Due to the influence of complex service environments, the bridge health monitoring system (BHMS) has to face issues such as sensor failures and power outages of data acquisition systems, leading to frequent occurrences of data missing events including continuous and discrete data missing. By comparison, the continuous data missing can cover up the time‐series characteristic and make the corresponding recovery present a greater difficulty, especially for the data with a large loss rate or complicated features. To this end, this paper develops a novel signal recovery method based on the combination of successive variational mode decomposition (SVMD) and TCN–MHA–BiGRU, which is the hybrid of temporal convolutional networks (TCNs), multihead attention (MHA), and bidirectional gated recurrent unit (BiGRU). In this method, SVMD with high reliability and strong robustness is initially employed to decompose the original signal into multiple stable and regular subseries. Then, TCN–MHA–BiGRU incorporating the concept of “extraction‐weighting‐description of crucial features” is designed for the independent recovery of each subseries, with the ultimate recovery result derived through the linear superposition of all individual recoveries. This method not only can effectively extract the data time‐frequency characteristics (e.g., nonstationarity) but also can accurately capture the data time‐series characteristics (e.g., linear and nonlinear dependences) within the data. The case study and the subsequent applicability analysis grounded in the monitoring data from BHMS are employed to comprehensively evaluate the effectiveness of the proposed method. The results indicate that this method outperforms compared methods for the recovery of continuous missing data with different missing rates.

To address the problem in model computations and the limited accuracy of current bridge deterioration prediction methods, this paper proposes a novel bridge deterioration prediction meth-od using the whale optimization algorithm and extreme learning machine (WOA-ELM). First, we collected a dataset consisting of 539 sets of bridge inspection data and determined the necessary influencing factors through correlation analysis. Subsequently, the WOA-ELM algorithm was applied to establish a nonlinear mapping relationship between each influencing factor and the bridge condition indicators. Furthermore, the extreme learning machine (ELM), back-propagation neural network (BPNN), decision trees (DT), and support vector machine (SVM) were employed for comparison to validate the superiority of the proposed method. In addition, this paper provides further substantiation of the model’s exceptional predictive capabilities across diverse bridge components. The results demonstrate the accurate predictive capability of the proposed method for bridge conditions. Compared with ELM, BPNN, DT, and SVM, the proposed method exhibits significant improvements in predictive accuracy, i.e., the correlation coefficient is increased by 4.1%, 11.4%, 24.5%, and 33.6%, and the root mean square error is reduced by 7.3%, 18.0%, 14.8%, and 18.1%, respectively. Moreover, the proposed method presents considerably enhanced generalization capabilities, resulting in the reduction in mean relative error by 11.6%, 15.3%, 6%, and 16.2%. The proposed method presents a robust framework for proactive bridge maintenance.