PSI - Issue 64

Lim Boon Xuan et al. / Procedia Structural Integrity 64 (2024) 791–798 Lim et al./ Structural Integrity Procedia 00 (2019) 000–000

797

7

Table 2. Type of error from different prediction outputs. Type of error Linear regression

Scheme 1

Scheme 2

Scheme 3

Scheme 4

RMSE

0.6527 0.5208 1.7599

0.3155 0.2572 0.8880

0.0919 0.0698 0.3113

0.7199 0.6226 1.3846

0.1908 0.1522 0.4588

MAE

Max-AE

Residual value (mm)

Residual value (mm)

Residual value (mm)

(a)

(b)

(c)

(d)

Fig. 6. Residual plot of prediction results: (a) Scheme 1; (b) Scheme 2; (c) Scheme 3; (d) Scheme 4

Count

(a)

(b)

(c)

(d)

Fig. 7. Histogram of residual values: (a) Scheme 1; (b) Scheme 2; (c) Scheme 3; (d) Scheme 4

6. Conclusion A temperature-induced deflection prediction model for continuous box girder bridges is established based on deep learning techniques, demonstrating high precision in the presented case study. The main conclusions of this paper are shown below: 1. Temperature-induced deflection is effectively extracted from raw deflection data in both span-1 and span-6 by employing the EMD method. 2. Despite span-1 and span-6 possessed different magnitude of temperature-induced deflection, the implemented LSTM model effectively captures the complex relationships between temperature and temperature-induced deflection, generating low prediction errors of RMSE, MAE, and Max-AE. Notably, utilizing girder temperature as the input variable results in more accurate predictions of temperature-induced deflection compared to using ambient temperature. This paper demonstrates the potential of the LSTM network in predicting the deflection of continuous box girder bridges. Future research could extend to annual analyses, expanding the data to encompass multiple years to study the effects of seasonal temperature variations on deflection. Furthermore, additional deep learning techniques, such as the Transformer model (Vaswani et al., 2017), could be investigated to enhance the prediction model. Acknowledgements The authors may wish to express their sincere appreciation for the financial support provided by the National Natural Science Foundation of China (No.U2005216).