基于概率神经网络针对一类动态系统传感器故障的分类识别法

Sensor Failure Detection in Dynamic System Based on PNN(Probabilistic Neural Network)Approach

基于概率神经网络,针对一类状态可观测的经典动态系统中传感器失效故障,给出了一种快速分类识别方法。利用概率神经网络训练速度快、分类更准确的优点,在多路传感器分别发生故障时,精确检测出发生故障的传感器,并通过实验给出最优spread值。

Based on PNN(Probabilistic Neural Network),this paper presents a fast detection method for sensor failure in dynamic fault system.Probabilistic Neural Network is famous for its quick training and more accurate classification.When multi-channel sensor malfunctions,the network can detect which channel is failure with high accuracy.The optimal value of spread is achieved through experiments.