2023 – Advanced Data-driven Model-based Estimates of Dam Breach Parameters

Monte Azmi, Kyle Thomson

Identification and examination of key breach parameters are the main components of a comprehensive dam break modelling and consequence assessments. The commonly considered breaching parameters are the peak discharge following the breach, the time required for failure completion, and the final average width of the breach. While the physical and logical breaching mechanisms are mostly understood, reliable estimates of these parameters are challenging, rooted from lack of extensive, reliable historical data along with predictive models with low uncertainties. In this research, the initial step involved compiling multiple datasets of historical dam break events followed by data cleaning procedures, which included filling in missing data and filtering out anomalous outliers. A data fusion technique was then deployed, an approach that combines data from multiple sources, for three main objectives: i) identify the most appropriate empirical equations for each breach parameter; ii) develop a proposed nonlinear regression-based models informed by the selected empirical equations; iii) conduct two types of uncertainty analyses on the results to assess the reliability and confidence of outcomes.

The newly developed data fusion-based models have shown marked improvement over the existing empirical equations when compared against the developed historical datasets, exhibiting reduced uncertainty across all breach parameters. To demonstrate the practical engineering implications of the findings, the proposed models’ outputs were applied to four real-world dam cases using a two-dimensional hydraulic model, specifically, the Hydrologic Engineering Center’s River Analysis System (HEC-RAS) model. This served two purposes: one, it exemplified the application of proposed models in practical settings, and two, it facilitated a comparison between the outputs of the hydraulic model and the estimates derived from our data fusion model. This comparison aids in verifying the consistency and reliability of our predictive models. The novel approach presented in this study sets a foundation for future advancements in dam breach prediction and risk management.

Buy this resource