The main purpose of this study is to apply a Deep Neural Network (DNN) method to linear systems and to predict in a relatively short time span the ultimate vertical bending moment (VBM) for damaged ships. A Deep Neural Network approach, which is composed of multiple fully connected layers with a Rectified Linear Unit (ReLU) which is a non-linear activation function, has been applied to more than 6000 samples and validated using leave-one-out technique. The ultimate strength has been predicted for a set of completely new damage scenarios of different cross sections, enhancing that the deep neural network method can estimate the residual hull girder strength for a correlated damage index general (DIG). The predicted residual hull girder strength as well as the shift of the neutral axis are validated against Smith’s method-based results.
A Deep Neural Network to Predict the Residual Hull Girder Strength
La Ferlita, Alessandro , Di Nardo, Emanuel , Macera, Massimo , Lindemann, Thomas , Ciaramella, Angelo , and Nikolaos Koulianos. "A Deep Neural Network to Predict the Residual Hull Girder Strength." Paper presented at the SNAME Maritime Convention, Houston, Texas, USA, September 2022. doi: https://doi.org/10.5957/SMC-2022-074
Download citation file: