Active diagnosis of MLD systems using distinguishable steady outputs

In active diagnosis the system is excited by a signal that aims to uncover latent errors. However, the diagnosis signal may destabilize the system, in particular in an open-loop structure, but also in a closed-loop structure, because the nominal controller is designed to stabilize the nominal system. This paper presents a method for active diagnosis of Mixed Logical Dynamical (MLD) systems where instability is avoided: The diagnoser looks for steady states of both the normal and faulty system which are reachable by the same input such that the corresponding outputs are distinguishable from each other. The input is applied to the system and the condition of the system is determined based on the output. Thus this excitation preserves stability. The method can be useful in a design phase to find a sensor allocation which guarantees diagnosability. The method is tested on the two tank benchmark example.