LPoverall

1
T.T. Narendran
Department of Management Studies Indian Institute of Technology Madras
TTN DoMS, IIT Madras, 3-Sep-09
Example Problem No. 1

(Linear Programming - Formulation) ( )
A small factory makes three products, soap, shampoo q p p g g and liquid soap. The processing is done in two stages at two plants, I and II, each of which works for 40 hours every week. h k The time required for processing each product at each plant and the profit of each product are indicated below
Plant I Soap Shampoo Liquid Soap Hours / week 3 4 4 40 hrs
Plant II 5 4 2 40 hrs
Profit / Kg Rs.10 Rs. 13 Rs. 12
Solution to Example Problem No 1 No.1

Let X1, X2, and X3 be the quantities of soap, shampoo and liquid soap manufactured per week. week The problem can now be modeled as follows
Maximize
Z = 10X1 +13X2 + 12X3 Objective Function
Subject to 3X1 + 4X2 + 4X3 < 40 5X1 + 4X2 + 2X3 < 40 X1 > 0, X2 > 0, X3 > 0 X 1, X 2 , X 3
Constraints Non-Negativity Constraint Decision Variables
What we observe in this mathematical statement of the problem is

(i) There is an objective function which describes what we desire (ii) A set of constraints, usually i th f t f t i t ll in the form of f inequalities (iii) Non negativity constraints for the decision Non-negativity variables (iv) The objective function and the constraints are linear.
So the general form of the L.P is g

Maximize Z = C1 X1 + C2 X2 +..+ Cn Xn Subject to a11 X1 + a12 X2 + + a1n Xn < b1 a21 X1 + a22 X2 + + a2n Xn < b2 an1 X1 + an2 X2 + + amn Xn < bm X1 > 0 X2 > 0 0, 0,.., Xn > 0 (Or)
. .
General f G l form of the L.P is: f th L P i
Example Problem No. 2:

A middle-aged woman, convalescing after a surgery, has been advised by her doctor to plan her diet in such a way as to ensure that she gets the prescribed quantities of vitamins. The h i Th choice of f d available t h f foods il bl to her, th amounts of the t f vitamins they contain, the cost of each food and the daily requirement of vitamins are indicated below
10
Cost
Vitamin A
Vitamin B
Vitamin D
Eggs Milk Cereals Daily requirement
Rs 1.50/unit 15 mg/unit 40 mg/unit 10 mg/unit Rs. 14.00/lit Rs. 18/unit (kg) 30 mg/lit 10 mg/kg 100 mg 25 mg/lit 55 mg/kg 250mg 20 mg/lit 30 mg/kg 120mg
11
Formulate
linear
programme
to
determine th quantity of each f d th t d t i the tit f h food that the woman needs to buy in order to minimize her total expenditure, ensuring at the same time that she meets her daily d il requirements of vitamins. i t f it i
12
Solution to Example Problem No.2 p

Let X1, X2, X3 be the quantities of eggs(in nos.), milk (i lit ) and cereals(in k ) b ilk (in lit.) d l (i kgs.) bought. ht The problem is of the form
13
Example Problem No.3

A senior executive of a company recently quit his job under the VRS with a hefty packet of Rs 1 crore. Messrs. Dhanachor Chit F d C Fund Company h has offered th ff d the following o o g investment es e sc e e scheme for o the e
benefit of such retired people:

14
"Invest a certain sum (in lakhs of rupees) in any month, invest half of that amount in the next month and you will get twice the amount invested originally in the first month". This scheme is available for the next six months months. The returns received at the end of any month can be used immediately for reinvesting either as a fresh investment or as a follow-up investment follow up investment.
15

h Let Xt = Fresh investment in the tth month
(t = 1 2 3 4 5) (in lakhs of Rs ) 1,2,3,4,5) Rs.) Noting that, in any month, the cash outflow g y must not exceed the cash on hand, the following linear programme is formulated ;
16
17
At the end of the 6 month, the cash on hand must be a maximum. Hence the objective function is
18
Example Problem No.4:

