection of these four half- planes, whieh is illustrated in Fig. 20.1(6). 3 can be shown that the fea~ sible region of @ linear progran- ning problen has a boundary con- sisting of a finite number of straight-line sognents, If the feasibie region can be enclosed in a sufficiently large circle, it is called bounded (Fig. 20.1(6}); otherise it is called inbounded (Fig. 20.5), I£ the feasible re- Bion is omiy (contains no points), then the constraints aze inconsis~ tent and the linear progranning problen has no solution (Fig. 20.6). Those boundary points of @ feasible region which are inter~ sections of two of the straight= Vine boundary segnents are called extvens points. (They are also called comen pointe or vertes pointe.) For example, fron Fig. 20-1(2), the feasible region of Example’ 20.1 has four extrene points: Linear Programing 1 / 291 © (2600) Figure 20.2 292 / A Geometric Approach (0,0), (0, 255), (180, 120), (260, 0) . (20.4) ‘The importance of the extrene points of a feasible region is shown by the following theoren: THEOREM 20.1 If the feasible region of a linear program ming problem ie nonempty and bowled, then the objective function attains both a maximum and mintmen value and these cocur at extvene points of the feasible region. If ion may or may not attain a maximun or minimen values however, if it attaine a mazimen ov minimm value, tt does 0 at an extrone point. Figure 20,2 suggests the idea behind the proof of this theoren. Since the objective function Ba on, Hog, of a linear programming problem is a linear function of 2; and 22, its level curves (the curves along which s has constant values) are straight lines. As we move in a direction perpendicular to these level curves, the objective function either increases or decreases monotonically. Within a bounded feasible region, the naximun and minimun values of 2 mst therefore occur at extreme points, as Figure 20.2 indicates. Figure 20.2 Linear Programming 1/293 In the next few examples, we use Theoren 20.1 to solve several Linear programming problens and illustrate the variations in the nature of the solutions which my occur. EXAMPLE 20.1 (REVISITED) From Figure 20.1(6), wo seo that the feasible rogion of Example 20.1 is hounded. Consequently, from ‘Theoren 20.1 the objective function 5 21000, +128, avtsins both its minimm and maximm values at extreme points. The four extreme points and the corresponding values of 2 are given in the following table Extreme point Value of @ 2) a= 2.00, + 1,282) GO 0 (0, 255) 318.75 (80; 120) 510.00 (260, 0) 520.00 We see that the largest value of is-$20.00, and the corresponding optinal solution is (260, 0). Thus, the candy manufacturer attains maximum sales of $520 when he produces 260 pounds of mixture A and none of mixture 2. EXAMPLE 20.4 sind vatues of =, and sy whlch misinize ae a+, subject to Ie i 294/ A Geometric Approach SOLUTION in Fig, 20.3 we have ray the feasible region of this problem. Since it is bounded, the naximn value of 3 is attained lat one of the five extreme points. The values of the objective function at the five extrene points are given in the following table: Extreme point | Value of (2, ,) Bea, + Se, 8) 18 Go a @2 15 9 0 Fron this table, the maximum value of s is 21, which is attained at a= Sand 2) "6. EXAMPLE 20,5 pind values of o) and sty which maxinize B= Any + 6, Linear Programming 1 / 295 subject to dey + Bey < 24 7 2) <6 2,20 SOLUTION the constraints in this problem are identical to the con straints in Exaxple 20.4, and so the feasible region of this problem is also given by Fig. 20.5. the values of the objective function at the extreme points are as Follows: Extrene point | Velue of (prey) | ede, + bey (0, 6) 36 6,6) 48 2) 46 (7,0) 28 (0,0) ° Wo soe that the objective function attains @ maximum value of 48 at ‘to adjacent extrene points, (5, 6) and (9, 2). This shows that an optinal solution to a linear programing problen need not be unique. AS we ask the reader to show in Exercise 20.9, if the objective Function has the sane value at two adjacent extrone points, it has the sane valuo at all points on the straight-line boundary” segment connecting the two extreue points. Thus, in this example the maxi- mum value of 2 is