Fault Detection and Diagnosis of Distributed Parameter Systems Based on Sensor Networks and Bayesian Networks

摘要

This paper presents some considerations related to fault detection and diagnosis, using Bayesian networks, in the complex distributed parameter systems with time and space variables, where the intelligent wireless sensor networks are used as a distributed sensor. These miniaturized intelligent sensors may be placed in the area of multivariable distributed parameter systems and even with limited resources of energy, memory, computational power and bandwidth they may add to solve applications on a large space. Multivariable estimation techniques are easier to applied when a multi-sensor network is used. Bayesian networks bring their main characteristics as graphic models with a node topology and treating information by probabilistic inference. The usage of Bayesian networks is chosen considering the distributed parameter system as a system with continuous variable, but digitally surveyed in discrete time, the sensor placed to measure the time variation of system variables been affected by random noises. The paper presents as an application how Bayesian networks could be applied to fault detection and diagnosis in an on-line estimation of a dynamic model for distributed parameter systems, with exemplification in the case of the city road traffic.