دانلود Intelligent Fault Management System for Wireless Sensor Networks with Reduction of Power Consumption

Abstract

In Wireless Sensor Networks (WSN), when a hard-ware error is not identified and promptly corrected, all monitoring  of  a  sensor  network  is  compromised.  This  work  presents a  new  approach  to  centralized  fault  management  system  for6LoWPAN  WSN.  The  system  is  based  on  two  fault  detection levels.  A  first  level  is  performed  locally,  by  all  sensors  within the   network,   using   statistical   methods.   The   second   level   is performed by the base station, through an ensemble of Multilayer Perceptron  type  Artificial  Neural  Networks  (ANN)  classifiers. One  of  them  is  continuously  trained  with  streaming  data,  while the other one is used to take actual decisions about fault detection. The  inputs  of  these  ANN  are  outputs  from  a  Kalman  Filter and  from  an  accelerometer.  Experimental  results  indicate  that this  approach  is  capable  of  reducing  message  traffic  and  power consumption within a WSN, keeping the detection accuracy rate higher  than  97%.