IMPLEMENTATION OF A NETWORK ADJUSTMENT APPLICATION FOR GEOMETRIC MONITORING OF ENGINEERING STRUCTURES: CASE OF THE NACHTIGAL DAM

Abstract

Accurate Precise dam deformation forecasting is vital to structural safety monitoring. This study creates an integrated framework comprised of: (1) an attention-enabled Long Short-Term Memory (LSTM) model with forecasting capability for deformation, and (2) high-level processing of altimetric leveling data with geometric monitoring, executed through a special-purpose geodetic network adjustment application applied to engineering structures. For forecasting deformation, Random Forest (RF) determines impact factor importance to filter input variables, while Density-Based Spatial Clustering (DBSCAN) removes equipment-induced anomalies to reduce measurement errors. The attention-enabled LSTM emphasizes significant temporal features, with superior nonlinear forecasting ability. In parallel, our built application processes digital leveling data through least squares adjustment with robust corrections applied to atmospheric refraction effects, Earth curvature effect, and instrumental/distance error effects. Validation on Nachtigal Dam returned millimetric precision: adjusted altitudes indicated 1.82 mm mean standard deviation, 0.25 mm dispersion, and homogeneity ratio of 1.41. Classification resulted in 66.7% "Good" and 33.3% "Acceptable" points with no outliers present. Comparison with Covadis software showed strong correlation (0.0229 m mean difference). This integrated solution deploying geodetic network adjustment applied to monitoring a damworks presently as an entire monitoring set-up, with future extension planned through geoid modeling and integration with gravity data.