Laboratory Evaluation of Anomaly-Aware MEMS Displacement Reconstruction for Slope Monitoring near Buildings

Reliable measurements of local displacement provide useful information for slope monitoring near buildings. Displacement reconstruction from low-cost MEMS acceleration is affected by gravity projection, sensor bias, and anomalous observations. This study presents a MEMS–LoRa sensing node and an anomaly-aware state estimation framework (LPD-RSE) for short-duration displacement reconstruction. The model accounts for gravity projection and estimates displacement, velocity, acceleration, residual bias, and bias-drift rate. Residual occupancy and robust slope criteria distinguish isolated spikes, persistent bias, and slowly varying drift before compensation and a robust extended Kalman filter update. Controlled platform experiments tested continuous 15 cm motion, intermittent 5 + 5 + 5 cm motion, and 15 cm motion with 30° and 45° installation inclination. Relative to commanded platform displacement, the maximum repeated-trial relative errors were 2.53% and 1.87% in the continuous and intermittent tests, respectively; the corresponding endpoint RMSE values were 0.325 and 0.206 cm. LPD-RSE produced the lowest endpoint RMSE among the four displacement reconstruction methods tested and limited rest period drift to 0.070 cm. The inclined tests showed larger errors. In the extended six-method comparison against an independent displacement reference, LPD-RSE yielded an endpoint RMSE of 0.1397 cm and a mean trajectory RMSE of 0.0775 cm. Paired tests characterized the uncertainty of the observed differences. These laboratory results support short-duration displacement reconstruction and suggest a potential role in monitoring ground movement near buildings.