基于神经网络的传感器冗余方法研究􀀁

本文把传感器硬件冗余技术、BP网络以及串并联Elman递归神经网络结合起来，提出了一种新方法用于构造具有传感器故障检测、分离且具有冗余能力的智能传感器系统。这可以减少硬件冗余技术中使用传感器的个数、消除硬件冗余技术中对冗余传感器应具有相同参数特性的要求，同时可以提高系统预报的准确性，仿真结果验证了这一方法的有效性。

Research on the Redundant Method for Sensors Based-on Neural Network

Combining the hardware redundant technology with BP neural network and Elman recurrent neural network, this paper presents a new method for constructing an intelligent sensor system with abilities of sensor fault detection, fault isolation and redundancy. By making use of the method, the numbers of sensors used in the hardware redundant technology can be reduced, the demands for the sensors with the same parameter properties can be ignored and at the same time the prediction accuracy for system output also could be improved. The result of simulation for a real temperature control system show that the method is correct and practicable.