Comparação entre duas abordagens para solução de sistemas de equações resultantes da discretização pelo método implícito adaptativo

Abstract

One critical point of a reservoir simulation is the total simulation time, that, depending on the model, may be days. After a series of discretizations the linear systems to be solved use incomplete factorization as preconditioning method. It is usual the application of matrix ordering to help convergence and decrease the amount of fill-in. In this work, we present the comparison between two preconditioners using the natural and induced orderings when applied to complete simulations of real cases using the adaptative implicit method. The results showed that, when compared to fully implicit simulation cases, the natural ordering preconditioner had a 2.46 speed-up, whereas the induced preconditioner had a 1.77 speed-up.