基于覆盖算法的雷达模拟电路故障诊断研究

Study on fault diagnosis in analog circuits of radar based on neighborhood covering algorithm

雷达是一个应用广泛的复杂系统,一旦雷达系统出现故障,可能出现难以预料的后果,造成巨大的损失。因此对雷达故障的定位和快速排除就显得非常重要。针对雷达中模拟电路的单元件软故障,应用领域覆盖算法(NCA)及其改进后的领域覆盖算法(INCA)构建神经网络进行诊断,并将其与BP神经网络诊断方法进行比较,可以发现领域覆盖算法及其改进后的领域覆盖算法在确定网络结构方面和运算速度等方面优于BP算法,且通过改进后的领域覆盖算法构建神经网络的诊断正确率明显高于BP神经网络。

Radar is a complex system with wide application. Once the radar system malfunctions, it may tend to unexpected consequences and cause huge economic losses. Therefore, it is very important to identify the fault and solve it quickly. This paper applies the neighborhood covering algorithm（NCA） and the improved neighborhood covering algorithm （INCA）to building neural networks and uses the neural networks to diagnose the soft fault of single element in analog circuits of radar. And through comparing it with back-propagation（BP） neural networks diagnostic method, it can be found that the NCA and the INCA are better than the BP in determining the structure of networks and in operation speed, and the diagnostic accuracy of the neural networks based on the INCA is superior to the BP neural networks.