Time-Domain Fixed-Structure Closed-Loop Model Identification of an Unstable Multivariable Maglev Nanopositioning System

This paper presents improved empirical representations of a general class of open-loop unstable systems using closed-loop system identification. A multi-axis magnetic-levitation (maglev) nanopositioning system with an extended translational travel range is used as a test model to verify the closed-loop system-identification method presented in this paper. A closed-loop identification technique employing a known controller structure is used for model identification and validation. Direct and coupling transfer functions (TFs) are then derived from the experimental input-output time sequences and the knowledge of controller dynamics. A persistently excited signal with a bandwidth in the frequency range of interest is used as a reference input. An order-reduction algorithm is developed to obtain TFs with predefined orders, which gives the closest match in the frequency range of interest without missing any significant plant dynamics. The entire analysis is performed in the discrete-time domain in order to avoid any errors due to continuous-to-discrete-time conversion and vice versa. Continuous-time TFs are used only for order-reduction and performance analysis of the identified TFs. Experimental results are presented in the time as well as frequency domains to verify the accuracy of the identified plant TFs. These results also demonstrate the effectiveness of the developed closed-loop identification method in meeting all of the three core objectives—(i) reduction in cross-axial coupling from 9.213 μm to 0.911 μm in translation and from 22.03 μrad to 1.353 μrad in rotation, (ii) large range motion capability with a travel range of ±2.9 mm, and (iii) improved robust stability.