Piezometric level prediction based on data mining techniques

Abstract

The safety assessment of dams is a complex task that is made possible thanks to a constant monitoring of pertinent parameters. Once collected, the data are processed by statistical analysis models in order to describe the behaviour of the structure. The aim of those models is to detect early signs of abnormal behaviour so as to take corrective actions when required. Because of the uniqueness of each structure, the behavioural models need to adapt to each of these structures, and thus flexibility is required. Simultaneously, generalization capacities are sought, so a trade-off has to be found. This flexibility is even more important when the analysed phenomenon is characterized by nonlinear features. This is notably the case of the piezometric levels (PL) monitored at the rock–concrete interface of arch dams, when this interface opens. In that case, the linear models that are classically used by engineers show poor performances. Consequently, interest naturally grows for the advanced learning algorithms known as machine learning techniques. In this work, the aim was to compare the predictive performances and generalization capacities of six different data mining algorithms that are likely to be used for monitoring purposes in the particular case of the piezometry at the interface of arch dams: artificial neural networks (ANN), support vector machines (SVM), decision tree, k-nearest neighbour, random forest and multiple regression. All six are used to analyse the same time series. The interpretation of those PL permits to understand the phenomenon of the aperture of the interface, which is highly nonlinear, and of great concern in dam safety. The achieved results show that SVM and ANN stand out as the most efficient algorithms, when it comes to analysing nonlinear monitored phenomenon. Through a global sensitivity analysis, the influence of the models’ attributes is measured, showing a high impact of Z (relative trough) in PL prediction.