The research of accident of the vessel with Artificial Neural Networks

• Main Authors: Cei-We Wu, 吳志文
• Other Authors: Tso-Ting Lin
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/wmq3eg

碩士 === 國立成功大學 === 交通管理學系碩博士班 === 90 === Abstract A considerable change in size and structure of marine traffic has taken place in the last two decades. This is mainly due to the jerky rise of cargo transport across the sea and new technical developments in the recent past. The number of tankers, for example, has doubled from 1960 to 1980 and their tonnage has increased sevenfold. New ship types have come into the world such as fast container ships and medium-speed vessels for the carriage of dangerous cargoes such as gas and chemical tankers. In addition, numerous structures for oil and gas production have been erected at sea which only serve to complicate marine traffic flow.With the growth of traffic density it has been assumed that the dangers of an accident would also increase. Accidents serve as an operational measure of marine safety, and specifically the safety of vessels, crew, and cargoes. The ability of Neural Network to accurately predict the type of vessel accident with the categorical input variables of time, season, position, weather, pilot, tonnage, and traffic could significantly reduce marine casualties by alerting port authorities and navigation groups. The performance of ANN model was affected by the elements including “training ＆ testing set”, “input variable”, “hidden layer”, “learning rule”, transfer function”, and “momentum factor” ect. These elements were considered to develop the best ANN model. The variables were selected by the contribution graph approach, and the multiple neurons and single neuron ANN models were developed to compare the similarities ＆ differences. This reseach shows the types of vessel accidents correct rate can achieve 85.78﹪（collision 96.55﹪、others(grounding,sinking) 75.00﹪）for multiple neurons and 72.41﹪（collision 94.83﹪、others(grounding,sinking)50.00﹪）for single neuron.