Fault diagnosis for uncertain networked systems

Fault diagnosis has been at the forefront of technological developments for several decades. Recent advances in many engineering fields have led to the networked interconnection of various systems. The increased complexity of modern systems leads to a larger number of sources of uncertainty which must be taken into consideration and addressed properly in the design of monitoring and fault diagnosis architectures. This chapter reviews a model-based distributed fault diagnosis approach for uncertain nonlinear large-scale networked systems to specifically address: (a) the presence of measurement noise by devising a filtering scheme for dampening the effect of noise; (b) the modeling of uncertainty by developing an adaptive learning scheme; (c) the uncertainty issues emerging when considering networked systems such as the presence of delays and packet dropouts in the communication networks. The proposed architecture considers in an integrated way the various components of complex distributed systems such as the physical environment, the sensor level, the fault diagnosers, and the communication networks. Finally, some actions taken after the detection of a fault, such as the identification of the fault location and its magnitude or the learning of the fault function, are illustrated