PSI - Issue 84

Lorenzo Principi et al. / Procedia Structural Integrity 84 (2026) 73–80

79

(a) ADT ≤ 10k

10k–25k ≥ 25k

(b) ADTT ≤ 300

300 – 700 ≥ 700 (c) Deck Material R.C. P.R.C. Hybrid (d) Obstacle Type

River Road Natural Disc .

(e) Seismic Design

(f) P.G.A.

Figure 5. Geospatial distribution of the input features along the SS77.

According to the predictions results (Table 9 ), most bridges are classified as M (53) and M - H (34), indicating an overall moderate - to - high seismic risk, likely influenced by regional PGA values and the presence or absence of seismic design provisions. Figure 6 shows that M - H and H class bridges are concentrated in the central section of the highway, whereas M class bridges are more frequent at the beginning and end. Higher - risk (M - H and H) bridges generally lack seismic design, in contrast to the modern, fully seismic - designed segment of SS77, which falls within the M class. The elevated seismic risk in the central section is attributed to both older, non - seismic structures and a greater number of crossings over primary and secondary roads (Figure 5d). Bridges along the coast are predominantly M class, likely due to a higher incidence of river crossings rather than roads, reducing overall exposure.

Table 9. Predictions of the CER.

L M-L M M-H H 0 0 53 34 8

(c) CER

Figure 6. Geospatial distribution of the CER predictions.

5. Conclusions This study developed an ANN model to predict the CER for existing bridges at a regional scale. The model relies on readily available input data, primarily from archives, eliminating the need for on-site inspections. Building on a previously established framework, the approach was extended to focus specifically on seismic risk prediction. The optimal ANN achieved an F1-score of 71%, demonstrating robust performance. In addition, a GIS-based tool was implemented to visualize the model outputs and support regional management. The proposed framework proves to be effective and scalable, with applicability to other hazards, such as floods and landslides.