An Optimized and Scalable Algorithm for the Fast Convergence of Steady 1-D Open-Channel Flows

Abstract

Calculating an open-channel steady flow is of main interest in many situations; this includes defining the initial conditions for the unsteady simulation or the computation of the water level for a given discharge. There are several applications that require a very short computation time in order to envisage a large number of runs, for example, uncertainty analysis or optimization. Here, an optimized algorithm was implemented for the fast and efficient computation of a 1-D steady flow. It merges several techniques: a pseudo-time version of the Saint-Venant equations, an evolutionary domain and the use of a non-linear Krylov accelerator. After validation of this new algorithm, we also showed that it performs well in scalability tests. The computation cost evolves linearly with the number of nodes. This was also corroborated when the execution time was compared to that obtained by the non-linear solver, CasADi. A real-world example using a 9.5 km stretch of river confirmed that the computation times were very short compared to a standard time-dependent computation.