Structural system identification - A study of different algorithms

Abstract

This paper uses Observability Range Space Extraction (ORSE) identification algorithm as a main thread to study various identification algorithms for structural systems. These algorithms include: Eigensystem Realization Algorithm (ERA), Q-Markov Covariance Equivalent Realization (Q-Markov Cover or QMC) algorithm, Least Squares algorithm for AutoRegressive eXogeneous model identification (LSARX), and ordinary Multi-input and multi-output Output Error State space (MOESP) model identificaiton algorithm. ERA, QMC, and ORSE algorithms have been successfully applied to structure system identifications. They have demonstrated robust noise suppression property and well behaved numerical properties. Comparison of these three algorithms shows that the ORSE algorithm is an extension of the ERA and QMC algorithms for general input and output data. LSARX algorithm was integrated with ERA algorithm for structural system identification. This paper investigates relationships between LSARX and ORSE algorithms and shows some potential disadvantages of using LSARX algorithm for high dimensional structural systems. Although ordinary MOESP algorithm has not been applied to structural systems, the algorithm and ORSE algorithm are very similar and differ only in computational implementation. Their similarities and differences are discussed. Finally, bias problems of the ORSE algorithm is studied and an unbiased version of the algorithm is derived.