摘要
针对一类模型不确定的非线性系统,提出了具有强鲁棒性和高灵敏度的在线故障检测与诊断方法。其中,系统只有输入、输出可检测,故障是关于输入和状态的非线性函数。将RBF神经网络和频谱分析相结合,由RBF神经网络来学习及存储电子电路的故障频谱和故障类型之间的映射关系,介绍了该算法的实现过程。并以某船舶电气设备放大电路为例建立仿真系统。仿真结果和实验实例表明,该算法可以快速有效地对故障元件进行定位,识别率较高。
Aiming at a kind of non-linear system whose model is uncertain, on-line fault diagnosis method which is robust and sensitive is put forward in this paper. Only the inputs and outputs of system can be detected. The fault is a non-linear function of inputs and states. RBF neural network is combined with frequency spectrum analysis. The mapping relationship between frequency spectrum of faulty circuit board and faulty form is studied and stored by RBF neural network. The procedure of this algorithm is introduced. An amplification circuit of certain watercraft electric equipment is simulated as an example. Simulation result indicates that this algorithm is efficient for fault diagnosis.
出处
《继电器》
CSCD
北大核心
2007年第21期51-54,58,共5页
Relay
基金
海军工程大学自然科学基金项目(HGDJJ07029)