Industrial IoT Technologies and Applications. International Conference, Industrial IoT 2016, GuangZhou, China, March 25-26, 2016, Revised Selected Papers

Research Article

Data Recovery and Alerting Schemes for Faulty Sensors in IWSNs

  • @INPROCEEDINGS{10.1007/978-3-319-44350-8_7,
        author={Huiru Cao and Junying Yuan and Yeqian Li and Wei Yuan},
        title={Data Recovery and Alerting Schemes for Faulty Sensors in IWSNs},
        proceedings={Industrial IoT Technologies and Applications. International Conference, Industrial IoT 2016, GuangZhou, China, March 25-26, 2016, Revised Selected Papers},
        proceedings_a={INDUSTRIALIOT},
        year={2016},
        month={9},
        keywords={Data recovery Hierarchical support vector machines IWSNS Alerting system},
        doi={10.1007/978-3-319-44350-8_7}
    }
    
  • Huiru Cao
    Junying Yuan
    Yeqian Li
    Wei Yuan
    Year: 2016
    Data Recovery and Alerting Schemes for Faulty Sensors in IWSNs
    INDUSTRIALIOT
    Springer
    DOI: 10.1007/978-3-319-44350-8_7
Huiru Cao1,*, Junying Yuan1,*, Yeqian Li1,*, Wei Yuan1,*
  • 1: Nanfang College of Sun Yat-sen University
*Contact email: caohuiru0624@163.com, cihisa@outlook.com, liyq@mail.nfu.edu.cn, yuanw@mail.nfu.edu.cn

Abstract

In monitoring and alerting industrial system, industrial wireless sensor networks play an important role. However, we usually have to face one critical issue that is to recover the data and emergency treatment schedule for the faulty sensors. In this paper, we target on monitoring industrial environments and deal with the problems caused by the failure or faulty sensors nodes. Firstly, based on industrial private cloud, an architecture of industrial environment monitoring system is proposed. Furthermore, a hierarchical support vector machines is adopted for faulty nodes’ data recovery. Unlike most previous works, we intend to address the problem from global and local data perspectives. Using the first layer Support Vector Machines is adopted to judge the types of missing data based on the monitoring system. In second layer of SVM is responsible for finishing the recovery local data in the light of the history records. Performance of the proposed SVM data recovery strategies are evaluated in terms of networks self-healing competence, and energy consumption. We also implement our schemes in a real-life monitoring and alerting network system to demonstrate the feasibility and validate the network detection capability of emergency events.