Identification Sensor Nodes Normal Operation Using Allan Variation
Volodymyr Kryzhanovskyi, V.F. Komarov, S.P. Serhiienko, L.V. Zahoruiko · Visnyk of Vinnytsia Politechnical Institute · 2021
In cyberphysical systems and the Internet of Things (IoT) the threat of failure of sensor nodes as well the threat of unauthorized interference during the transmission of measurement data exists. Thus, the need to develop and improve methods of monitoring sensor nodes is relevant given the large number of such nodes, energy constraints that limit traffic and computing resources of systems, as well as the need for full automation of work due to the economic inexpediency of their maintenance. It is proposed to use one of the statistical methods, namely the Allan variance or a more general concept of structural functions, to study stochastic processes in time sequences representing the implementation of random measurement results in such systems. The ability to calculate the statistical characteristics of nonstationary processes and to determine the spectral dependences of various kinds of noise at moderate computational costs is the advantage of that method. The use of Allan dispersion to process model temperature and humidity data from Arduino-based sensors in the Internet of Things system is considered. The dependences of Allan deviation and its modifications as functions of various averaging interval are calculated. For the typical modes of operation of the considered sensors in the Internet of Things applications the dependences of the Allan deviation correspond to the type of noise "Random Walk FM". It is confirmed by calculations at different measurement intervals. The influence of different types of noise on these dependences is investigated. The minimum set of parameters that can with a limited data set characterize the normal measurement process, possible malfunctioning or natural and artificial interference to normal operation was chosen. The possibility of splitting the process of statistical processing between the sensor node and the controller of the Internet of Things is indicated, which will reduce the traffic and the appropriate energy consumption of the sensor node.