25.12.2014 Views

Sustainable Construction A Life Cycle Approach in Engineering

Sustainable Construction A Life Cycle Approach in Engineering

Sustainable Construction A Life Cycle Approach in Engineering

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

The results of pairwise comparisons are used to def<strong>in</strong>e judgment matrices, also called Pairwise<br />

matrices or P matrices.<br />

Each entry p ij of these matrices corresponds to the number that estimates accord<strong>in</strong>g to selected<br />

scale the relative performance (or weights of importance) of entity E i when it is compared<br />

with E j . Entities can be represented by alternatives A i or criteria C i that have to be compared<br />

<strong>in</strong> terms of selected criterion or decision goal.<br />

E E E E E<br />

1 2 3 4 n<br />

E p p p<br />

... ... ...<br />

E<br />

<br />

2<br />

p21<br />

P <br />

nxn<br />

<br />

1 11 12 1n<br />

<br />

E<br />

<br />

p p<br />

n m1 mn<br />

<br />

<br />

<br />

<br />

<br />

<br />

(5)<br />

The matrix of pairwise comparisons is reciprocal, be<strong>in</strong>g p ij =1/p ij and p ii =1. In case it is possible<br />

to have exact value of p ij = w i /w j , with w i actual weigth of importance of entity E i , the matrix<br />

[P] is consistent, that is p ij =p ik·p kj . It can be shown that for real values of p ij the rank of [P] is 1<br />

and that λ=n is the pr<strong>in</strong>cipal eigenvalue. When the values of p ij are close to actual ones but different,<br />

which is the most common case <strong>in</strong> applications, the [P] matrix is not consistent, the rank<br />

is greater than 1 and the maximum eigenvalue λ max is greater than or equal to n.<br />

It can be demonstrated that the normalized pr<strong>in</strong>cipal eigenvector w of P matrix that corresponds<br />

to the maximum eigenvalue λ max represents the vector (a i1 , a i2 ,…., a <strong>in</strong> ) with relative performance,<br />

or weights, of compared entities, that is the i-th colum of matrix [A] or the vector of<br />

relative weights of criteria.<br />

In particular, it results:<br />

<br />

P w w with max n<br />

(6)<br />

max<br />

w=(a<br />

i1, a<br />

i2,..., a<br />

<strong>in</strong>)<br />

(7)<br />

A simplified calculation of eigenvector w can be performed by multiply<strong>in</strong>g the entries <strong>in</strong><br />

each row of matrix P together and tak<strong>in</strong>g the n-th root. In order to have values that add up to 1,<br />

a normalization by the sum of obta<strong>in</strong>ed values has to be carried out. A simplified procedure to<br />

obta<strong>in</strong> λ max is to add the columns of matrix P and the multiply<strong>in</strong>g the result<strong>in</strong>g vector with w.<br />

To estimate the consistency of pairwise comparison matrix, Saaty proposed the Consistency<br />

Ratio CR, obta<strong>in</strong>ed divid<strong>in</strong>g the consistency <strong>in</strong>dex by the Random Consistency Index RCI given<br />

<strong>in</strong> table 2.<br />

CI is def<strong>in</strong>ed as follows:<br />

CI ( n)/(n 1)<br />

(8)<br />

max<br />

RCI is an average random consistency <strong>in</strong>dex derived from randomly generated reciprocal matrices.<br />

Accord<strong>in</strong>g to Saaty, the CR should be less than 10% <strong>in</strong> order to have consistency <strong>in</strong> pairwise<br />

judgments.<br />

Table 2. RCI values of sets of different order n.<br />

n 1 2 3 4 5 6 7 8 9<br />

RCI 0 0 0.58 0.9 1.12 1.24 1.32 1.41 1.45<br />

180

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!