On-line Model-free Sensor Fault Identification and Dictionary Learning in Cyber-Physical Systems

This paper presents a model-free algorithm for the online fault identification and fault dictionary learning in Cyber-Physical Systems (CPSs). The algorithm, which relies on the modelling of functional relationships among the datastreams acquired by CPSs sensing units and on model-free change detection mechanisms to inspect for faults, is able to automatically identify the fault type of the detected fault (among the set of autonomously defined faults present in the fault dictionary) or recognize the occurrence of a new type of fault (that is then automatically inserted in the fault dictionary). The main characteristics of the proposed algorithm are the ability to operate without requiring

any a-priori information about the system under inspection or the possibly occurring faults, i.e., following the model-free approach, and the capability to autonomously build the fault dictionary over time once a new type of fault is recognized, i.e., operating in an on-line manner. A human supervisor could be considered in the loop to associate each autonomously defined fault type (in the parameter space) to its behaviour in the data space (to estimate fault type parameters). Experimental results on both synthetic and real datasets corroborate the effectiveness of the proposed solution.

The proposed system including the LTI modelling, HMM-CDT, fault identification and dictionary updating.