| 基于贝叶斯网络的内河船舶碰撞人的失误分析 |
| |
| 引用本文： | 陈亚东,张延猛,胡昊.基于贝叶斯网络的内河船舶碰撞人的失误分析[J].上海海事大学学报,2017,38(3):41-46. |
| |
| 作者姓名： | 陈亚东 张延猛 胡昊 |
| |
| 作者单位： | 上海交通大学船舶海洋与建筑工程学院,上海交通大学海洋水下工程科学研究院,上海交通大学船舶海洋与建筑工程学院 |
| |
| 摘 要： | 为分析对船舶航行安全有显著影响的人的因素,引入SwainGuttmann模型和贝叶斯网络分析内河船舶碰撞事故。依据SwainGuttmann模型的信息处理过程(感知、决策和行动)梳理内河船舶碰撞形成机理。利用领域专家知识和小样本量数据学习法构建贝叶斯网络,计算网络节点间的条件概率。利用贝叶斯网络的不确定性知识推理方法,得出影响船舶航行安全的关键人的因素。该结果与真实事故案例统计结果相吻合,这说明SwainGuttmann模型和贝叶斯网络适用于对内河船舶航行安全的综合分析。
|
| 关 键 词： | 航运安全 内河船舶碰撞 贝叶斯网络 最大期望算法 安全风险 人的失误 |
| 收稿时间： | 2016/10/9 0:00:00 |
| 修稿时间： | 2016/12/22 0:00:00 |
Human error analysis of inland ship collision based on Bayesian network |
| |
| Institution: | School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University,Chinese Underwater Technology Institute, Shanghai Jiao Tong University and School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University |
| |
| Abstract: | In order to analyze human factors that have a significant impact on the ship navigation safety, the Swain & Guttmann model and Bayesian network are proposed to analyze inland ship collision accidents. The collision mechanism is summarized according to the information processing procedure (perception, decision making and action) of the Swain & Guttmann model. The Bayesian network is established with the combination of the domain expert knowledge and the learning method with small amount of sample data, and the conditional probabilities between the network nodes are computed. The main human factors, obtained using the uncertain knowledge inference method of the Bayesian network, are consistent with the statistical results from real accident cases, which reveals that the Swain & Guttmann model and Bayesian network are applicable for comprehensive analysis of inland ship navigation safety. |
| |
| Keywords: | navigation safety inland ship collision Bayesian network expectation maximization algorithm safety risk human error |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《上海海事大学学报》浏览原始摘要信息 |
| 点击此处可从《上海海事大学学报》下载免费的PDF全文 |
|
