摘要
选用合适的训练、选择BP神经网络结构、连接权系数的方法和船舶实航数据,建立BP神经网络.用同一艘船的另两段实航数据验证该神经网络的泛化性能,在其中一段数据中人为加入缓变故障信号,用以检验其对缓变故障的敏感性.结果表明,系统能够跟踪同一条船不同航行状态的动态特性,对缓变故障也相当敏感,适用于一般海况船舶正常航行时平台罗经故障检测.
A certain method and a set of data from a ship when it voyaged over the East China Sea are used to determine the input delay and the numbers of hidden units of BP neural networks,and train them. And then,the generalization of BP neural networks is verified by making use of the other two sets of data of the same ship in different voyages,and the sensitivity to slow malfunction of stabilized gyrocompass is verified by adding factitious slow fault in one of the two sets of data. The determined BP neural networ...
出处
《郑州大学学报(理学版)》
CAS
北大核心
2009年第3期30-33,49,共5页
Journal of Zhengzhou University:Natural Science Edition
基金
河南省自然科学基金资助项目
编号0411012700
关键词
平台罗经
故障检测
神经网络
stabilized gyrocompass
fault detection
neural networks