Young-Moo Son, Dong Woo Kim, Sung Gyu Jeon, Yong Tai Kim, Hoon Kyu Oh, Jin-Hyeok Kim, In-Chang Yeo, In-Su Han, Myung-Il Roh, "A Computational Framework for Automation and Optimization Algorithms for Structural Assessment and Design Development of Pump Tower", Proceedings of ISOPE 2026, Orlando, USA, 2026.05.31-06.05
| Abstract | LNG (Liquefied Natural Gas) ships are essential to guarantee both durability and safety during their operational life cycle. Pump Tower is a structure that enable the loading and unloading of LNG within cargo tanks. In general, It is a tripod structure composed of multiple bundled pipes, which is installed from the tank top to the bottom. To ensure the safety of pump tower, evaluations of strength, fatigue, and vibration are required, following the standards provided by class or API. Pump Tower in LNG cargo tank experiences two principal forces: thermal and sloshing load. Thermal load is influenced by the LNG filling level in the cargo tank, whereas the sloshing load is governed by the ship’s motion. The ship’s motion can be obtained through CFD analysis, and the output consists of extensive time-dependent data. Therefore, the process is time-consuming due to the consideration of all load cases and the screening of all pump tower components, consequently extending the duration of the structural safety evaluation and further constraining the development of an optimized design incorporating multiple variables. Among the numerous seconds of data, the time intervals required by the class society can be extracted based on the mass distribution of the pump tower. From the CFD data, the calculated structure loads applied to the pump tower and evaluates the safety of pump tower: strength, fatigue and resonance. All these processes are carried out accurately and efficiently through the implementation of an automated algorithm. Along with the automated algorithm, analysis data have been accumulated to construct a surrogate model capable of predicting pump tower performance. The surrogate model is an ensemble model combining a maximum value classification model and a distribution parameter regression model, implemented using the XGBoost (eXtreme Gradient Boosting) model. Using the model, NSGA-II based global optimization allows for forecasting optimal designs in different ship types and maritime conditions, substantially decreasing the time and effort associated with lightweighting and optimization. With the incorporation of constraints such as production feasibility and safety regulations, a practically applicable pump tower model for shipyard can be developed. |
|---|---|
| Publication Date | 2026-06-01 |
