摘要
在基于神经网络的平台罗经故障检测中,为了提高故障检测灵敏度,根据船载平台罗经故障检测的特点,提出了以模糊逻辑和指数加权平均处理估计误差的故障检测方法,并用实船航行数据仿真。该方法对未知输入等于扰不敏感而对故障敏感,且可根据故障的大小自动调节检测时间的长短。对不易检测的小故障,自动延长检测时间以利用更多的信息从而提高检测的正确率;对手较大的故障,自动缩短检测时间从而减少检测延时和累积误差。
In this article, a certain kind of method for stabilized gyrocompass failure detection was brought forward in order to improve the sensitivity of failure detection with high noises, which processes estimation errors with fuzzy logic and exponent-weighted average. And simulation was performed with real stabilized gyrocompass readings during a voyage. The failure detection method is sensitive to errors other than disturbance such as system' s unknown input. Furthermore, the method may adjust the length of detection time adaptively. When the error magnitude is small and intermixed with noises, the detection time is increased to make use of more information, which results in higher probability of correct detection. When the error magnitude is bigger, the detection time is decreased, so that the detecting delay is decreased and the error accumulation is reduced.
出处
《信号处理》
CSCD
北大核心
2008年第6期1044-1047,共4页
Journal of Signal Processing
基金
河南省自然科学基金资助项目(编号:0411012700)
关键词
神经网络
模糊逻辑
平台罗经
故障检测
stabilized gyrocompass
failure detection
neural networks
fuzzy logic