A Strategy for Fault Recovery of Wireless Sensor Network Based on v-SVC
Jian Kang Yang · Journal of Information and Computational Science · 2013
Wireless Sensor Network (WSN) is applied widely and large-scale, which is usually used to collect physical parameters in wicked environment. It’s difficult for human to intervene when a WSN has faults. That is to say, fault diagnosis, isolation and recovery which exist automatically in WSN itself can solve this problem. In clustering algorithm, a Cluster Head (CH) is necessary for the whole network. When the CH can’t work, due to the energy, spatial barrier, strength of signal and so on, we have to replace the CH efficiently; when there is no suitable backup CH for one cluster, the cluster should be re-clustering which needs an effective strategy for the nodes joining other clusters. This paper proposed a strategy for the WSN in that situation. Based on the Support Vector Machine (SVM), we suggested a decision-function. In this function, by using support vector we can balance multiple causes of a fault except energy. This light calculation can make fault management of WSN more efficient, practical and easy-deployed to improve the lifetime and robustness of the whole network. The simulation experiment showed that, based on the v-SVC, the CH replacing and re-clustering algorithm are more efficient than the traditional solutions in way of optimizing the training sets and kernel function.