DATA-DRIVEN AUTOREGRESSIVE MODEL IDENTIFICATION FOR STRUCTURAL HEALTH MONITORING IN ANISOTROPIC COMPOSITE PLATES

Abstract

A simple data-driven AutoRegressive (AR) model may be used to assess a model to describe and to predict the time-series outputs of the PZT sensors receiving Lamb waves for different operating conditions in composite structures. Thus, this paper presents the potentiality of the use of a set of AR models to detect, locate, and, manly, to extrapolate a damage sensitive index based on changes in onestep- ahead prediction errors. To illustrate this proposal, an aeronautical composite panel with bonded piezoelectric elements, that act both as sensors and actuators, is used to study the relationship between the variation of the parameters of the identified model and the presence of various simulated damage. A damage progression evaluation by extrapolating the AR parameters is also suggested and examined based on cubic spline functions to verify the future state and to observe how the damage could evolute, based on some simplified assumptions. This step could help to make a decision about a possible required repair without adopting a complicated and costly physical model.