Real measurement data-driven correlated hysteresis monitoring model for concrete arch dam displacement

Abstract

Displacement is an important indicator reflecting the safety performance of concrete arch dams and displacement monitoring models are widely adopted to analyze the health and serviceability of concrete arch dams. The measured data shows that the displacement changes of concrete arch dams are integral and the displacement changes in adjacent areas have high similarity. However, existing displacement monitoring models mostly focus on prediction accuracy and often ignore the correlation of displacement changes, thus failing to reflect the integrity of arch dam displacement changes. In this article, considering the hysteresis of displacement changes a real measurement data-driven correlated hysteresis monitoring model is established with dual objectives of prediction accuracy and displacement similarity. Firstly, the clustering analysis method was adopted to partition the monitoring points of concrete arch dams according to the displacement change speed. Secondly, a dual-objective prediction model was established with the objectives of prediction accuracy and displacement similarity based on Tanimoto measure. Finally, considering the trend of monitoring point displacement changes and displacement correlation, the Holt-winters algorithm combined with the Pearson correlation coefficient was adopted to modify the results of the dual-objective prediction model, thereby establishing a real measurement data-driven correlated hysteresis monitoring model. Engineering examples show that compared with other prediction models, the real measurement data-driven correlated hysteresis monitoring model comprehensively considers the influences of typical factors and displacement hysteresis effects, and can significantly improve the prediction accuracy while reflecting the correlation of displacement changes well, which is a trustworthy displacement monitoring model for concrete arch dams.