Machine Learning-Assisted Mechanical Analysis and Intelligent Design Methods for Civil Structures

This study presents a machine learning-assisted approach for the mechanical analysis and intelligent design of civil structures. The proposed method addresses key challenges such as multi-source data fusion, dynamic updates, and hybrid variable modeling. The mechanical behavior of a cable-stayed bridge is first analyzed to identify critical safety factors, including cable force deviations, main girder stress, and temperature gradients. These factors are incorporated into a unified probabilistic graphical model using a hybrid Bayesian network. In this model, continuous parent nodes are discretized via hidden gate nodes, enabling effective hybrid variable representation. Dependencies among variables are defined using conditional probability tables, and real-time updates are achieved through Bayesian inference. To validate the method, a cable-stayed bridge model with 20 cable nodes and one load node is developed.