A robust fault detection approach on the basis of virtual sensors collocated with fault excitations

Robustness against disturbances and model uncertainties are important problems in fault detection processes. On the basis of different objectives, parameter optimization of the residual generator can be done to enhance robustness. If disturbances and model uncertainties are modeled as unknown inputs, the purposes of the optimization process can be formulated as to attenuate the influence of unknown input and enhance the influence of fault on the residual. Since both objectives have to be considered in the optimization process and their optima are usually different, the solution of the optimization problem is a compromise between these two objectives. Despite different methods and tools to solve the multi-objective optimization problem, an approach using a single objective function on the basis of virtual sensors, which collocate with the fault excitations is introduced in this paper. Both optimization purposes are considered in one objective function. The objective function is easy to apply and the optimization problem can be solved in a single optimization routine.