Displacement is the most intuitive reflection of the structural behaviour of concrete dams, and it is of great significance to predict future displacement for dam health diagnosis. The mathematical models used to predict displacement are mostly established with the only regression objective of minimizing the mean square error between the measured and fitted displacements, whereas the spatial associations between the displacements of multiple monitoring points of an arch dam are ignored. To increase the prediction accuracy of machine learning technique-based mathematical models, a spatial association-coupled support vector machine model is proposed in this article to predict the displacement of high arch dams. This approach is conducted by performing an incremental distance-based spatial clustering for dam displacement field in the first step. The displacement spatial association is quantified by the integrated shape similarity index between the measured time series of multiple monitoring points and is then coupled with the fitting mean square error to optimize the parameters of the support vector machine model. A case study of an engineering example indicates that the prediction accuracy and generalization ability of the proposed double objective support vector machine model have been greatly improved compared to the traditional single objective support vector machine model. For the total 34 plumb line monitoring points on the dam body of the Jinping-I arch dam, when using the hydraulic, seasonal and time- and hydraulic, hysteretic, seasonal and time-based double objective support vector machine models, the prediction accuracy of 25 and 21 monitoring points increases with an average rate of 50.8% and 47.4%, and the degrees of overfitting are evenly reduced by 44.3% and 70.9%, respectively.
Analysis on Finite Element Grid Size and Discrete Error for Super-high Arch Dams
