A Kalman Filtering and Least Absolute Residuals based Time Series Data Reconstruction Strategy for Structural Health Monitoring

With the acceleration of urbanization and the aging of infrastructure, structural health monitoring systems (SHMS) has become increasingly important in civil engineering. Traditional SHMS rely on Internet of things (IoT) sensors mounted on structures. However, this approach faces challenges such as high maintenance cost and limited accessibility. Using unmanned aerial vehicles (UAVs) to carry sensors for data collection can provide greater flexibility and a clearer monitoring perspective. However, both approaches have frequent sensor failures because of the harsh operating environments and equipment wear. This leads to data loss or unreliability of the SHMS. To address this problem, this paper proposes a data reconstruction strategy using Kalman filtering and least absolute residual (KF-LAR) for the reconstruction of time series fault data. Kalman filtering is adopted to preprocess the time series data to eliminate noise and uncertainty in the data. Then, time series from sensors that are highly correlated with the faulty sensor are accurately identified through Pearson correlation coefficient analysis. Finally, a LAR reconstruction method is proposed to optimize the data reconstruction process to ensure the accuracy and robustness of the reconstruction result. This study proposes a novel data reconstruction solution to improve the reliability and efficiency of SHMS. It can also be widely adopted in IoT applications that facing similar data reconstruction challenges, including drones and aerial network systems.