PSI - Issue 78

Pooria Mesbahi et al. / Procedia Structural Integrity 78 (2026) 1839–1846

1844

13.411 m

13.411 m

18.288 m

1.676 m 0.19 m

" " " " "

H 1.143 m 1.500 m 1.000 m

1.143 m 1.000 m H 1.500 m

0.762 m

1.143 m

6 ф 35.8 mm

0.914 m Pier section 36 ф 25.4 mm

4 ф 25.4 mm

6.704 m

Cap beam section

6 ф 35.8 mm

0.914 m

14 m

ф 12.7 mm @ 89 mm Hoops

Fig. 3: Considered layout of multispan continuous concrete I-girder bridge for Bridge1 (H = 6.706 m) and Bridge2 (H = 7.5 m) as assets in the target bridge network

10

# 10

11

# 10

2.8

10

11

2.15

# 10

# 10

2.4 2.6 2.8 3 3.2

1.6 1.8 2 2.2 2.4 3.5 4 4.5 5 5.5

Burn-in

2.1

Burn-in

Burn-in

2.7

Burn-in

2.05

E s [Pa]

E c [Pa]

2.6

E s [Pa]

E c [Pa]

2

0 1000 2000 3000 4000 5000 6000 7000 8000 Iteration 2.5

0 1000 2000 3000 4000 5000 6000 7000 8000 Iteration 1.95

0

500

1000

1500

0

500

1000

1500

Iteration

Iteration

7

8

7

8

# 10

# 10

# 10

# 10

3.2

4.9

3.5

Burn-in

3

Burn-in

4.8

Burn-in

Burn-in

3

2.8

f c 0 [Pa]

f y [Pa]

f c 0 [Pa]

f y [Pa]

4.7

2.6

2.5

0 1000 2000 3000 4000 5000 6000 7000 8000 Iteration 4.6

0 1000 2000 3000 4000 5000 6000 7000 8000 Iteration 2.4

0

500

1000

1500

0

500

1000

1500

Iteration

Iteration

4

# 10

4

2.6

# 10

0.055

0.05

2.5

Burn-in

Burn-in

2.2 . [N/m 3 ] 2.4

0.05

0.048

Burn-in

Burn-in

2.4 2.45 . [N/m 3 ]

9 [-]

9 [-]

0.045

0.046

0 1000 2000 3000 4000 5000 6000 7000 8000 Iteration 0.044

2

0.04

0 1000 2000 3000 4000 5000 6000 7000 8000 Iteration 2.35

0

500

1000

1500

0

500

1000

1500

Iteration

Iteration

(a) Trace plots of the samples of six model parameters generated by MCMC and DRAM algoritm in the probabilistic FEMU process

(b) Trace plots of the samples of six model parameters generated by DE-MC in the probabilistic FEMU process

Fig. 4: Trace plots of the Markov chains of six model parameters generated by DRAM (a) and DE-MC (b) in the probabilistic FEMU process

Fig. 5: Prior and posterior distributions of model parameters before and after Bayesian FEMU for Bridge1

After obtaining the posterior distributions of all model parameters for B1 via Bayesian FEMU, the BN presented in Fig. 2 is used to estimate the posterior distributions of the corresponding parameters for B2. For simplicity, the BN structure is assumed to be identical for all model parameters. To construct the BN, CPDs are first defined for each node. Then posterior of each B1 parameter, θ i , 1 , is discretized into bins corresponding to the levels of the associated CPD. After running the BN in Python and incorporating the e ff ects of comparative nodes, the discretized posterior distribution for each B2 parameter, θ i , 2 , is obtained (Fig. 6).