A brewery blends three raw materials, "H. P. Solution", "B.C. powder" and varnish in varying proportions to obtain three final products, Arrack products "Arrack" , "Pattai" and "Sarayam". Pattai Sarayam The salient data are shown below.
19
Selling Price / Litre
Ingredient
Cost per litre
Quantity available (litres)
Arrack Pattai Sarayam
Rs. 15 Rs. 9 Rs. 12
HP Solution BC powder Varnish
Rs. 10 Rs. 8 Rs. 7
10,000 12,000 15,000
20
Besides, the following specifications on the proportions of ingredients must be met: ti fi di t tb t
Formulate a linear programme to maximize the total revenue of the brewery brewery.
21
Solution to Example Problem No.4

Let Xij be the quantity of the ith raw material i jth th product t i l in the d t
Constraints on raw materials available :

HP Solution BC Powder Varnish
22
Constraints on raw material proportions : p p
23
Objective function must maximize the net revenue:
24
Example 5:
A shipping company has to lift three types of cargo whose details are given below.
Type of Cargo A B C Quantity (tons) 300 500 400 Volume per ton 1.2 m3 0.8 0 8 m3 1 m3 Profit per ton Rs. 10,000 Rs. 7,000 Rs 7 000 Rs. 8,000
25
26
Weight in aft, center, forward must be in the same proportion as the holding capacities by weight, i.e., weight i e 100 : 200 : 75 Formulate a linear programme to maximize the revenues of the shipping company.
27

Let Xij be the quantity (in tons) of the ith cargo in the jth portion of ship. The linear programme is formulated as follows:
28
Constraints on tonnage capacity :
Constraints on holding space:
29
Constraints on proportion of loads in the three holds :
Example Problem No. 6 Trim - Loss Problem:

A roll of paper, 215 cms. wide, has to be cut according to p p g the following widths ( in cms.) for the requirements expressed in terms of the lengths in metres metres.
35 cm X 90 m 60 cm X 90 m 64 cm X 180 m
30
Paper strips can be cut and pasted lengthwise but not widthwise. widthwise
31
Formulate a linear programme to minimize the trim losses incurred while cutting and pasting the q g p p required lengths of paper.
32

This problem requires nominal decision variables -each variable denotes the length, in metres, to which the roll of paper is cut according to pattern j. There are ten possible patterns which have a Trim Loss less than 35 cms. These are listed below
33
35 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 0 0 0 1 2 4 4 2 2 6
60 0 1 2 3 2 1 0 0 1 0
64 3 2 1 0 0 0 1 2 1 0
Loss 23 27 31 0 25 15 11 17 21 5
34
Let Xj be the length, in meters, of the jth pattern cut out from the given roll of paper. t tf th i ll f
Minimizing this trim loss is equivalent to the following objective :
35
Example Problem No.7: p

Messrs. Messrs Hife and Daw are two scientists who conducted a series of experiments to record the values of the output . parameter i for a temperature setting ti. They Th expect th relationship t b of th f t the l ti hi to be f the form
36
They have concluded that the coefficients ey a e co c uded a e coe c e s have to add up to 1. They wish to determine the values of the coefficients such th t th sum of th modulus of th h that the f the d l f the
.
errors is a minimum, i.e., they wish to , , y minimize
Formulate a linear programme for this purpose

37

L.P - decision variables are the coefficients a b ,c and d a, c d.
Minimize total absolute error in calculating a,b,c a b c and d d. i.e., Minimize the sum of the moduli of error
38
We have to minimize
39
We introduce a term yi which is designed to take the positive difference between and
The problem is now formulated as follows :
Example Problem No. 8 :

(Manipulation of a problem, which is not inherently linear)
40
A final product is assembled with 4 units of component A and 3 units of component B. The man fact ring shop r ns three different processes manufacturing runs processes, each of which requires varying amounts of raw materials and produce different amounts of A and B B. Two types of raw materials are used. 100 units of raw material I (RM I) and 200 units of raw material II (RM II) t i l d it f t i l are available to the shop each day
41
The following table gives the information on the quantities of raw materials consumed by each process and the yield of A and B from each of them
42
Shop
Input per run (units) RM I RM II 5 8 7 200
Output per run (units) A 6 5 7 B 4 8 3
No. of runs X1 X2 X3
I II III
7 4 2 100
Formulate a linear programme to maximize the number of completed assemblies produced each day day.
43

