Data-driven model identification of guided wave propagation in composite structures

Abstract

This paper shows the applicability of data-driven model identification to describe guided wave propagation in composite structures. The model identified can be useful to predict waveform or further conditions considering the dynamics of wave propagation in composites media, mainly when the geometry or boundary conditions are complex and make it very difficult to propose an analytical model. Thus, the main purpose of this paper is to use a simple autoregressive with exogenous terms model to fit the experimental data assuming different frequencies of excitation and the effect of temperature changes. An example using a composite plate is performed in this paper to show the steps and applicability. After these analyses, the results obtained demonstrate the practical benefits of the black-box model for modeling of guided wave propagation in complex composite structures for aeronautical applications.

Delamination area quantification in composite structures using Gaussian process regression and auto-regressive models

“…Figueiredo et al (2012) proved the capability of autoregressive models with eXogenous inputs (ARX) as a sensitive feature for damage detection in a composite plate. Silva (2018) illustrated some practical aspects of the implementation of ARX models using Lamb wave for system identification of a composite plate assuming temperature effects. Moreover, the authors discuss the potential of ARX models regarding the purpose of SHM.…”

Extrapolation of AR models using cubic splines for damage progression evaluation in composite structures

“…Among these approaches, one very effective way to address this issue is by using data-driven model identification based on guided Lamb wave propagation or random inputs provided by PZT active-sensing. In particular, Auto-Regressive (AR) models to describe Lamb waves are already abundantly used in the SHM of composite structures (da Silva, 2018; Figueiredo et al, 2012; Nardi et al, 2016; Paixão et al, 2020).…”

“…Figueiredo et al (2012) showed the use of time-series predictive models for piezoelectric active-sensing using autoregressive with exogenous input (ARX) models and machine learning to detect damages. da Silva (2018) also applied the ARX model to perform predictions, and a waveform generator in a 10 layers carbon-epoxy plate excited by guided Lamb waves assuming different central frequencies and environmental conditions, such as temperatures changes, and noted some benefits and disadvantages of the possible performance of this strategy for SHM.…”

“…Identified data-driven model, as suggested by Nardi et al (2016) or da Silva (2018) using AR family models, could be attractive to be adopted to interpolate the states between the healthy and damaged condition and to extrapolate or to quantify a damage progression. These models can be used as a surrogate model to reach a subsequent application of higher forms of SHM’s hierarchy.…”

Fuzzy clustering and AR models for damage detection in CFRP coupons considering loading effect

“…Additionally, composite materials have been used in different industrial applications due to their features such as lightweight, high strength-to-weight ratio, and long-term durability. However, the level of requirements in SHM of these materials is more complicated compared with metallic structures due to complex damage, such as matrix cracks, delamination, and debonding, and significant effects of load variations in service life [14].…”