摘要
针对神经网络在故障测试中因训练样本集大而导致的诊断速度慢和性能下降等不足,提出区域法则,即将全系统化解为若干个子模块并用神经网络去分别匹配,以此达到加快诊断速度和提高诊断能力的目的,同时给出了典型训练样本的建立思路和网络结构的学习算法.实验表明,该方法在火炮随动系统测试中体现出了较好的诊断效果,具有一定的工程应用价值.
This article indicates that the fault diagnosis technology of using the local area rules and neural net may detect the ship-gun circuit faults. An approach about how to construct the typical training samples and the learning arithmetic of the neural net was presented. The result of experi ments show that the diagnosis is of good effect when the tolerative differences of the elements are bigger.
出处
《海军工程大学学报》
CAS
北大核心
2006年第6期79-83,共5页
Journal of Naval University of Engineering