Let X1 X2 X3 be the number of runs of X1, X2, each process operated per day Total quantity of A produced = q y p
Total quantity of B produced =
44
No. of units of finished product p
45
We have to maximize the above expression. This is Thi i accomplished b f li h d by formulating the l i h problem as follows :
46
T.T. Narendran
Department of Management Studies Indian Institute of Technology Madras
47
Example Problem No. 8 :

A furniture shop manufactures tables and chairs. The operations take place
sequentially in two work centers. The associated profits and the man hours required by each product at each work center q y p are shown in the table below :
48
Furniture Shop
Tables
Chairs
Available man hours
Profit / unit Work Center-I Work Center-II
8 4 2
6 2 4 60 48
Formulate a linear programme to determine the p g optimal number of tables and chairs to be p manufactured so as to maximize the profit.
49

The following linear programme is formulated
This problem can be solved graphically. The f ll i figure shows th constraints ; Th following fi h the t i t
50
The shaded region is the feasible solution set to the problem
51
Before we indicate the procedure to solve this problem, a few concepts are introduced. introduced
52
Convex Set:
If there exists a set in which a straight line joining any two points in the set is also contained in the set, then such a set is called a convex set
53
The extreme of a convex set will always lie e e e e o co e se a ays e at a corner point. This property is used for determining the solution to the given linear programme. The expression
54
A series of parallel lines for various assumed values of Z can be drawn. These are called Isoprofit lines. The profit Z along each line is the same. The value of Z is seen to increase as the isoprofit lines move farther away from the origin. The Th corner point th i t through which th l t i h hi h the last isoprofit fit line passes, gives the optimal solution. In the figure OABC, the corner points are evaluated as follows:
Iso Profit Line
55
56
Z0 = 0 ZA = 72 ZB = 132 ZC = 120 For the given problem, the optimal solution g p p occurs at corner point B (12,6) i.e., i e X1 = 12 X2 = 6 with the corresponding profit 12, of 132 which is the maximum.
57
Note: The graphical method, obviously, can g y solve problems with just two variables. The need, however, is for a method that can solve problems of realistic size. p For this purpose, there exists a method called the Simplex Algorithm developed by George.B. Dantzig.
58
A few basic concepts must be learnt before the algorithm is introduced.

Basic Solution: Consider the following set of equations :
59
There are four variables and two equations. Hence this thi cannot b solved as simultaneous equations. t be l d i lt ti However, it is possible to find a set of solutions called basic solutions which are obtained as follows In the given system of equations if we set the 'extra' equations, extra variables = 0, we can solve for the remaining variables. i bl
60
For instance, if we set o s a ce, e se We get This is called a basic solution solution. We can see that there are as many as solutions. . Of these, the solutions that also satisfy the , y non-negativity condition are called basic feasible solutions
61
In general if we have a system of m general, equations with n variables (where n > m), we can obtain basic solutions by setting (n - m) variables = 0 and solving for the remaining variables. i i i bl There will be basic solutions
62
Canonical form :
If there exists a system of equations such that each equation h one variable with coefficient h ti has i bl ith ffi i t of 1 in that equation and a coefficient of 0 in all the other equations, such a system is said to be in Canonical form form. The given system is seen to be in Canonical form since satisfy this condition.
63
Other Definitions :
The variables that are set = 0 are called non basic variables while the remaining variables are called basic variables. Now let us consider the same example:
64
Remove the inequalities and rewrite the constraints as equations b i i by introducing ''slack variables'' are d i l k i bl shown below
This is a system of two equations with four variables. Choose the variables in canonical form, i.e., as the basic variables for the initial solution. Rewrite the equations, expressing the basic variables in terms of non-basic variables variables.
65
Both the non basic variables have positive coefficients in the objective function and hence have the potential to increase the value of Z . l f
66
Let us now consider making X1 a basic variable. , g To do this, one of the existing basic variables must become non basic since there can be only two basic variables at any stage. i bl t t To find out which variable is to be replaced, we find the maximum possible value for X1 in equations (1) and (2) (2).
67
68
Of these, only the lower value will satisfy both these the constraints. Hence
2
69
70
When we examine for the maximum possible value of X2 i (3) and (4) we see that X2 l f in d (4), h displaces X4 as basic variable.
71
Here, Here the coefficients of both the non nonbasic variables in the objective function are negative negative. Hence, there is no further scope for increasing the value of Z. So the final solution is
72
73
Simplex Al Si l Algorithm : Tabular form ith T b l f

