Abstract: During bridge reinforcement construction, construction control relies on the integrity and accuracy of monitoring data. However, monitoring data often suffer from long-term continuous data loss due to sensor failures or construction disturbances.Based on linear regression imputation methods, this paper adopts the global Pearson correlation coefficient and local effective data coverage as dual thresholds: a candidate sequence of reference points is first constructed in descending order of correlation with the target point, and each candidate is then verified for local data completeness during the missing period. The first candidate satisfying both thresholds is selected as the regression reference point to perform the imputation. With the threshold values determined based on statistical characteristics of the dataset and information requirements of construction monitoring,experimental validation using measured data shows that the coefficient of determination (r²) between the imputed and actual values is 0.8432, and the root mean square error (RMSE) is 0.7516. This method was applied to impute 1440 periods of arch rib stress monitoring data during the tie-bar replacement construction of an arch bridge in Guangdong Province. The repair rate for long-term continuous missing data reached 75.75%. The repaired curve restored the original variation characteristics of the target point, satisfying the minimum information continuity requirement for construction control. The method provides a reference for similar engineering applications.