Design of an optimal input signal for plant-friendly identification of inertial systems

摘要

System identification, in practice is carried out by perturbing processes or plants in operation. That is why, in many industrial applications a plant-friendly input signal would be preferred for system identification. The objective of this kind of experiment design is to minimise the variance of the parameters to be estimated and to maximise the input friendliness. In this case, the objective function was formulated through maximisation of the Fisher information matrix determinant (D-optimaliiy) in the form of a conventional integral criterion with amplitude constraints. The D-efficiency which quantifies the input trajectories friendliness was verified.