Steps (i) to (iii) were shown in the previous sections. The remaining steps are shown in the following sections.
6.2. Comparison of Criteria
It is worth noting that each of the selected criteria needs to be weighted with respect to the others in order to set their relative importance in the full evaluation of different retrofit solutions. This is achieved by using the method proposed by Saaty [
47], which is based on pairwise comparisons of criteria and the analysis of eigenvalues. Considering two criteria, Cj and Ck (j, k = 1, 2, …, 7), the relative importance of Cj with respect to Ck must be defined according to different linguistic definitions, reported in
Table 4, which are then converted into numerical values.
By comparing criteria according to the mentioned method on a pairwise basis, a series of values that populate an n × n matrix (A) is obtained, where n is the number of selected criteria. Values on the matrix diagonal are always 1, since derives from the comparison of Cj with itself. Moreover, it is always the case that .
Comparing criteria for weighting is not an easy task, since it can be influenced by the subjective evaluation of the decision-maker. In case of private buildings, the final decision makers are the tenants, who are often the owners, according to the widespread Italian custom of buying a home. It is true, of course, that choices of tenants are affected by counselling from building administrators, engineers in charge of the structural design (eventually complemented by an energy efficiency upgrade), and architects in charge of resolving architectural and functional issues. For these reasons, four subjects were charged with filling the A matrix to represent an architect (Arch.), a structural engineer (Eng.), a building administrator (Admin.) and a private citizen who could be a tenant (Ten.) (
Figure 8). All subjects were adequately educated on the TOPSIS method and, in particular, on how to make pairwise comparisons.
In addition, external factors can influence the comparison of criteria. For example, the existence of financial incentives, as is currently the case in Italy [
10], could make it possible to attribute a lower importance to criterion C1, related to the intervention cost, which is completely supported by public funds in the form of tax discounts. However, this will not be a permanent condition, since incentives usually expire after a certain period. In case of the presence of stores on the ground floor of the building, the work duration can assume high importance. On the other hand, disruption due to invasive retrofit work is recognized as one of the most significant restraints on the massive rehabilitation programme, supported by the current Italian incentive legislation, which is aimed at economic growth to recover from the COVID-19 crisis. Therefore, many people would identify disruption as the most important criterion for evaluating a retrofit solution.
From the above background, different versions of matrix A were considered in this study. Basically, the four individuals filled in two versions of the matrix, considering either the presence or absence of financial incentives, to finally obtain eight versions of the consequent relative weights for matrix A (
Table 5,
Table 6,
Table 7 and
Table 8).
Once the pairwise comparisons are carried out to determine the A matrices, a consistency evaluation related to the DMs is needed. The consistency evaluation is satisfied when, given a criterion Ci that is judged to be more important than criterion Cj by factor , and a criterion Cj that is in turn judged to be more important than criterion Ck by factor , the result is that Ci is judged to be more important than Ck by factor .
To this end, the so-called consistency index (
CI) can be calculated as follows:
where
is the maximum eigenvalue of the A matrix.
CI has to be normalized by the random consistency index (
RCI), which is an average random consistency factor depending on
n (
Table 9). The ratio
is the consistency ratio (CR). Consistency control is considered satisfactory when CR is less than 5% if
n = 3, 9% if
n = 4, and 10% if
n > 4.
Once the consistency control is passed, the vector (W) of relative weights w1, w2, …, w7 can be determined as the principal right eigenvector. Otherwise, pairwise comparisons must be redone to achieve the minimum required consistency.
For the case under consideration, the vectors of weights and the related CR values are reported in
Table 10 and
Table 11.
In the absence of incentives, the highest weights are attributed to criteria C1 (intervention cost), C3 (seismic performance) and C6 (invasiveness), and in the presence of incentives they are C3, C6 and C7 (durability). The presence of incentives shifts the decision-maker’s attention from intervention cost to durability. In fact, a strengthening solution with low durability would force tenants to face maintenance expenses in a probable future period without incentives.
Moreover, the best consistency (lower CR values) is obtained by technicians (Eng. and Arch.), as would be expected, while for non-technical individuals (Admin. and Ten.) CR values are near the upper limit of 10%.
Table 10 and
Table 11 also show columns with the average of the four weight vectors that will be assumed in the following evaluations. It is assumed that the final decision is influenced by all actors involved in the decision-making process.
6.3. Ranking of Retrofit Solutions
The TOPSIS method [
16] is based on a concept which states that the best solution is the one with the minimum distance from an ideal solution A* and, therefore, the maximum distance from the “anti-ideal” solution A-. Assuming
xij as the measure of the performance of the
ith solution (i = 1, 2, 3) with respect to the
jth criterion (j = 1, 2, …, 7), all values
xij (3 × 7 = 21) are collected in decision matrix D = [
xij] (
Table 12).
Regarding criteria C5, C6 and C7, due to their “qualitative” nature, only linguistic judgments can be expressed. To apply the TOPSIS method, it is, therefore, essential to convert these qualitative variables into “quantitative” ones by once again adopting the Saaty eigenvector method [
47].