The algorithm explained above can be
implemented in a tabular form for greater p g working convenience. This form is also easier for coding and automation purposes. The key features of the tabular form are as follows:
74
Each variable is represented as column. The corresponding coefficients appear in the boxes of the table table. CB = Coefficient of Basic Variable It can be seen that the tabular form is just another way of representing the same set of equations used above
75
76
The so u o a this s age reads as e solution at s stage eads X3 = 60 X4 = 48 Z = 0 60, 48,
77
In the row Cj Zj , the positive coefficients for the non basic variables X1 and X2 indicate that the value of Z will increase if one of these variables becomes a basic variable. We choose X1to become the basic variable. This requires one of the existing basic s equ es o e o t e e st g bas c variables to become non basic. In th I other words, X1i th entering variable. W d is the t i i bl We have to find the departing variable.
78
Now, evaluate the ratio bi / aij for all aij >0 in this column. The ratios are 60 / 4 = 15 and 48 / 2 = 24 The minimum ratio determines the departing variable. Hence, i this case X3 i the d H in hi is h departing variable. i i bl The coefficient 4 corresponding operation is called the pivot. to this
79
The next table is obtained by performing the following operations. In the 'basis' column X1 replaces X3. Divide the pivotal row (4 2 1 0 | 60) by the pivot (4). 15) We get (1 0 |
80
To get the X4 row, do the following : Multiply the M lti l th new row j t obtained b th just bt i d by the coefficient in the pivotal column (2). We get (2 1 0 t | 30) Subtract corresponding elements from the old X4 row. That is
81
This constitutes the new X4 row in the table. The Cj Zj row is evaluated as follows :
Zj = CB aij
For example, Z2 = 8 () + 0 (3) = 4 and C2-Z2 = 6-4 = 2 In this manner, all the values of Cj Zj are computed
82
83
At the end of this iteration, the solution reads as X1 = 15 X4 = 18 Z = 120 15, 18, 120, Since the X2 column in the Cj Zj row shows a positive value, we now bring in X2 as the entering variable. t i i bl Proceeding as before, we find 18 / 3 = 6 is less than 15 / () = 30
84
Therefore, 3 is the pivot and X4 is the departing variable. p g The next table is obtained using the same steps described earlier
85
86
Now th solution reads N the l ti d , , X1 = 12, X2 = 6, Z = 132 The coefficients of the non basic variables in the th Cj Zj row are all non positive. ll iti , p Therefore, there is no further scope for increasing the value of Z. Hence th algorithm stops at thi stage. Th H the l ith t t this t The given solution is the optimal solution
87
The steps involved in the simplex algorithm are given below :

Rewrite inequalities as equations using slack variables. Set up L.P in tabular form LP form. Determine the entering variable as one with a positive Cj Zj coefficient ; Generally, the variable with the maximum Cj Zj is y chosen as the entering variable though this is not a strict criterion.
88
Determine the departing variable as the row p g in which the aij coefficient yields the minimum positive bi / aij (aij >0). 0). The aij corresponding to the departing variable is called the pi ot ariable pivot. In order to obtain the next table, perform the following steps. Divide the pivotal row throughout by the pivot.
89
Multiply this new row by coefficients in the pivotal column and subtract the product from the corresponding old row; Repeat this process till the table is complete. Check if there is any Cj Zj > 0. If so, repeat steps (3) - (7). If not, read off the optimal solution from the last table table.
90
Complications and their resolution: p

