Adaptive Modeling and Eigen-Parameter Identification of Structures based on the MX Filter and CMX-LMS Algorithm

Abstract

The paper considers the adaptive modelling and eigen-parameter identification of structures that may have time-varying dynamic properties. After the introduction of the MX filter and the MX-LMS algorithm, the relationships between the MX filter and structural physical properties are derived. A new adaptive algorithm, the CMX-LMS algorithm, is then presented. The CMX-LMS algorithm combines the MX-LMS algorithm with the deduced equivalent, optimising the eigen-frequency, damping ratio and the forward path gain directly. Compared to the MX-LMS algorithm, the CMX-LMS algorithm has the following advantages: ensures that the MX filter is stable during optimisation, a reduced number of optimised parameters, easy setting of the initial parameters and their search ranges and fast convergence. The simulations for a single-mode structure, a three-mode structure and a real-time experiment exhibit the good performance of the CMX-LMS algorithm.