Goalprogramming Goal programming

Objectivesofthetopic: Obj ti f th t i Indentifyindustrialbasedsituationswheretwoor moreobjectivefunctionsarerequired. WriteamultiobjectivefunctionmodelasagoalLP l b f d l l Useweightingsumandpreemptivemethodstosolve goalLP Seechapter8inthetextbook.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 There are situations in industrial planning that require finding the optimal parameters for two or more objective functions: Reduce labor cost and improve efficiency Increase production and reduce energy usage Goal programming provides a mechanism to solving these situations whose objective functions might be j g conflicting.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Production mix
A company is producing industrial detergents with four grades. The revenues (price cost) of the four grades are
Grad 1 2 3 4 Revenue(SR perm3) 5500 350 550 0.75

The management of the company has decided to collect SR 160,000 in revenues for the next month. But, they asked that the revenue from grade 2 should not exceed 10% of the total revenue. f
DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Also, revenues from grade 3 should not exceed 20% of the total revenue. Also, the production of grade 4 should not exceed 20 m3. The industrial engineer has been asked to find the production mix that will closely satisfy all the requirements of management. Define x1 = m2 of grade 1 produced x2 = m2 of grade 2 produced x3 = m2 of grade 3 produced x4 = m2 of grade 4 produced

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The goals of the management can be expressed mathematically as

5500 x1 + 350 x2 + 550 x3 + 0.75 x4 160,000 350 x2 0.1(5500 x1 + 350 x2 + 550 x3 + 0.75 x4 ) 550 x3 0.2(5500 x1 + 350 x2 + 550 x3 + 0.75 x4 ) x4 20 x1 , x2 , x3 , x4 0

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The goals can be written as

5500 x1 + 350 x2 + 550 x3 + 0.75 x4 160,000 550 x1 315 x2 + 55 x3 + 0.075 x4 0 1100 x1 + 70 x2 440 x3 + 0.15 x4 0 x4 20 x1 , x2 , x3 , x4 0

The industrial engineers finds difficult to meet all goals of the management with a single production mix. Hence, he is looking for a compromise solution that may not necessarily satisfy all the constraints.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The flexible goals can be written as

5500 x1 + 350 x2 + 550 x3 + 0.75 x4 + s1 s1+ = 160,000
+ 550 x1 315 x2 + 55 x3 + 0.075 x4 + s2 s2 = 0 + 1100 x1 + 70 x2 440 x3 + 0.15 x4 + s3 s3 = 0 + x4 + s4 s4 = 20

x1 , x2 , x3 , x4 0 s , s+ 0

The variable s- and s+ are called deviational variables, and by their definition, only one of them, not both, can be basic (> 0). The deviational variables s- and s+ resemble the slack and surplus variables in their interpretation.
DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 For example, for the first goal, if s+ is > 0, then the constraints left side is greater than the right side, and s+ merely absorbs the excess to make the left side equal the right side. When s- is > 0 then the left side is less than the right 0, side, and s- adds to the left side to make the two sides equal. Therefore, any solution for which s+ 0, the solution is feasible for the first constraint. The violation of the constraint is measured by s+, which should be minimized.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 A compromise solution to the production mix will be to minimize the deviations: minimize G1 = s-1 minimize G2 = s-2 minimize G3 = s-3 minimize G4 = s+4

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Goal programming algorithms

Two algorithm are discussed: 1. Weighted sum method g 2. Preemptive method Both methods represent the goal programming problem with a single objective optimization problem. The two methods do not generally produce the same solution because of the subjectivity involved involved.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Weighted sum method

The weighted sum method assigns weights to the deviations and minimizes the sum of the weighted deviations. The weight assigned to a specific deviation signifies the importance of that deviation and how important to reduce it in comparison with the other deviations. Assume there are n goals: minimize Gi, i = 1,, n Then, the combined objective function is defined as minimize z = w1 G1 + w2 G2 + + wn Gn wi > 0 is the weight given to goal i i.
DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Example
In a single-stage assembly line, there are 10 workers available. The plant produces two models of a product. The revenues, storage space requirements, and worker utilization per model are as:
Model1 SR4 5m2 1worker Model2 SR8 24m2 2workers

Revenue(x100) Storagespacerequirement Workerutilization

The available storage space is 100 m2.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Because of the nature of the demand, production of model 1 should not exceed 6 parts. It has been decided that the total revenue is not to be less than SR 4,500. The industrial engineer has to plan the assembly that will increase the companys revenue and reduce inventory, since it is a common storage.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Define x1 : units of model 1 assembled x2 : units of model 2 assembled The production mix problem can be written in math as
4 x1 + 8 x2 45 8 x1 + 24 x2 100 x1 + 2 x2 10 x1 6 x0

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Add deviational variables to both goals:

