Author(s): V. Christelis; V. Bellos; G. Tsakiris
Linked Author(s): Vasilis Bellos
Keywords: No keywords
Abstract: In this paper, the automatic calibration of the in-house FLOW-R2D hydrodynamic model is addressed, for a real world flood event created after the Tous dam break (Spain). The Manning coefficient in the entire computational domain and the effective slope of the upstream boundary cross-section (for deriving the upstream boundary steady flow condition) are selected for calibration. Due to the long simulation run times of hydrodynamic models, conventional optimisation techniques cannot be easily implemented for automatic calibration while the commonly applied trial and error method has several limitations. To cope with the excess computational burden, a surrogate-assisted optimisation algorithm (namely, MLMSRBF) is employed in this paper. To set a benchmark solution, a FLOW-R2D-based automatic calibration is also performed and the results are compared to the MLMSRBF method. Results indicate that for a computational budget of 100 FLOW-R2D simulations, both the conventional and the surrogate-assisted algorithm perform equally well while they outperform the traditional trial and error approach.