基于免疫神经网络的无线传感器网络节点的故障诊断

在无线传感器网络（WSN）的应用中，经常要通过对网络节点的测量数据进行处理来判断WSN运行是否可靠。为了利用BP网络和免疫算法处理数据的优点，本文提出了BP免疫算法。根据免疫的思想，先用免疫算法对BP网络训练所需的权值进行全局搜索优化，再用BP算法进行局部搜索，并输出满足给定网络误差的权值。最后输入检验数据，计算BP网络的输出，并根据输出结果判断WSN发生了何种异常。实验结果显示该方法比单独使用BP网络更客观、准确和可靠。

Fault Diagnosis of Nodes in Wireless Sensor Network Based on Immune Neural Network

In the application of WSN, the operation reliability is usually evaluated by processing the measured data at network nodes. To make use of the advantages of BP network and immune algorithm in data processing, the immune BP algorithm is proposed in the paper. According to the immune theory, the immune algorithm is used to globally search and optimize the weight required by BP network training, and then BP algorithm is used to locally search for outputting the weight which meets the given accuracy of BP network. Finally, the testing data are imported in the trained BP network, and thus, anything abnormal of WSN can be diagnosed by calculating the BP network output. The experiment shows that the result obtained with this method is more accurate and reliable than that by using BP network only.