摘要
为解决航舵故障诊断的复杂非线性模式分类问题,提出一种基于自组织特征映射(SOM)神经网络的航舵故障诊断方法,构造一个2层SOM神经网络,训练后多个权值向量位于输入向量聚类中心,实现快速有效的自适应分类。仿真结果表明:SOM网络经过100次训练即可实现聚类,对有限故障测试样本分类准确率可达90%,对航舵故障诊断具有一定的参考价值。
It is difficult to collect inductive fault patterns of the steer system for diagnosis. In order to solve such a complicated nonlinear pattern classification problem, a kind of fault diagnosis method for steer system is proposed based on the Self-Organizing feature Map(SOM) neural network structure. A two- layer SOM neural network is built. Many weight vectors are in the clustering center of input vector after training. The adaptive classification is realized fast and effectively. The simulation results show that the SOM network can realize clustering after 100 times of training with the accuracy rate up to 90% for the classification of finite fault test samples.
出处
《信息与电子工程》
2012年第3期339-342,共4页
information and electronic engineering
关键词
自组织特征映射
人工神经网络
故障诊断
航舵
Self-Organizing feature Map
artificial neural networks
fault diagnosis
nautical steer