SYSTEM IDENTIFICATION AND CLASSIFICATION OF STOCHASTIC CHANGE DETECTION OF UNCERTAIN NONLINEAR SYSTEMS WITH REDUCED-ORDER MODELS

摘要

A reliable structural health monitoring (SHM) methodology is proposed to detect relatively small changes in uncertain,time-varying,nonlinear systems. Using a nonlinear magneto-rheological (MR) damper,effective system changes and uncertainties were precisely controlled over the course of 4000 physical tests. The tested MR damper was identified with the Restoring Force Method (RFM),a nonparametric system identification method involving two-dimensional orthogonal polynomials. Classification results shows that the identified coefficients with orthogonal basis functions can be used as reliable indicators for detecting small,genuine system changes with reduced-order models. An optimal design procedure is also proposed and for the classification of detected system changes. The use of the MR damper to create physical change detection data,while suitable in this study,does not have complete freedom to simulate a wide range of nonlinearities. To allow for the physical modeling of a wide range of nonlinear models,the authors also introduce an experimental setup whose restoring force can accurately be controlled and reprogrammed (with software) based upon directly measured displacement and velocity readings at each time step.