Autonomous prediction framework for the seismic risk of RC girder bridge clusters considering fatigue and mainshock-aftershock sequences

To study the seismic fragility and failure behavior of reinforced concrete continuous bridges (RCCBs) under the influence of fatigue and mainshock-aftershock sequences, a simplified calculation model for the fatigue level of RCCBs considering pedestrian and traffic loads is proposed. A probabilistic earthquake risk function considering the mainshock-aftershock sequence is developed via the theories of earthquake reliability and probability hazards. A monitoring intensity equation is proposed that considers the spatial vector synthesis of ground motion indicators under the mainshock-aftershock sequence. A distribution cloud map of vector-synthesized earthquake intensity considering the mainshock-aftershock sequence is generated using network monitoring data from the 2008 Wenchuan earthquake in China and the 2024 Ishikawa earthquake in Japan. A failure probability matrix is established on the basis of onsite inspections of 386 RCCBs affected by the Wenchuan earthquake, considering updated fragility and fatigue levels. A nonlinear regression algorithm for determining the seismic fragility of RCCBs considering the mainshock-aftershock sequence and fatigue level is proposed. Using deep learning and ensemble learning algorithms, ten feature parameters that significantly affect the seismic fragility of RCCBs are used as input variables for the automation model, and the optimized seismic fragility level of the bridge is used as the output variable of the model. Using random forest (RF), multilayer perceptron (MLP), convolutional neural network (CNN), and recurrent neural network (RNN) algorithms, four autonomous models have been developed to predict the seismic fragility of RCCBs.