Maritime inspection analysis is recently an emerging topic. Practical solutions are sought to improve the post-inspection process (i.e. OCIMF SIRE) in tanker operations. This study aims at to analyse the reported observations based on natural language processing (NLP) integrated with artificial neural network (ANN). The developed methodology also takes the advantage of Maritime Root Cause Analysis Tool (MARCAT) to systematically initiate the potential causes database including the number of 2383 observations. Then, an NLP-based ANN classification algorithm was produced that predicts the causes of new entries to the inspection database. The classification algorithm gives high-accuracy results varying between 0.90 and 0.97 in different causation segments. An inspection analysis for prediction validation was carried out both by the company authority and by the algorithm. The results show that the model can predict with a high success rate. Consequently, this study developed a post-inspection analysis model with high accuracy that is expected to contribute maritime executives to improve fleet safety and efficiency. Providing a third-party solution to the tanker industry, further studies are planned to conceptualise the model as platform as a service (PaaS).
A deep learning approach to analyse ship inspection reports via natural language processing integrated with artificial neural network
Journal of Marine Engineering & Technology ; 23 , 4 ; 291-304
2024-07-03
14 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Natural language processing-based approach for automatically coding ship sensor data
DOAJ | 2024
|An Artificial Neural Network Model for Preliminary Ship Design
British Library Conference Proceedings | 1994
|NATURAL FREQUENCIES OF SUSPENSION BRIDGES: AN ARTIFICIAL NEURAL NETWORK APPROACH
Online Contents | 2002
|Consistently Trained Artificial Neural Network for Automatic Ship Berthing Control
DOAJ | 2015
|Identifying mathematical model of ship movement using artificial neural network
Tema Archiv | 2000
|