4 x1 + 8 x2 + s1 s1+ = 45
+ 8 x1 + 24 x2 + s2 s2 = 100

x1 + 2 x2 10 x1 6 x, s , s + 0

Through extensive meetings market studies and meetings, studies, inventory size, it has been decided that the total revenue is twice as important as overstocking.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Hence, the objective function is

min s.t.
+ z = 2 s1 + s2

4 x1 + 8 x2 + s1 s1+ = 45
+ 8 x1 + 24 x2 + s2 s2 = 100

x1 + 2 x2 10 x1 6 x, s , s + 0

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The problem is then solved using the simplex method:

min i s.t. z = 2 x3 + x6 + 100 x7 + 100 x8 4 x1 + 8 x2 + x3 x4 8 x1 + 24 x2 + x5 x6 x1 + 2 x2 x1 x0 + x7 + x8 + x9 + x10 = 45 = 100 = 10 =6

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The optimal solution is z = 10 x1 = 6 x2 = 2 s-1 = 5 s-2 = 4 rest of the variables = 0 t f th i bl This is to say, with this production mix, the revenue will be in short by SR 500 and storage utilization will be less than the maximum preferred by 4 m2.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Notice that there are other solutions for which z =10: s-1 = 4 s+2 = 2 x1 = 4.5 x2 = 2.75 H Hence, we call th solutions ffi i t solutions rather ll the l ti efficient l ti th than optimal solutions.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Preemptive method
The modeler might have a preference for the most important goal and the least important goal. With the preemptive method, the goals are ranked in the order of importance, so that any solution will not satisfy low priority goals on the expense of high priority goals goals. Given n goals, the objectives of the problem are Minimize G1 (highest p ( g priority) y) Minimize Gn (lowest priority) The method considers one goal at a time, starting with the highest priority, and the process is carried out such that the solution obtained from a lower priority goal never degrades any higher-priority solutions.
DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 There are two ways to solve a goal problem using preemptive method: Successive addition of constraints Goal simplex

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Successive addition of constraints

The successive addition of constraints works best if implemented in an optimization software. The method works by solving for highest priority goal first, then the value of the goal is put as a restriction (constraint) for the solution of the problem for the second-highest priority goal, and so on. Take the general goal model with the objective function being the highest priority goal: Min G1 s.t. constraints t t i t Let the solution be G*1.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Then, the second problem is to solve Min G2 s.t. G1 G*1 constraints

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Example
Consider the production mix problem:
i min G1 = s1
+ min G2 = s2

s.t. st

4 x1 + 8 x2 + s1 s1+ = 45
+ 8 x1 + 24 x2 + s2 s2 = 100

x1 + 2 x2 10 x1 6 x0

Assume that meeting the revenue goal is more important than the inventory goal.
DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The first LP that has to be solved is

min G1 = s1 s.t. 4 x1 + 8 x2 + s1 s1+ = 45
+ 8 x1 + 24 x2 + s2 s2 = 100

x1 + 2 x2 10 x1 6 x0

A solution to this LP is G*1 = 5 5.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The second LP that has to be solved is

+ min G2 = s2 i

s.t.

s1 5 4 x1 + 8 x2 + s1 s1+ = 45
+ 8 x1 + 24 x2 + s2 s2 = 100

x1 + 2 x2 10 x1 6 x0

A solution is x1 = 5, x2 = 2.5.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Goal simplex
The goal simplex optimizes, rather than satisfies, the goals. It adds an objective row to every goal in the goal model, and the goals are optimized sequentially one by one. When the process of optimizing one goal reaches the end, before the next optimization is started, any nonbasic variable having a reduced cost in the row of the goal just been optimized different from zero must be removed.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

Example
Consider the production mix problem to maximize revenue and reduce inventory:
max min s.t. 4 x1 + 8 x2 8 x1 + 24 x2 x1 + 2 x2 10 x1 6 x0

Assume that maximizing revenue is more important than minimizing inventory. Hence, the first objective will be optimized first.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Adding slack variable, the simplex tableau with two objectives is:
Basic P1 P2 s1 s2 x1 4 8 1 1 x2 8 24 2 0 s1 0 0 1 0 s2 0 0 0 1 Solution 0 0 10 6

To improve P1, x2 has to be basic, so it will be considered g as the entering variable. By applying the minimum ratio, s1 has to leave the basis.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 Apply the row operations to eliminate x2 from all rows except its row:
Basic P1 P2 x2 2 s2 x1 0 4 1 x2 0 0 1 0 s1 4 12 0.5 0 s2 0 0 0 1 Solution 40 120 5 6

This tableau is optimal, all reduced costs in P1 are 0.

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The nonbasic variable s1 has reduced costs of 4 in P1. Hence, s1 should be eliminated before P2 can be , optimized. The tableau becomes:
Basic P1 P2 x2 2 s2 x1 0 4 1 x2 0 0 1 0 s2 0 0 0 1 Solution 40 120 5 6

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The row of P1 can be eliminated as it has no effect on the solution. To improve P2, x1 should be basic (> 0). By applying the minimum ratio, s2 should become nonbasic:

Basic P1 P2 2 x2 x1

x1 0 0 1

x2 0 1 0

s2 4 4 1

Solution 40 96 2 6

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM

 The tableau is optimal and the solution has been reached: x1 = 6 x2 = 2

DrMuhammadAlSalamah,IndustrialEngineering,KFUPM