Ultimate Strength Assessment of Cracked Stiffened Plates Using Optimized XGBoost Method

Assessing the ultimate strength of the stiffened plates forming the ship structure is the first step in assessing its ultimate strength. Over time and increase the life of the structure, failures such as cracks reduce the load-bearing capacity of the structure. The main purpose of this paper is to present a machine learning method based on XGBoost algorithm to calculate the ultimate compressive strength of stiffened plates with crack failure using the results of multiple finite element analyzes. To achieve the best possible results from the XGBoost algorithm, some of the hyperparameters in this algorithm have been optimized using the Bayesian optimization method. The results of this method show that the accuracy of using the optimized XGBoost algorithm is much higher than conventional methods based on linear regression.