PSI - Issue 44

Gabriella Tocchi et al. / Procedia Structural Integrity 44 (2023) 1972–1979 Gabriella Tocchi et al./ Structural Integrity Procedia 00 (2022) 000–000

1977

6

municipalities are already included in Cartis database and for both only two TCs are identified, the HC and the Expansion Area. The aim of this application is to evaluate the reliability of the proposed approach for delimitation of TCs and to test the possible use of Cartis-based statistical information about building typologies to compile building inventory also in municipalities for which the Cartis form was not compiled. The logistic regression described in paragraph 3 is used to predict census tracts belonging to TC01; such predicted compartments are then compared to the ones identified as HC (also called TC01) by effective compilation of Cartis form. In Fig. 3 the comparison between the TC01 as reported in the Cartis form and the one obtained according to the proposed model for the municipality of Frignano is shown. The latter can be graphically obtained grouping all census tracts classified as belonging to HC. The accuracy of the model for the considered case study is 0.87, meaning that the 87% of census tracts are well classified. As it can be noted by the figure, the TC in Cartis form was delineated using the street layout without considering the extension of census tracts. Thus, assuming HC also includes census tract only partially covered by it, only two census tracts are wrongly classified by the algorithm, i.e., census tract number 1 and number 3, that are characterized by relative low number of pre 1945 buildings (22% and 4% respectively) and high population density (85% and 86%). Table 1. Input and output census tracts’ data for the municipality of San Cipriano d’Aversa. Census tracts wrongly assigned by the model are highlighted in red.

Census tract

Pop Density

Buildings <1945

Real TC

Predicted TC

1 1 1 1 2 2 2 2 1 2 1 1 2 2 2 1

1 2 3 4 5 6 7 8

76% 96% 96% 87% 77% 46% 79% 53% 79% 88% 30% 97% 79% 77% 0%

7%

1 1 1 1 1 1 2 2 1 2 1 1 2 1 1 2

11% 24%

100%

9% 0% 4% 0% 0%

10 11 12 13 14 15 16 17

32%

0%

10% 13%

2% 2%

30%

1%

The application of the proposed model to San Cipriano d’Aversa municipality shows very similar results: the accuracy on the unseen set of data is 0.75; this means that only 1/4 of census tracts are wrongly assigned to relative TCs (5, 6, 15 and 16, see Table 1). However, in the whole municipal area the number of buildings built before 1945 is very low (approximately 10% of residential buildings) therefore in this town the association to HC is mostly governed by population density. Table 1 shows input and output data for each census tract of San Cipriano d’Aversa municipality. It is also worth noting that even if tract number 16 originally is not included in HC, this tract shows a very high percentage of pre 1945 buildings (30%) with a relative high population density (79%). Once delimitated, the total number of masonry and RC buildings in the TC can be assumed as the sum of buildings in census tracts belonging to it. However, as already mentioned, information on the sole construction material of the main load resisting structure may be not enough for the application of some vulnerability models. For defining a suitable exposure model at TC level, Cartis-based statistics on building typologies may be adopted. The analysis of the database in the considered region shows that the diffusion of main structural features mostly varies based on the geographic area (Fig. 4). For example, considering the administrative areas identified by provinces, irregular layout masonry buildings are widespread in the province of Benevento (BN), Salerno (SA) and Avellino (AV), where most of these structures are included in the first TC; on the contrary, they are not very common in the province of Caserta (CE) and Naples (NA), regardless the considered TC. The municipalities could