You are on page 1of 7

Group: Zhiyuan Zheng; Shaonan Bi

Instructor: Dr. Elise Miller-Hooks

Course Project: Metaheuristic Project

CEIE 601: Infrastructure Systems Modeling

George Mason University

Maximum the Profit of a Network of Processing Raw Material

Abstract:

To get a comprehensive on Metaheuristic, our group design a problem solved by

Genetic Algorithm. This report demonstrates the process about how we figure it out in

details by six steps including formulation, definitions, constraints, object, data, and data

analysis. After getting the solution of this problem, our group use the GAMS to check

the validity of this program. In the discussion, our group apply this kind of problem

into more cases with more ideas.

Introduction

There are two kinds of raw materials. Including wood and stone, the total amount

of these two kinds of raw material is 10000 tons. There are five factories that can

process these raw materials. The factory 1, 2, and 3 can only process wood. The factory

3, 4, and 5 can process stone. The factory 3 can only process the mixture of wood and

stone as proportion of 1:2 as well as that 1 ton of wood combining 2 tons of stone can
make 1 ton of this specific commodity. In addition, the maximum processing capacity

of factory 1, 2, 3, 4 and 5 is 2000tons, 1800 tons, 3600 tons, 1900 tons and 1600 tons.

The material must transport from the warehouses to factories and the commodities must

transport to the dealership. The aim of this project is to maximum the profit of this

network. The price of commodities produced by these five factories are as following:
Factory /Price/Material Wood stone
1 530
2 480
3 800 800
4 500
5 540
The distances of each site are as following:
Distance Wood Warehouse Stone Warehouse Dealership
1 16mile 28 mile
2 24 mile 17 mile
3 17 mile 14 mile 22 mile
4 26 mile 13 mile
5 13 mile 30 mile

Fig. 1
Solution

For this problem, the metaheuristic is chosen to solve this problem. The specific

steps of Genetic Algorithm are as following.

Step 1: Initialization

Definition:

Xij = the amount from i to j, i I = {, , 1,2,3,4,5}, j J = {1,2,3,4,5, }

Cij = the cost of transportation from i to j, i I = {, , 1,2,3,4,5} , j J =

{1,2,3,4,5, }

Vn = the value of production of factory n, n N = {1,2,3,4,5}

G = the total profit of this project

Constraints:

Xw12000 Xs51600

Xw21800 Xs41900 Xw3+Xs3 3600

Xw1+Xw2+Xw3+Xs3+Xs4+Xs510000

Xw3:Xs3:X3t=1:2:1

Object:
1 1 + 2 2 + 3 3 + 3 3 + 4 4 + 5 5
G=Max[ 1 (1 + 1) 2 (2 + 2) 3 (3 + 3) ]
3 (3 + 3) 4 (4 + 4) 5 5 + 5

Get population of solutions:

It is hard to change the conditions of the factory 3. So we change the amount

transport to factory 1,2,4, and 5 (for example, if we change the amount of factory1, we

make other factories to gain the maximum amount). Then gain four parent solutions.
Number Factory 1 Factory 2 Factory 3 Factory 4 Factory 5 Profit
1 1100 1800 3600 1900 1600 2488000
2 2000 900 3600 1900 1600 2488000
3 2000 1800 3600 1000 1600 2506000
4 2000 1800 3600 1900 700 2497000
Step 2: Crossover

Each parent solution has 5 elements, but the amount of factory 3 is the same, so

we use the four elements of parent solution to crossover. We cut parent solution as two

fragments and crossover it, then get the offspring solutions.

Parents 1: Factory 1 Factory 2 Factory 4 Factory 5

Parents 2: Factory 1 Factory 2 Factory 4 Factory 5

Crossover

Offspring1: Factory 1 Factory 2 Factory 4 Factory 5

Offspring2: Factory 1 Factory 2 Factory 4 Factory 5

The offspring solutions are as following:

Number Factory 1 Factory 2 Factory 3 Factory 4 Factory 5 Total Feasibility


1 1100 1800 3600 1900 1600 10000
2 1100 1800 3600 1000 1600 9100
3 1100 1800 3600 1900 700 9100
4 2000 900 3600 1900 1600 10000
5 2000 900 3600 1000 1600 9100
6 2000 900 3600 1900 700 9100
7 2000 1800 3600 1900 700 10000
8 2000 1800 3600 1900 1600 10900
9 2000 1800 3600 1000 1600 10000

There should be 16 offspring solutions, but we abandon 7 solutions which are

repetitive.

Step 3: Mutation

We check the feasibility of each solution, and find the No. 8 offspring solution is
not feasible, so we make a mutation on it. We change the amount of factory 3 to make

it feasible.

Step 4: Evaluation

We use the final profits as the criterion. Then compute the profit of each solution

using the following equation:


1 1 + 2 2 + 3 3 + 3 3 + 4 4 + 5 5
G1=[ 1 (1 + 1) 2 (2 + 2) 3 (3 + 3) ]
3 (3 + 3) 4 (4 + 4) 5 5 + 5

The profits of solutions are as follow.

Offspring solutions:
Number Factory 1 Factory 2 Factory 3 Factory 4 Factory 5 Total Profit
2 1100 1800 3600 1000 1600 9100 2227000
3 1100 1800 3600 1900 700 9100 2218000
5 2000 900 3600 1000 1600 9100 2506000
6 2000 900 3600 1900 700 9100 2488000
8 2000 1800 2700 1900 1600 10000 2627500
Parent solutions:
Number Factory 1 Factory 2 Factory 3 Factory 4 Factory 5 Total Profit
1 1100 1800 3600 1900 1600 10000 2488000
2 2000 900 3600 1900 1600 10000 2488000
3 2000 1800 3600 1000 1600 10000 2506000
4 2000 1800 3600 1900 700 10000 2497000
Step 5: Selection

We make selection to choose the solution which can gain the highest profit.

Obviously, the No. 8 offspring solutions is optimal.

Step 6: Evolution

We choose the best two solution as parent solution. Then do the step 2 to 5.
Number Factory 1 Factory 2 Factory 3 Factory 4 Factory 5 Total
1 2000 1800 2700 1900 1600 10000
2 2000 1800 3600 1000 1600 10000
Crossover
Number Factory 1 Factory 2 Factory 3 Factory 4 Factory 5 Total
3 2000 1800 3600 1900 1600 10900
4 2000 1800 2700 1000 1600 9100

If we make mutation on solution No. 3, it will be same as the previous solutions,

so we abandon it. On the other, we check the profit of solution No. 4, it is 2366500

dollars. Finally, the best solution is offspring solutions No. 8.

Result

Double Check by GAMS:


Thus, the final solution is as following:

The quantities of wood shipped to factory 1 are 2000 tons, to factory 2 are 1800

tons, and to factory 3 are 900 tons. The quantities of stone shipped to factory 3 are 1800

tons, to factory 4 are 1900 tons, and to factory 5 are 1600 tons. The total profit of this

project is 2627500 dollars.

Discussion

By doing this program, we found that transportation management is one of the

difficult and complex problems. After finishing this project, we also found that there

are many variables to be considered in this kind of problem, such as total amount of

martial, distance, process capacity, profit and martial waste. We used the genetic

algorithm to solve problem, and found that if there are many uncertainties and optional

solutions in one question, you can use this algorithm, it is easy for people to mix and

find an optimal solution. In addition, academic research is a rigorous mission, it is

always beneficial to use an excel or GAMS to check the results you have already gained.

In conclusion, this kind of algorithm that we use can also be used to solve some

questions like minimum cost of building airlines to send passengers or supply chain

management.

You might also like