attained at ali points’ on the straight-Line seg- nent connecting the extrene points (5, 6) and (9, 2). EXAMPLE 20.6 ind vatues of =, end ey which minimize ae 22 subject to 296 / A Geometric Approach SOLUTION rq Fig. 20.4 we have drawn the feasible region of this problem. Because one of the constraints is an equality constraint, the feasible rogion is a straight Line segment with two extreme points. The values of 3 at the two extreme points are as follows: Extrene point | Value of Gye) [a= 2s, -2, 2 (6,0) 2 ‘The minimum value of = is thus 4 and is attained at 2, =5 and Figure 2004 Linear Programming 1/ 297 EXAMPLE 20.7 ping yatuss of oy and sy which maxinize 5 20,4 Se subject to SOLITON sa esciie eptia(oe nae Uioeed estgentiog potion te Oy inectas Wa ales ets wjntarst itctee cetain hase tions a= 2c, +52, can be made arbitrarily large and positive. This probles has no op- timal solution. Instead, we say the problem has an wibounded solu tion. Figure 20.5 298 / A Geometric Averoach EXAMPLE 20:8 Find values of =, and =) which maximize sess, subject to a 2 SOLUTION the above constraints are the sate 2s those in Example 20.7, so that the feasible region of this problem is also given by Fig. 20.5. In Exercise 20.10, we ask the reader to show that the objective function of this problem attains a maximum within the fea- sible region. By Theoren 20.1, this maximum must be attained at an extreme Point. The values of 3 at the two extrene points of the feasible rogion are given by Extrene point | Value of (ey) 252 +2, a, 6) 1 G2) as ‘The maximm value of 3 is thus 1 and is attained at the extreme point <,*1, 2)= 6. ExAWPLE 20.9 sina ehe vatues ©, and 2, which minimize 1 subject to Linear Programing 1/299 setts, < 55 Bayt de, > 24 SOLUTION As can be seen from Fig. 20.6, the intersection of the five half-planes defined by the Five constraints is empty. This Linear programming problen has no feasible solutions since the con~ straints are inconsistant. Figure 20.6 EXERCISES 20.4 Find values of 2, and #, which aaximize B= Say +22, subject to 300/ A Geometric Approach 20.2 Find values of «, and 2, which minimize 5 = Sey- Sey subject to 20.3 Find values of a and subject to 20.4 Solve the Linear programming problon posed in Example 20, 20.5 Solve the Linear programming problem posed in Example 20 Linear Programming 1 / 304 20.6 A trucking firm ships the containers of two companies, 4 and B. Bach container of Company 4 weighs 40 pounds and is 2 cubie fect in volume. Each container of Company 2 weighs 50 pounds and is 3 cubic feet in volune. The trucking firm charges Com- pany 4 $2.20 for each container shipped and charges Company B $3.00 for each container shipped. If one of the firm's trucks canmot carry more than 37,000 pounds and cannot hold more than 2000 cubic feet, how many containers from companies and should a truck carry to maximize the shipping charges? 20.7 Repoat Exerciso 20.6 if the trucking firm raises its price for shipping a container of Conpany 4 t0 $2.50. 20.8 A manufacturer produces sacks of chicken feed fron two ingre- dients, 4 and 8. Each sack is to contain at least 10 ounces of nutrient , et least 8 ounces of nutrient 2, and at least 12 ounces of futrient 3, Each pound of ingredient 4 contains 2 ounces of nutrient iy, 2 ounces of mutrient M7, and 6 ounces Of nutrient Ws. Each pound of ingredient 2 contains 5 ounces of nutrient Nj. 3 ounces of nutrient #1, and 4 ounces of nutri- ent 13. If ingredient 4 costs 8¢ per pound and ingredient 3 costs 9¢ per pound, how much of each ingredient should the man- ufacturer use in each sack of feed to minimize his costs? 20.9 If the objective function of a Linear programming problem has ‘the same Value at tno adjacent extreme points, show that it hes the sane value at all points on the straight-line segment con- necting the two extrene points. Hint: if (#), 1) end (et, #3) are any two points in the plane, a point (#,,,) Lies on the stralght-Line segment connecting them if = tel + - Dat and = t+ 2ey where ¢ is a nunber in the interval [0,1]. 20.10 Show that the objective function in Example 20.8 attains a max- dmua value in the feasible set. Hint: Examine the level curves of the objective function.

You might also like