Interpretation of dam deformation and leakage with boosted regression trees

Abstract

Predictive models are essential in dam safety assessment. They have been traditionally based on simple statistical tools such as the hydrostatic-seasontime (HST) model. These tools are well known to have limitations in terms of accuracy and reliability. In the recent years, the examples of application of machine learning and related techniques are becoming more frequent as an alternative to HST. While they proved to feature higher flexibility and prediction accuracy, they are also more difficult to interpret. As a consequence, the vast majority of the research is limited to prediction accuracy estimation. In this work, one of the most popular machine learning techniques (boosted regression trees), was applied to model 8 radial displacements and 4 leakage flows at La Baells Dam. The possibilities of model interpretation were explored: the relative influence of each predictor was computed, and the partial dependence plots were obtained. Both results were analysed to draw conclusions on dam response to environmental variables, and its evolution over time. The results show that this technique can efficiently identify dam performance changes with higher flexibility and reliability than simple regression models.

Hysteretic effect considered monitoring model for interpreting abnormal deformation behavior of arch dams: A case study

“…In recent years, there is a tendency towards employing advanced tools in the machine learning community to establish prediction models, such as neural network, 12,13 support vector machine, 14,15 random forests, 16 extreme learning machine, 17 and boosted regression tree. 18 Dam effect quantity and its influencing factors are respectively taken as the output and input variables, and according to the observation data, monitoring models can be established by training the learning rules of the used artificial intelligence algorithm. 10 However, the machine learning technique-based monitoring models are without considering the structure characteristics of concrete dams, and there are no direct mathematical expressions, so that they can only be used for prediction, rather than causal interpretation of dam deformation like statistical model.…”

“…To conquer such drawbacks, the second category of data‐based model has been developed. In recent years, there is a tendency towards employing advanced tools in the machine learning community to establish prediction models, such as neural network, support vector machine, random forests, extreme learning machine, and boosted regression tree . Dam effect quantity and its influencing factors are respectively taken as the output and input variables, and according to the observation data, monitoring models can be established by training the learning rules of the used artificial intelligence algorithm .…”

“…Dam effect quantity and its influencing factors are respectively taken as the output and input variables, and according to the observation data, monitoring models can be established by training the learning rules of the used artificial intelligence algorithm . However, the machine learning technique‐based monitoring models are without considering the structure characteristics of concrete dams, and there are no direct mathematical expressions, so that they can only be used for prediction, rather than causal interpretation of dam deformation like statistical model …”

Statistical model optimized random forest regression model for concrete dam deformation monitoring

“…On the other hand, powerful tools such as neural networks and support vector machines have been developed, which make use of observed data for interpreting complex systems . These more flexible and accurate models are available but are more difficult to implement and analyze …”

A deformation separation method for gravity dam body and foundation based on the observed displacements

“…Forward analysis models are fundamental approaches that play a major role in dam safety assessment systems . By using these models, it is possible to determine the expected response from the prototype observations.…”