Recovery of missing strain data under multiple time-varying effects using IMM Kalman filtering

Data loss usually occurs in field structural health monitoring (SHM), which leads to difficulties in the data analysis and safety assessment of structures during the construction stage. Since the state model of structural deformation is complicated with multiple time-varying effects coupled, existing data recovery method based on a single state model is not capable to accurately recover long-term strain loss due to the perturbations and uncertainties in the state model. This study proposes an effective data recovery method for recovering missing strain based on the interactive multi-model (IMM) Kalman filtering. The estimation error brought by the model uncertainty is reduced by the fusion of multiple models. The accuracy and effectiveness of the proposed method were verified through a concrete shrinkage and creep experiment. The results showed that the recovered data is in good agreement with the real data within long-term missing zone. Furthermore, this method was applied to a real super high-rise building to recover the missing strain data during the construction stage. The recovered strain data provide valuable assistance to the construction managers to understand the complete evolution of the structural strain under time-varying effects during the construction process.