小样本情况下的船舶溢油事故风险评价研究  

Study on ship oil-spill risk assessment based on small samples

作　　者：张欣[1] 

机构地区：[1]上海海事大学交通运输学院,上海200135

出　　处：《船舶工程》2009年第2期76-80,共5页Ship Engineering

基　　金：上海市基础研究重点项目(05JC14073)

摘　　要：船舶溢油风险评价是一项复杂的多因素问题,是船舶溢油应急管理的关键环节.作为智能搜索算法的代表理论,BP神经网络被认为是进行不确定风险评价的较好方法之一,然而船舶溢油事故属于小样本事件,统计数据往往难以满足BP神经网络要求的样本容量.针对这一困境,首先提出一种利用B样条最小二乘理论的数据拟合法,显著增加样本数.其次,根据船舶溢油特点建立了基于BP神经网络的船舶溢油风险评价模型.最后以上海港近年发生的10起溢油事故为实例,检验了模型的可行性.Risk assessment of ship oil-spill is a complex multi-factor issue, which plays a key role of ship oil-spill emergency response. As the classical theory of intelligent search algorithm, BP neural network has been regarded as one of the preferred methods to solve risk assessment problem with uncertainty. However ship oil-spill accidents belong to small-sample events and data collection hardly satisfies sample requirements of BP neural networks. In order to conquer this problem, the paper firstly presents B spline least square method combined with BP neural network to increase the amount of samples. Secondly, according to the features of ship oil-spill, a risk assessment model based on BP network is established At last, ten accidents of Shanghai Port are taken as assessment examples to test the feasibility of the model.

关 键 词：船舶溢油 风险评估 小样本 BP神经网络 B样条最小二乘拟合 

分 类 号：U698.7[交通运输工程—港口、海岸及近海工程]