A Maritime Accident Forecast Based On BP Neural Network And Carried Out By MATLAB

With the development of maritime trade, maritime accidents often take place because of more and more watercraft. It is an important part for maritime traffic safety that exact and effective maritime accident forecast. Our country is a great shipping country. Exact maritime accident forecast will reduce personnel casualty and property damnification.The forecast theory is explained first in the thesis, which is based on BP neural network theory. A good many factors that effecting maritime accidents are analyzed and are disposed by quantity. Backpropagation neural network are composed of a good many simple elements operating in parallel. The cores of Backpropagation Neural Network are the capacity of parallel computing, self-studying, fault-tolerant and nonlinear function approximating. Input vectors and the corresponding target vectors are used to train a network until it can approximate a function, so that the model is established. At present, during the course of BP neural network's learning and training, we often adopt the algorithm of back-error propagation , which is based on global error function's gradient descent. In order to solve the Bp neural network's inherent deficiency of slowly converging and easily falling into local minimum, we propose to use an algorithm combined self-adapting learning rate and extra momentum. Environment sea-route traffic density ship sailor and management by oneself or synthetically act on maritime accident, which composed by human-ship-environment system. Some factor does not assort with system by oneself movement, so results in maritime accident. A good many factors that effecting maritime accidents are analyzed and are disposed by quantity in the thesis.But, to establish BP neural network model, you must have rich program experiences, which hampers the development of neural networks in a way. An available Neural Network Toolbox (NNT) in MATLAB, however, solves the problem. In the thesis, how to use the backpropagation training functions in theMATLAB toolbox to train feedforward neural networks to solve specific problems is explained. Besides that, in order to make up the MATLAB alternation deficiency of man and computer, how to design the interface between MATLAB and VC++ is presented.Through the thesis, we can see that it's easy and effective to make BP neural network model by using the Neural Network Toolbox (NNT) based on MATLAB and VC++ to evaluate the predictive model. We adopt a three-layer BP neural network. Input-layer contains seven nerve cells (sailor average maritime experience, average shipping years , average standard wind days > average river velocity , visibility, water depth and traffic density) and output-layer contains four nerve cells (great accident, large accident, common accident and small accident).The test that use the data of historical occurrence level of maritime accidents in ChongQing water area is done, the result shows the feasibility of the predictive model. So it's valuable and has a bright future.The thesis includes five chapters. In the first chapter, it is the preface having stated the background, the purpose and meanings of the study and so on. In the second chapter, the theories of maritime accidents forecast based on BP neural network are talked about. In the third chapter, it is the maritime accidents forecast MATLAB programme part based on BP neural network. In the fourth chapter, the design of the interface, some functions of maritime accidents forecast system and emulate experiment are explained. In the last chapter, the conclusion and further work are talked out. Maritime accidents forecast system will advance the level of maritime accidents forecast, if it is used. However, its interface and functions need improving on account of the limited time and capacity of the author.