In recent years, the demand for minerals and rare-earth elements are escalating due to rapid technological advancements and developments. This condition raises the importance of Deep-Sea Mining (DSM) as an option to fulfill the global demands, for the sake of future ambitious projects. On the other hand, DSM still faces some drawbacks and obstacles in its operations, e.g. environmental impact of its tailings discharge. Thus, the presence of tools for predicting the behavior and environmental impact of DSM tailings becomes crucial for the sake of conducting sustainable DSM operations. Researches, both numerical and laboratory experiments, are then done to achieve this goal. In the study of DSM tailings behavior through numerical simulation, the challenge lies in the ability to implement the complex physics phenomena around DSM plumes to a numerical model. This research is thus aimed to observe one of the parts of the so-called physical phenomena, and numerical constraints on the simulation of DSM plumes: the effects of implementing arbitrary non-orthogonal mesh. Arbitrary non-orthogonal mesh would give users the freedom to refine the mesh based on the required resolution on a certain area in the simulation domain. In this research, an arbitrary non-orthogonal mesh is constructed forming a domain geometry of 3D tank with round pipe as a source of tailings discharge, adapting the laboratory setup of Byishimo (2018). Mesh refinement is done around the pipe discharge area, where the jet mainstream expected to occur, and near the bottom, where settling of solids fraction and high velocity-gradients are expected to occur. Top-hat approximation theory is then used for defining the inlet boundary condition of the simulation domain, and smooth solid wall is used for the bottom boundary. In observing the effect of arbitrary mesh, two parameters are varied: solids settling rate, and differentiation scheme. Following this, six simulation cases are prepared, containing three solids-settling conditions (minimum, realistic, and extreme solids fraction settling) and two differentiation schemes (Gauss Gamma and Gauss Linear). These cases are then simulated using the CFD drift-flux model in OpenFOAM for two incompressible fluids, with ambient fluid and tailings mixture as the two incompressible fluids. Turbulence is modeled using LES method, with Wall-Adapting Local Eddy-viscosity (WALE) LES model for modeling the subgrid-scale of the turbulence. Two simulated field variables are picked to be observed and compared with the laboratory measurement data: flow velocities and local Suspended Solids Concentration (SSC). Simulations show that the constructed domain able to generate stable results using not only Gauss Gamma differencing scheme but also Gauss Linear, which originally expected to give unbounded results. From simulations with various solids settling conditions, it is analyzed that the implementation of solids settling using the mentioned function leads to constant settling rate with inability to re-suspend the settled solids. Thus, simulation of cases with extreme solids settling leads to hyper-concentration of solids fraction in the cells on the bottom. The top-hat profile of the inlet boundary condition is constantly sustained throughout the simulation, resulting in uniform jet vertical velocity profile. The simulations also show that momentum-driven jet-like flow can be observed around the impingement point, while gravity current generated further from the impingement point. The simulation results show that the mesh refinement is enabling simulating and validating the flow with the required resolutions. Moreover, the constructed domain also shows that Gauss Linear can be used for simulating DSM plumes, furthermore using full tank domain. The simulated SSC also turns out not only affected by the simulation domain and mesh, but also the differencing scheme, and the presence of settling and pick-up functions.