Observers and redundancy equations generation for systems with unknown inputs
Résumé
In order to detect sensor and actuator failures or process disturbances, we use analytical redundancy in order to generate a variable that might contain information about the defects. This variable, also called a residual, is excepted to be significant of the defects as it is computed such that its value is equal to zero when no failure occurs and quite different from it otherwise. Several ways to generate the residuals have been developed but we specially study two techniques. The first one is a direct residual generation based on the parity space approach whereas the second one is based on an output estimation of the implied process which is further compared to its measure to create the residual. This paper describes a procedure for the residual generation using an observer in both cases of known and unknown input systems. Moreover; we show that direct redundancy, i.e. the input-output system transfer function, is not only useful to generate the residual but also provides a signal linked with one generated by the previous observer. Our results are put into theorem and lemma format.