Dynamic Response Prediction for Steel Arch Bridges Using a CEEMDAN–VMD–CNN–BiLSTM Framework

Vehicle loads and environmental effects during service produce nonlinear and non-stationary bridge response signals that are also affected by noise. These characteristics make accurate response prediction challenging for conventional approaches. This study proposes a hybrid prediction framework that integrates Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Variational Mode Decomposition (VMD), and a Convolutional Neural Network–Bidirectional Long Short-Term Memory Network (CNN-BiLSTM). The framework is evaluated using multi-source monitoring data collected by the structural health monitoring system of the Furong Town Bridge in China. CEEMDAN first decomposes the original structural response signals into multiple intrinsic mode functions (IMFs). These are classified using sample entropy and K-means clustering, and the high-frequency composite component is further decomposed by VMD. The CNN-BiLSTM network predicts the resulting subcomponents individually. Their predictions are then combined to reconstruct the structural response. Validation using arch-crown displacement and mid-span girder deflection shows that the framework achieves [Formula: see text] values above 0.99 for both tasks, with lower prediction errors than those of the five comparison models. The results suggest that the proposed approach has potential for predicting bridge displacement and deflection under field conditions.