Using a PANS Simulation Approach for the Transient Flow Around the Japan Bulk Carrier

Abstract

In this study, a partially averaged Navier-Stokes (PANS) modeling approach developed based on the k-ω turbulence model has been applied to the flow around the Japan Bulk Carrier. Three different meshes have been used for a fixed physical resolution for the PANS modeling. The results are encouraging, with small-scale flow dynamics being allowed to develop on reasonably small mesh sizes, but more studies are required before reliable predictive simulations can be performed. 1. Introduction Steady resistance and wake simulations in ship hydrodynamics are currently being performed with relatively high reliability using the Reynolds averaged Navier-Stokes (RANS) simulation approach. There are, however, a number of cases where RANS is not sufficient and a scale-resolving transient simulation approach is needed. This is well established in cases of flow around bluff bodies, e.g., vortex-induced vibrations and flow across cylinders (Pereira 2018), or for ship to ship interaction (Arslan et al. 2015), where the main interest is on global parameters, such as shedding frequency or force variation, but the transient separating flow is not well predicted by RANS simulation. It could also be that details of the transient flow behavior is part of the study, e.g. relating to propeller dynamics (Jang & Mahesh 2013; Bensow 2015; Guilmineau et al. 2017), vortex dynamics (Fureby et al. 2016; Visonneau et al. 2018), or in cavitation (Bensow 2011); in these cases, RANS modeling is inherently not able to capture the details of the governing flow physics because of the underlying averaging. However, it may also be the case where the flow is smoothly separating, or close to, as is the case for the Japan Bulk Carrier (JBC), see Visonneau et al. (2016) and further the discussion below. Using a wall-resolved large eddy simulation (LES) approach is computationally very expensive, with an unfavorable scaling toward full-scale Re numbers (Liefvendahl & Fureby 2017), thus the attention has rather turned toward hybrid RANS/LES approaches or wall-modeled LES; this was used in many of the studies referenced previously, e.g., Guilmineau et al. (2017); Pereira (2018); Visonneau et al. (2018). Grid size estimates for different modeling approaches for simulating the JBC in model scale are given by Liefvendahl and Johansson (2018), amounting to around 2•109 for the wall-resolved LES, 150•106 for the wall-modeled LES, 50•106 for the hybrid RANS/LES, and finally 30•106 for the RANS approach, where the distance between wall-resolved LES and the others will greatly increase with the Re number. In terms of computational effort, one needs to remember that the transient flow approaches, compared with RANS, further needs to include long time-series for sampling statistics.