Qualitative pairwise comparisons were made between the different solutions Si as a function of the degree of compliance with the considered criterion, and the procedure was carried out separately for C5, C6 and C7. The pairwise comparison was made by only two individuals out of the selected four, the engineer and the architect, since it requires technical skills in the building sector.
By adopting the linear Saaty scale (
Table 4), these judgments were then converted into numbers
aij in the range 1/9–9. In this case, with reference to
Table 4, the “relative importance”
aij should be interpreted as the relative measure of the degree of compliance with the considered criterion of the alternative Si with respect to Sj. Once all values
aij are in a 3 × 3 matrix (reported in
Table 13 and
Table 14 for each qualitative criterion), Saaty’s method defines the elements of the main eigenvector as a quantitative measure of the performance of each retrofit solution with respect to the qualitative criterion under consideration. The binary comparisons in the matrices were entrusted to the expert judgment of the architect (
Table 13) and the structural engineer (
Table 14). The quantitative measure of each solution Si is reported in
Table 15.
A consistency ratio lower than 5% (acceptability threshold) indicates that there is a good degree of consistency among the expert judgments provided by the two compilers.
Accounting for the heterogeneity of the
xij D matrix coefficients, normalization is necessary in order to obtain the R = [
rij] matrix. The latter represents the normalized decision matrix.
The normalized decision matrix R = [
rij] is reported in
Table 16.
Subsequently, all values of the
ith column of the R matrix are multiplied by the weight w
i of the
ith criterion (“Average w
i” column in
Table 10). The weighted normalized decision matrix V = [w
j rij =
vij] is reported in
Table 17.
The ideal A* and the anti-ideal A- solutions (
Table 18) are obtained by putting together the best and worst performance with respect to each of the seven criteria for the three retrofit solutions (S1, S, S3).
Criteria from C1 to C6 can be classified as “cost” criteria, which require minimisation of their values to improve the intervention performance with respect to them. The opposite is true for criterion C7 (durability), which is a “benefit” type and should be maximized. Therefore, in defining the ideal and anti-ideal solutions, it is necessary to interpret the maximum value as the “best” performance for “benefit” type criteria and the minimum value for “cost” type criteria.
Each real or ideal retrofit solution can be represented by a point in a seven-dimensional space where the
jth axis measures the normalized and weighted performance of the considered alternative with respect to the generic criterion Cj. Then, the geometric distance
of the solution Si (i = 1, 2, 3) from the ideal A*, and the distance di- of Si from the anti-ideal A-, can be obtained with the following expression:
where
vj* and
vj- represent the
jth element of the vectors of seven elements (
Table 18) constituting the alternatives A* and A-. The values of the distances
and
di− are shown in
Table 19. Once the distances
and
di− of the alternative Si (
i = 1, 2, 3) are obtained from the solutions A* and A-, respectively, the relative distance of the retrofit solution from the ideal solution is obtained from the following ratio:
values for the three retrofit solutions are in the range [0.0–1.0]. According to the TOPSIS method, the solution ranking is conducted according to
values. The most suitable is the one with maximum
. In this case, in the absence of incentives, the most suitable solution is S3 (base isolation), with
= 0.732 (
Table 19). The exoskeleton solution (S2) is second, and local strengthening (S1) would be the least favourable solution.
Starting from
Table 16 reporting the normalized decision-making matrix
Rij, the same procedure was again applied, using the “Average w
i” column in
Table 11, which was obtained considering the availability of financial incentives (
Table 20 and
Table 21). This allowed to obtain
Table 20 reporting the weighted decision-making matrix and
Table 21 reporting the ideal and anti-ideal solutions.
In this case, the more suitable solution (
Table 22) is S2 (exoskeleton), while retrofit solution S1 (local strengthening) is always ranked third.
By applying the TOPSIS multi-criteria decision-making method, local strengthening retrofit options result in less suitable solutions for the building type under consideration. In both assumed scenarios (presence/absence of incentives), global techniques such as exoskeleton (S2) and seismic isolation (S3) are more suitable because they are less invasive. The possibility of intervening from the outside without having to relocate tenants has great value and can be important in the final decision.
When external or public financial resources are available in the form of either tax discounts or grants, decision makers shift their focus to the durability of the retrofit solution. In this regard, the exoskeleton solution is judged to be more durable compared to base isolation, which may involve expensive replacement of isolation devices due to effects of aging on their constitutive materials. As shown in
Table 22,
is equal to 0.735 for S2 and 0.684 for S3, with a 7% difference.
In the opposite situation, that is, without financial incentives, the most suitable solution is base isolation. In this case, the cost of system adaptation has a negative influence on S2 (exoskeleton), which requires a new foundation system with demanding excavation work along the building perimeter, while the cost of the same work for S3 is lower. However, careful evaluation is required, since, as can be seen from
Table 19, the difference in
between S3 and S2 is only 3%, and a slight increase in this parameter could invert the ranking positions.
The results obtained in terms of ranking can be considered to be generally valid for building types such as the one considered here, which is largely widespread in Italy and the Mediterranean area. However, considering a different building type and varying levels of seismic hazard (which is assumed high in this study) could lead to different results.