An Adaptive FN-SPH (AFN-SPH) Method to Computational Efficiency in Simulating Problem Deforming Boundaries.

Abstract

In numerical modelling, both grid-based and meshfree IJlethods have their own advantages as well as disadvantages. However, compared' to grid-based methods, meshfree methods are much capable of simulating largely deforming discrete problem domains. In this regard, Smooth Particle Hydrodynamics (SPH) is a commonly used meshfree method to simulate complex fluid dynamics, but with additional computational effort. As an alternative, the Fixed Neighbourhood based SPH (FN-SPH) has been recently developed by combining the features of both grid­based methods and meshfree methods, to reduce the computational effort compared to the Nearest Neighbour Particle Searching (NNPS) based conventional SPH approach. Here, only a predefined portion of the whole problem domain is searched in each instance to find the neighbouring particles, rather than searching the whole problem domain for neighbours as in NNPS. However, the FN-SPH technique is only applicable when there is much limited deformation in the fluid domain due to the fixed-neighbourhood nature of the FN-SPH. It restricts the use of FN-SPH in generic fluid engineering applications involving large deformation. As a solution, this research investigated the applicability of a novel approach which is introduced as, the Adaptive FN-SPH (AFN-SPH), where the fixed neighbourhood of FN-SPH gets timely updated in order to accommodate deforming problem domains. Accordingly, mechanical compression of a single plant cell was simulated using the AFN-SPH for both fresh and dried conditions in a High Performance Computer (HPC). It was observed that AFN-SPH can reduce the computational time by 40% over the conventional SPH or FN-SPH solutions for the same problem, while reliably reproducing the original solution, both qualitatively and quantitatively. Accordingly, the prosed adaptive FN-SPH approach can be recommended as a computationally efficient, yet versatile meshfree based modelling approach applicable for problem domain with deforming boundaries.