PSI - Issue 52

Muhammad Raihan Firdaus et al. / Procedia Structural Integrity 52 (2024) 309–322 M.R. Firdaus et al. / Structural Integrity Procedia 00 (2023) 000–000

321

13

lations. A concise summary of the computational resource utilization for these simulations is conveniently presented inTable 1. Based on the collected data, it is evident that when using the same type of CPU for the static simulations, the full solid modelling approach consumed the highest amount of computational time. On the other hand, the full shell modelling method proved to be the most e ffi cient and least time-consuming. As for the dynamic simulations, various CPU types were employed. However, it is noteworthy that only the full solid modelling employed 24 cores in the High-Performance Computer (HPC). Despite utilizing what could be considered the most powerful CPU within the dynamic simulation category, this model still demanded the longest computational time, taking approximately 20 days to complete. The extensive computational requirements of the full solid modelling necessitate a high number of CPU cores and a substantial amount of RAM, surpassing the capabilities of a regular PC equipped with a core i7-7700 CPU and 16 GB of RAM, as indicated by the inability of the latter to run the simulation according to the status report. Continuing within the same table, it is noteworthy that the computing time for both the multi-stage multi-scale and concurrent multi-scale methods falls within the range delineated by the Full Shell and Full Solid simulations. By taking into account this recorded computational time and resource utilization, along with the simulation results, it becomes apparent that the multi-stage multi-scale and concurrent multi-scale models e ff ectively serve as a bridge between the resource-intensive Full Solid model and the accuracy of the Full Shell model. These two multi-scale models successfully capture the out-of-plane stress characteristics exhibited by the Full Solid model, which necessi tates substantial computing time and resources, while also o ff ering in-plane stress accuracy closer to that of the Full Shell model, which requires comparatively lower computing time and resources. This multi-scale methods resulting in better result outcomes compared to the Full Solid method, while still maintaining acceptable amount of computational expenses. In this paper, we recognize that simulating fluid-structure interaction, particularly involving the hydrodynamic im pact of a float structure as studied here, incurs exceptionally high computational costs. Consequently, it is essential to explore alternative approaches that reduce this burden while still delivering valid outcomes. Thus, we have de vised two methods, namely the multi-stage multi-scale and concurrent multi-scale modelling methods, as resolution strategies for simulating the fluid-structure interaction of an amphibious aircraft float structure. Both the multi-stage multi-scale and the concurrent multi-scale methods, as alternative modelling approaches for FSI in the amphibious aircraft float section, demonstrate their capability to compute complex structures and yield acceptable stress values using reasonable computational resources, all the while maintaining computational accuracy and e ffi ciency. Furthermore, these methods are well-suited for modelling complex structures with numerous regions requiring detailed consideration. Specifically for the concurrent multi-scale method, the pre-computation phase can be cut for a quite plentiful amount, as only one model file is needed to simulate the entire structure. However, there are some disadvantages to consider. The concurrent multi-scale method entails more computational time compared to the multi-stage multi-scale model for each region that needs to be simulated in detail. This is because the region of interest must be modeled using solid elements, which increases the computational load on ABAQUS and extends the overall computational time. Additionally, the shell-to-solid coupling must be defined in all locations where shell edges meet the solid surfaces. When dealing with complex structures that require a large number of detailed regions, defining such couplings becomes impractical and susceptible to operator errors. As a result, the solid submodelling approach may be more convenient. 4. Conclusions