The campaign length of a blast furnace (BF) is limited by the hearth inner lining lifetime. In order to maximize it and ensure a good draining of hot metal and slag, understanding the flow in the hearth is essential. Despite its necessity, information about the state of the hearth is limited due to a lack of accessibility and direct measurements during operation. Thus, numerical methods are valuable tools for such investigations. Challenges in modeling the flow involve several continuous phases (hot metal, slag and hot blast) as well as the presence of the deadman, a dense bed of coke particles. Although the deadman takes on various shapes and positions which depend on several operational conditions (e.g. the weight of the burden, the deadman porosity as well as the liquid levels in the hearth), and thus shows dynamic behavior, previous research treated the deadman as a simplified static porous medium. To account for the deadmans transient characteristics, we present in this work a computational fluid dynamics (CFD) - discrete element method (DEM) coupled method. A volume of fluid (VOF) approach is utilized to model the multiple fluids, while the DEM is used to describe the deadman by treating the coke particles as discrete entities, resulting in a complete four-way coupled model. The CFD-DEM model was validated on a few selected cases where well-documented solutions could be found. Additionally, in analogy to the BF tapping, we performed lab-scale experiments on the drainage of water through sitting and floating particle beds, which served as additional validation to the model. Furthermore, the CFD-DEM model was employed and demonstrated on a small-scale hearth, where we compared it to a static, floating particle bed and highlighted how the dynamic deadman alter the drainage behavior. ^Although the CFD-DEM model offers several advantages, it comes at a significantly higher computational cost than the Eulerian approaches. Especially, if we consider the huge industrial-scale BFs, scale-up of the model becomes a serious concern due to not only the large spatial scale variations present, but also the large separation of temporal scales (from particle collisions of fractions of a second, to several hours for a single tapping cycle). Attempting to address the issues of spatial and temporal scale separations, we introduce a model intended to bridge the gap between the computationally cheap Eulerian methods and the expensive but accurate CFD-DEM ones. By utilizing the CFD-DEM model to provide spatially detailed, inhomogeneous porosity distributions of the deadman, and by relating them to various liquid levels in the hearth, we could describe the dynamic deadman in a simplified manner within a Eulerian framework. This model, referred to as the dynamic void fraction model, was validated by comparisons with the previously obtained CFD-DEM simulation results of the experimental setup. Ultimately, it enabled us to successfully simulate several hours of real-time for a full-scale BF hearth. Moreover, it was reported in literature that natural convection due to density variations in the hot metal significantly alters the flow field as compared to the isothermal conditions. In order to account for this effect, we extended the dynamic void fraction model by incorporating fluid and solid temperatures, and subsequently a simplified model for the natural convection.