There are many deviations from the standard form of the linear programme that was used to illustrate the simplex algorithm. We now consider each of these variations and explain how to work with them.
91
(1) Minimization:
One of the most important variations to be p addressed is the minimization problem Consider the following example :
92
We remove the inequalities and rewrite the constraints using 'surplus variables' X3 and X4 as shown below.
93
Unfortunately, we still do not have a canonical form to get this simplex algorithm started. Therefore, Therefore we introduce one more set of variables called 'artificial variables' to create the required canonical form. However, However we also ensure that the artificial variables will eventually leave the basis by assigning them highly unfavorable coefficient values in the objective function function.
94
The problem is now written as follows:
95
We now proceed to construct the simplex table as p p shown below. The steps are the same except for the following: f ll i (i) The criterion for choosing the entering variable is 'most negative Cj
Zj Zj
(ii) The criterion for optimality is that all Cj coefficients must be > 0 0.
96
97
(2) Constraint of > form in a maximization problem:

Consider the following example:
98
The surplus variable in the second e su p us a ab e e seco d constraint does not give the canonical form. Hence we introduce an artificial variable in this constraint. In the objective function the artificial function, variable has a coefficient -M in order to make it undesirable variable in the final solution.
99
100
3) N Negative bi : ti
101
The negative sign in the R.H.S of constraint (1) posses a problem. Hence it is rewritten, equivalently, as
102
4) E Equation constraint : ti t i t
Consider the following example
103
When a constraint is in equation form, we e co s a s equa o o , e cannot add a slack variable.
Either manipulate the existing constraints to obtain a canonical form or simply introduce or, as artificial variable as shown below:
104
105
(5) V i bl unrestricted in sign: Variable t i t di i
106
Let X2 = X3 X4, and X3 > 0, X4 > 0. e a d 0 Now rewrite the problem as follows :
107
(6) Ti for entering variable : Tie f t i i bl
108
Tie can be broken arbitrarily. y

109
(7) Ti f d Tie for departing variable: ti i bl
110
111
X2 is the entering variable. Both 6 and 9 can be pivot since 24 / 6 = 36 / 9
Choosing 6 as pivot (arbitrarily) we iterate and get the next table:
112
113
Here, Here the optimal solution exists but a basic variable has take a zero value. Such a solution is called a degenerate basic feasible solution. Degeneracy occurs whenever we encounter D h t a tie for the departing variable.
114
Consider the graphical solution to the same problem:
115
We can see that Constraint 2 is totally redundant redundant. Constraint 3 is redundant but it touches one corner point point.
116
Degeneracy occurs when:

More than 2 constraints pass through a corner point [Constraints i l d non-negativity conditions also]. [C t i t include ti it diti l ] The practical implication is that more than one resource can be completely utilized at a particular corner point solution, leaving no slack at all. Note: Some times, during an iteration, a degenerate so u o solution may appear & then vanish in a subsequent ay appea e a s subseque iteration
117
Try this out in the following example:
118
(8) Alternative optimal solution:
119
Though we have reached the optimal solution, we see that Cj Zj = 0 for a non-basic variable. This is unusual. If we force X1 as an entering variable and iterate further, we get the following solution
120
121
The solution is different but the value of Z is the same. Plotting the constraints on a graph we find graph, that any point between (0,3) and (7/3, 7/3) is optimal; this is not indicated by 'Simplex' which Simplex gives only corner points.
122
123
In general alternative optimal solution is general, detected whenever for a non basic variable in the optimal simplex variable. We can iterate and find another corner point solution that is optimal. Any point on the line joining those two corner points also gives th optimum. i t l i the ti
124
(9) Infeasibility: I f ibilit

Consider the following example:-
125
Simplex Table for the problem:
126
After one iteration, we get a table in which , g the optimality condition seems to be satisfied, i.e., satisfied i e all Cj Zj < 0 . However, since the artificial variable is still there in the basis, no feasible solution exists. exists Consider the graphical solution to the problem:
127
128
C ea y, eas b e eg o Clearly, a feasible region does not exist. o e s In general, the presence of one or more artificial variable in the final table confirms infeasibility.
129
(10) Unbounded solution: U b d d l ti

