No abstract

Abstract The proposed study investigated the applicability of the finite volume method (FVM) based on the Godunov scheme to transient water hammer with shock front simulation, in which intermediate fluxes were computed using either first-order or second-order Riemann solvers. Finite volume (FV) schemes are known to conserve mass and momentum and produce the efficient and accurate realization of shock wave. The second-order solution of the Godunov scheme requires an efficient slope or a flux limiter for error minimization and time optimization. The study examined a range of limiters and found that the MINMOD limiter is the best for modeling water hammer in terms of computational time and accuracy. The first- and second-order FVMs were compared with the method of characteristics (MOCs) and experimental water hammer measurements available in the literature. Both the FV methods accurately predicted the numerical and experimental results. Parallelization of the second-order FVM reduced the computational time similar to that of first-order. Thus, the study presented a faster and more accurate FVM which is comparable to that of MOC in terms of computational time and precision therefore a good substitute for the MOC. The proposed study also investigated the implementation of a more complex convolution-based unsteady friction model in the FVM to capture real pressure dissipation. The comparison with experimental data proved that the first-order FV scheme with the convolution integral method is highly accurate for computing unsteady friction for sudden valve closures.

Here, recent developments in the key numerical approaches to water hammer modelling are summarized and critiqued. This paper summarizes one-dimensional modelling using the finite difference method (FDM), the method of characteristics (MOC), and especially the more recent finite volume method (FVM). The discussion is briefly extended to two-dimensional modelling, as well as to computational fluid dynamics (CFD) approaches. Finite volume methods are of particular note, since they approximate the governing partial differential equations (PDEs) in a volume integral form, thus intrinsically conserving mass and momentum fluxes. Accuracy in transient modelling is particularly important in certain (typically more nuanced) applications, including fault (leakage and blockage) detection. The FVM, first advanced using Godunov’s scheme, is preferred in cases where wave celerity evolves over time (e.g., due to the release of air) or due to spatial changes (e.g., due to changes in wall thickness). Both numerical and experimental studies demonstrate that the first-order Godunov’s scheme compares favourably with the MOC in terms of accuracy and computational speed; with further advances in the FVM schemes, it progressively achieves faster and more accurate codes. The current range of numerical methods is discussed and illustrated, including highlighting both their limitations and their advantages.

No abstract

No abstract