Diagnosing faults and illnesses with a quasi-time-invariant filter transfer function model

The problem of diagnosing a fault or a perturbation in a dynamical system (be it industrial or natural) can be restated as the characterization of the transfer function of a filter that acts on a “sane” or “unperturbed” system to transform it into a “faulted” or “perturbed” one. This filter would then be approximated as time-invariant (and linear) only for short time windows, while the time evolution of such transfer function on longer time scales leads to a simple but powerful general framework that allows to diagnose the onset and/or the time evolution of a fault or perturbation of the system.