The city of Cali is the main urban center of the south-western part of Colombia. It is located along the Cauca river, the second largest in the country. At the moment, the embankment located along the left margin that protects the city is designed for a return period of 100 years. In this study is intended to make a probabilistic analysis for dike failure for the present state of this flood defence. The water levels inside the main stream are highly sensitive to the peak discharge of upstream near tributaries which must be included in the analysis. An existent hydrodynamic model can be used for the estimation of the possible water levels but it requires large amounts of input data and is highly time consuming. Simulation of hydrodynamics and geotechnical stability models are quite advanced right now but for reliability studies might not be the most efficient solution. Therefore meta-modeling procedures could be implemented if accurate and sufficient data is available. For this study, different data-driven surrogate models where coupled as a solution for flood defense failure calculation tools. These models where used to estimate the loads and resistance combined effects of the actual flood defense for several different failure mechanisms. Algorithms like ANN and M5P where applied for building the data-driven surrogate models for each levee failure mechanism. Calculation time was reduced sufficiently to make it a feasible solution for this kind of studies. The final models can also be used as design tools if their application ranges is sufficiently understood.
Aguilar, Juan; Warmink, Jord; Schielen, Ralph; and Hulscher, Suzanne, "Data-Driven Surrogate Models For Flood Defence Failure Estimation : “Jarillon” Of Cali, Colombia Case Study" (2014). CUNY Academic Works.