The present paper proposes a novel approach for the representation of the imperfect plate geometry of ships structures and assess its impact on the probabilistic ultimate strength of plates and hull-girders. The description of the imperfect geometry is basically implemented using the theory of random fields. Evidence for the selection and robustness of the proposed model is documented from literature data and real measurement campaigns conducted on ships. A preliminary study is first presented upon the prediction of plates’ probabilistic ultimate strength comparing the efficiency of the proposed model with existing imperfection models from literature. Afterwards, a case study on a VLCC oil tanker for the prediction of hull-girder ultimate strength applying different imperfection models takes place. In order to evaluate the variability of ultimate strength under the effect of stochastic imperfections, artificial neural networks (ANNs), which trained appropriately with relatively limited results of non-linear finite element calculations, are used. The main objective of this research study is to provide a methodology for the evaluation of probabilistic-based ultimate strength assessment of plates and hull-girders in view of stochastic geometric imperfections. In doing so, and incorporating other sources of uncertainties, the designer/engineer should be able to perform a valuable reliability-based analysis of the structure.

本文提出了一种新颖的方法来表示不理想的船板几何形状，并评估了其对板和船体梁概率极限强度的影响。缺陷几何的描述基本上是使用随机场理论来实现的。从文献数据和在舰船上进行的实际测量活动中，可以证明所选择模型的选择和鲁棒性。首先对板的概率极限强度进行了初步研究，将所提出的模型与现有文献中的缺陷模型的效率进行了比较。之后，进行了一个VLCC油轮的案例研究，使用不同的缺陷模型来预测船体梁的极限强度。为了评估在随机缺陷的影响下极限强度的变异性，使用了经过适当训练的非线性有限元计算结果相对有限的人工神经网络（ANN）。这项研究的主要目的是鉴于随机几何缺陷，为基于概率的板和船体梁极限强度评估提供一种方法。在这样做的过程中，并结合其他不确定因素，设计人员/工程师应能够对结构进行有价值的基于可靠性的分析。这项研究的主要目的是鉴于随机几何缺陷，为基于概率的板和船体梁极限强度评估提供一种方法。在这样做的过程中，并结合其他不确定因素，设计人员/工程师应能够对结构进行有价值的基于可靠性的分析。这项研究的主要目的是鉴于随机几何缺陷，为基于概率的板和船体梁极限强度评估提供一种方法。在这样做的过程中，并结合其他不确定因素，设计人员/工程师应能够对结构进行有价值的基于可靠性的分析。