Consider the example:
130
The following Simplex table is Th f ll i Si l t bl i constructed: t t d
131
After the first iteration we find that there iteration, is scope to improve the value of Z but
we are unable to find a pivot since there is no aij > 0 the entering variable column. When we plot the constraint, constraint the
following graph is obtained:
132
133
We see that there is no feasible finite optimum since the solution space is unbounded and Z p g y keeps on increasing indefinitely. Generally, the inability to find a pivot indicates unbounded solution solution. But unbounded solution space does not mean infeasibility. i f ibilit Besides, there are problems with unbounded solution space and a finite solution. p Most minimization problems can be seen to be of this kind.
134
Note: In rare cases, a situation like this may indicate infeasibility. While unbounded ness nbo nded true. non-existence non e istence of a pivot, the converse is not always
135
Exercise Problems: No.1 E i P bl N 1
136
137
138
139
Thank You !!!

LPoverall

Uploaded by

Copyright:

Available Formats

LPoverall

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

LPoverall

Uploaded by

Copyright:

Available Formats

1

TTN DoMS, IIT Madras, 3-Sep-09

Example Problem No. 1

Plant I Soap Shampoo Liquid Soap Hours / week 3 4 4 40 hrs

Profit / Kg Rs.10 Rs. 13 Rs. 12

TTN DoMS, IIT Madras, 3-Sep-09

Solution to Example Problem No 1 No.1

TTN DoMS, IIT Madras, 3-Sep-09

Constraints Non-Negativity Constraint Decision Variables

TTN DoMS, IIT Madras, 3-Sep-09

What we observe in this mathematical statement of the problem is

TTN DoMS, IIT Madras, 3-Sep-09

So the general form of the L.P is g

General f G l form of the L.P is: f th L P i

TTN DoMS, IIT Madras, 3-Sep-09

Example Problem No. 2:

TTN DoMS, IIT Madras, 3-Sep-09

Eggs Milk Cereals Daily requirement

TTN DoMS, IIT Madras, 3-Sep-09

TTN DoMS, IIT Madras, 3-Sep-09

Solution to Example Problem No.2 p

TTN DoMS, IIT Madras, 3-Sep-09

Example Problem No.3

benefit of such retired people:

TTN DoMS, IIT Madras, 3-Sep-09

Solution to Example Problem No 3 No.3

TTN DoMS, IIT Madras, 3-Sep-09

TTN DoMS, IIT Madras, 3-Sep-09

TTN DoMS, IIT Madras, 3-Sep-09

Example Problem No.4:

TTN DoMS, IIT Madras, 3-Sep-09

Selling Price / Litre

Cost per litre

Quantity available (litres)

Arrack Pattai Sarayam

Rs. 15 Rs. 9 Rs. 12

HP Solution BC powder Varnish

Rs. 10 Rs. 8 Rs. 7

10,000 12,000 15,000

TTN DoMS, IIT Madras, 3-Sep-09

Besides, the following specifications on the proportions of ingredients must be met: ti fi di t tb t

Solution to Example Problem No.4

Constraints on raw materials available :

Constraints on raw material proportions : p p

TTN DoMS, IIT Madras, 3-Sep-09

Objective function must maximize the net revenue:

TTN DoMS, IIT Madras, 3-Sep-09

TTN DoMS, IIT Madras, 3-Sep-09

TTN DoMS, IIT Madras, 3-Sep-09

TTN DoMS, IIT Madras, 3-Sep-09

Solution to Example Problem No.5

TTN DoMS, IIT Madras, 3-Sep-09

Constraints on tonnage capacity :

Constraints on holding space:

TTN DoMS, IIT Madras, 3-Sep-09

Constraints on proportion of loads in the three holds :

TTN DoMS, IIT Madras, 3-Sep-09

Example Problem No. 6 Trim - Loss Problem:

TTN DoMS, IIT Madras, 3-Sep-09

Solution to Example Problem No.6