Maintenance Study Based on Bayesian Network and Expectation-Maximum Algorithm

With the applications of large scale wireless sensor networks (WSNs), the number of sensor nodes grows from a few nodes of twenty years ago, a thousand nodes of ten years ago, to today's tens of thousands nodes. As the number of nodes increases, it is a difficult task to maintain their normal working. It is crucial to improve the maintainability of a large-scale WSN due its high cost of maintenance. In this paper, a maintenance technique based on Bayesian network and expectation-maximum (EM) algorithm is proposed. The proposed technique has learning and reasoning capability. It has merits of high efficiency and low cost, which is our research focus for the maintenance of large scale wireless sensor networks. In the proposed technique, Bayesian network and data-driven are used to reduce human subjectivity. The expectation-maximum algorithm is adopted to fully extract the information contained in historical data and reduce the cost of the collection of high-quality data.