Rj Robbert Voorhoeve

On numerically reliable frequency-domain system identification: new connections and a comparison of methods

Frequency domain identification of complex systems imposes important challenges with respect to numerically reliable algorithms. This is evidenced by the use of different rational and data-dependent basis functions in the literature. The aim of this paper is to compare these different methods and to establish new connections. This leads to two new identification algorithms. The conditioning and convergence properties of the considered methods are investigated on simulated and experimental data. The results reveal interesting convergence differences between (nonlinear) least squares and instrumental variable methods. In addition, the results shed light on the conditioning associated with so-called frequency localising basis functions, vector fitting algorithms, and (bi)-orthonormal basis functions.

Accurate FRF Identification of LPV Systems: nD-LPM With Application to a Medical X-Ray System

Linear parameter varying (LPV) controller synthesis is a systematic approach for designing gain-scheduled controllers. The advances in LPV controller synthesis have spurred the development of system identification techniques that deliver the required models. The aim of this paper is to present an accurate and fast frequency response function (FRF) identification methodology for LPV systems. A local parametric modeling approach is developed that exploits smoothness over frequencies and scheduling parameters. By exploiting the smoothness over frequency as well as over the scheduling parameters, increased time efficiency in experimentation time and accuracy of the FRF is obtained. Traditional approaches, i.e., the local polynomial method/local rational method, are recovered as a special case of the proposed approach. The application potential is illustrated by a simulation example as well as real-life experiments on a medical X-ray system.

Frequency response function identification of LPV systems: A 2D-LRM approach with application to a medical X-ray system

LPV control has emerged as a systematic approach in the design of gain-scheduled controllers. This requires the identification of LPV models. The aim of this paper is to develope a flexible and accurate 2D-LRM approach, to enable fast and accurate non-parametric system identification of a frequency response function. The scope of this paper is on the identification of open-loop SISO LPV systems. Smoothness between several frozen LTI conditions within the LPV system is exploited to enable accurate pre-testing for parametric LPV modeling. The proposed approach achieves smoother estimations of the LPV behavior without an increase in estimation errors and with reduced variances. Traditional LPM and LRM approaches can be recovered as a special case of the proposed approach. The potential of the approach is shown by virtue of a simulation of a medical X-ray system.

Evaluating performance of multivariable vibration isolators: A frequency domain identification approach applied to an industrial AVIS

Vibration isolation is essential for industrial high-precision systems in suppressing the influence of external disturbances. The aim of this paper is to develop an identification method to estimate the transmissibility matrix for such systems. The transmissibility matrix is a key performance indicator in vibration isolation, but its identification is severely limited by the heavy weight and size of many industrial systems. Two non-parametric system identification methods based on periodic and spectral analysis are compared. It is shown that spectral analysis can benefit from random floor excitations at low frequencies and periodic shaker excitations at high frequencies. Using this method, a transmissibility matrix between 1 and 100 Hz is successfully measured on an industrial active vibration isolation system (AVIS), demonstrating that the proposed method is suitable for identification of these heavy-weight systems.

Identification for motion control: Incorporating constraints and numerical considerations

Frequency domain identification is a common starting point for model based motion control. The aim of this paper is to tailor parametric identification methods to the specific class of motion systems. The proposed method involves two aspects: 1) incorporating prior knowledge, e.g. it is often known beforehand that the system exhibits rigid-body behavior, and 2) numerical reliability, since the considered class of systems is challenging to identify. The result is a frequency domain algorithm that is particularly suited for the identification of lightly damped motion systems with rigid body behavior, a high model order, and large input- output dimensions. Experimental results on a prototype next-generation motion system clearly demonstrate the advantages of the proposed approach.