Optimal Dimensions of Post-Tensioned Concrete Cylindrical Walls Using Harmony Search and Ensemble Learning with SHAP

The optimal design of prestressed concrete cylindrical walls is greatly beneficial for economic and environmental impact. However, the lack available big enough datasets training robust machine learning models one factors that prevents wide adoption techniques in structural design. current study demonstrates application well-established harmony search methodology to create a large database configurations. unit costs steel used construction, specific weight stored fluid, height wall are input variables whereas optimum thicknesses with without post-tensioning output variables. Based on this database, some most efficient ensemble like Extreme Gradient Boosting (XGBoost), Light Machine (LightGBM), Categorical (CatBoost) Random Forest algorithms have been trained. An R2 score greater than 0.98 could be achieved by all models. Furthermore, impacts different features predictions analyzed using SHapley Additive exPlanations (SHAP) methodology. was found greatest impact thickness, followed fluid. Also, help individual conditional expectation (ICE) plots variations predictive model outputs respect each feature visualized. By genetic programming methodology, equations obtained thickness.