A stochastic change detection methodology for reliable monitoring complex nonlinear dynamic systems is proposed. For a semi-active magneto-rheological (MR) damper, the non-parametric, data-driven restoring force method was used to identify the nonlinear dynamic damping device. Both supervised and unsupervised statistical pattern recognition techniques were used to detect the changes in the physical characteristics of the MR damper with different input currents. The classification errors were analyzed to find the optimal strategy for designing change detection classifiers for reliable structural health monitoring (SHM) applications.
展开▼