PSI - Issue 84

Stefano Bozza et al. / Procedia Structural Integrity 84 (2026) 852–858

857

Regulation Bridge class Model a 0 D. 1933 First class BL

a 1

a 2

a 3

̅

Table 1. Fitted model parameters of the best model for each regulation and bridge class considered.

R 2

0,493 0,271 0,322 0,504 0,319 0,706 0,359 0,887 0,716 1,539 0,948 0,833

1,31E-02m -1 6,13E-03m -1 1,27E-03m -1 6,42E-03m -1 1,03E-03m -1 2,00E-03m -1 8,15E-04m -1 -4,95E-03 m -1 -5,12E-03 m -1 -5,42E-02 m -1 -9,07E-03 m -1 3,07E-04m -1

-9,13E-03 m -1 2,95E-03m -1 6,63E-03m -1 4,11E-03m -1 6,30E-03m -1 -1,19E-03 m -1 3,34E-03m -1 4,96E-03m -1 3,40E-03m -1 1,61E-03m -2 5,75E-03m -1 -2,02E-04 m -1

21,5 m 0,965 24,5 m 0,741 14,5 m 0,798 25,7 m 0,853 14,4 m 0,759 25,4 m 0,153 25,0 m 0,243 20,1 m 0,354 20,1 m 0,504

0,011 0,025 0,024 0,017 0,024 0,025 0,021 0,020 0,020 0,006 0,016

D. 1933 C. 1945 C. 1962 C. 1962

Second class

BL BL BL BL BL BL BL BL

First class First class

Second class

D.M. 1980 First class D.M. 1980 Second class D.M. 1990 First class D.M. 1990 Second class D.M. 2005 First class D.M. 2005 Second class D.M. 2008 Second class

C

-1,69E-05 m -3

0,975

BL BL

19,2 m 0,818 29,6 m 0,069

0,009 The fitting models with the lowest coefficient of determination were those related to D.M.1980, D.M.1990 and D.M.2008 second class bridges (since D.M.2008 first class ones show ρ=1.00 ). In particular, models fitted to D.M.1980 and D.M.1990 datasets show R 2 values between 0.15 and 0.50 and mean absolute error in the range 0.020÷0.025 (comparable with the error related to second class bridges of previous regulations), while the fitted model of D.M.2008 second class bridges shows the worst coefficient of determination among all models (0.07) but also a low mean absolute error (0.01). The small error is due to a very low dispersion of the entire dataset, with ρ varying between 0.81 and 0.89. The bridge span as only independent variable is not sufficient to explain the dataset dispersion for all the regulations considered, and more complex models should be investigated in order to improve the fast preliminary assessment. However, since the mean absolute error in all fitted models are comparable, the presented models could be used to perform an approximated fast preliminary assessment of simply supported PC girder bridges. 4. Conclusion The “Guidelines for risk classification and management, safety assessment and monitoring of existing bridges” requires a preliminary assessment for bridges with Medium or higher Class of Attention, which could be time consuming for large bridge portfolios. In this paper, a tool for the fast preliminary assessment of simply supported PC girder bridges is presented. Several outdated codes were considered in order to take into account the age of construction, and a large number of simply supported decks with different geometry were analysed following the Guyon-Massonnet- Bareš method. The maximum bending moment in the most stressed girder was used as a demand parameter to compare the effects induced by outdated traffic load models and the traffic load model of the current technical code, performing a preliminary analysis on every deck considered. The deck span was assumed as the only independent variable of the proposed models, and then different functions were fitted to each dataset minimising the quadratic deviation, and the best fitting model was chosen as a tool for the fast preliminary assessment of PC girder bridges. Coefficient of determination and mean absolute error of the models were used to measure the goodness of fit of the models: models obtained for D.1933, C.1945, C.1962 and D.M.2005 shown a good fit to the dataset, while those related to D.M.1980, D.M.1990 and D.M.2008 (second class bridges) shown low R 2 values. However, in all models the mean absolute error was equal or less than 0.025, so the proposed models can be used for an approximate fast preliminary assessment of simply supported PC girder bridges. To improve the proposed models, more complex functions (considering more parameters or more independent variables) should be considered in further investigation. Moreover, other bridge typologies could be taken into account to generalize the fast preliminary assessment of existing bridges.