Subspace identification with prior steady-state information

Abstract

In system identification, the quality of data is important for obtaining good models, but there are situations where the available data are highly corrupted with noise. However, some prior information about the system to be identified, such as dc gain and settling time, may be available to obtain improved model identification despite data noise. In this paper, a subspace identification scheme incorporating known dc gain is investigated. The prior process information is incorporated into system identification through using the least square with the equality constraint that the sum of impulse response parameters is equal to the dc gain. In comparison with the existing approaches, the proposed identification strategy provides unbiased parameter estimations, is applicable to Multi-Input Multi-Output (MIMO) systems, and is computationally efficient.

Model predictive control of atomic force microscope for fast image scanning

“…From this FRF using the prediction error method (PEM) best fit is chosen with the measured plot as described in references [18], [19], [20]. The system behavior is characterized by processing the data from the identified model.…”

High performance control of atomic force microscope for high-speed image scanning

“…From this FRF using the prediction error method (PEM) best fit with the measured plot is chosen as described in references [17], [18]. The system behavior is characterized by processing the data from the identified model.…”

Regression techniques for subspace-based black-box state-space system identification: an overview

“…Initially developed for systems working under open-loop conditions [62,76,74], this linear regression approach has allowed the development as well as the analysis of some subspace-based identification algorithms for closedloop data [63,45,75,46,59,23,24,44,17]. More recently, this formulation of the subspace-based identification as a linear regression-based problem has been used to incorporate prior information into some subspace-based identification algorithms [80,81,1].…”