Research On Ship Risk Assessment Based On The BP Neural Network

As an important carrier of maritime transportation, ship is a complicated and huge system engineering. Modern ship is not only integrated by most advanced communications, navigation, machinery, automation and other high-techs, but also equipped with high-qualified seafarers, rich sailing chart data and high-level management system. Ship operation and cargo transportation by sea constitute a high-value commercial activities. Another attribute of marine transportation is of high risk, personnel on board, cargo, machines and energy systems occupy the limited space on board ship. According to various accident causation theory, whether it is "Hayne theory", "energy release" or "matter of insecurity and human unsafe acts" and so on, the interference between different elements on board ship easily leads to accidents. In addition, it is inevitable to avoid maritime accidents sailing across the narrow waterways such as reef, shoal, canals, as well as sailing in the wind, waves, torrent or fog. The consequence of accidents brought ship owners, shipping companies, cargo owners and insurance companies huge economic losses, and what’s worse, severe chemical leakage or oil spill caused ecological disaster which will not be restored in short-term or long period of time.Based on research of evaluation index system and assessment method of ship risk in and abroad, analysis of the current deficiencies of ship overall risk assessment method and reference to the application of BP neural network in Shanghai ship oil spill risk assessment (2009) and the ship Collision Risk Assessment (1998), this paper proposed ship overall risk assessment method on the basis of BP neural network model. The method is of highly non-linear line approximation, fuzzy inference, massively parallel processing, self-training study, self-organized and relatively good fault tolerance and so on. To some extent, it avoided the interference of some subjective and random factors, meanwhile, the comparison output risk value of BP neural network model with the expected or previous ship overall risk value, verified BP neural risk assessment of effectiveness and operability in ship assessment field.This paper selected 5 tankers owned by a Tanker Corporation as a sample, also invited 10 experienced Captains who are serving in this company to complete relative questionnaires. Then, data extracted from the questionnaires are being trained and processed via BP neural network theory by MATLAB software tools made the model error reach the predetermined rang inner. Making risk assessment of M/V Bailuzhou provided a basis of risk control and decision-making for the further steps by means of well-trained BP network assessment model.