.png){kind=icon}
.png){kind=icon}
Return
Methodology for mapping bridge temperature-deflection considering uncertainty
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engstruct.2024.119580.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Bridge deflection is a critical measure of structural stiffness, influenced by a variety of factors, making it challenging to establish a clear temperature-deflection relationship. Current research separates the temperature- induced deflection component from deflection data. However, the authenticity of this separation has not been adequately verified. This study introduces a novel neural network, the Residual Connection Mixture Density Network (ResMDN), which eliminates the need for preprocessing deflection data. The ResMDN allows for the construction of a probability distribution prediction model for temperature-induced deflection, effectively addressing uncertainties with limited prior knowledge. The research findings demonstrate a significant linear relationship between the ResMDN-predicted variance of temperature-induced deflection and vehicular weight, thereby validating the accuracy of the ResMDN model. Furthermore, the study proposes using the maximum probability density of temperature-induced deflection as the theoretical value. A comparative analysis of deflection components extracted using different methods revealed that traditional approaches tend to capture the combined effect of temperature- and vehicle-induced static deflections. In contrast, the ResMDN accurately predicts the actual temperature-induced deflections, outperforming existing methodologies.
Keywords:
Structural health monitoring
Intelligent algorithm
Mixture density network
Bridge structure
Temperature-deflection mapping