Abstract: Fault detection of deployed sensor network is difficult due to constrained resources. A fault detection method based on anomaly task operational log is proposed. Behavior-state model of sensor nodes is devised through analyzing execution characteristics of the node program. The model combining with pre-deployment test records and chi-square test is used for fault detection during post-deployment. Experimental results show that the method can effectively detect the unknown type of fault. The communication storage and computing cost are small.

Key words: fault detection, chi-square test, anomaly detection, concurrency computation model

摘要： 针对传感器网络在资源受限部署后难以实施故障检测的问题，提出基于异常任务运行记录的故障检测方法。通过分析节点程序运行特征，建立节点行为状态模型，结合系统部署前的测试执行记录和卡方检验技术进行故障检测。实验结果表明，该方法与已有方法相比，能有效检测未知类型故障，且通信、存储和计算开销均较小。

关键词: 故障检测, 卡方检验, 异常检测, 并发计算模型

CLC Number: 

• TP309.2