ABSTRACT
ARMA model is constructed for each data sample and the model residual variance is calculated. The health and damage condition monitoring data are served as the inspected samples and substituted into the constructed ARMA model to calculate the new residual variance. It reveals that there are significant differences between two state model residual variances. K-L distance is used as indicators of damage severity. The hypothesis test is proposed to identify the state of the structure and supply the early warning the damage. Numerical results show that the damage indicator based on ARMA model residual variance can accurately distinguish between health and damaged states of the structure.
