Ship Flooding Time Prediction Based on Composite Neural Network

When a ship sailing on the sea encounters flooding events, quickly predicting time of compartments in damaged area is beneficial to making evacuation decisions and reducing losses. At present, decision-makers obtain data through various sensors arranged board predict compartment flooding. These help with calculation emergency situations. This paper proposes new approach obtaining time. Specifically damage scenarios, based Convolutional Neural Network Recurrent (CNN-RNN), using composite neural network framework estimates when compartment’s water reaches target height. The input images compartment. Transfer learning utilized paper. ResNet18 model Pytorch used extract spatial information from images. Long Short-Term Memory (LSTM) then applied Experimental results show that, for compartment, predicted by 85% accurate while others’ accuracy more than 91%. Intuitively, it comes actual event, network’s average prediction error approximately 1 min. To summarize, these suggest that proposed above can provide assist