The high strength-to-density ratio and specific stiffness make carbon fiber reinforced plastics widely used in various industrial applications. However, damage is largely inevitable during the processing, preparation, or use of composite structures [1]. Generally, macroscopic damage modes can be reduced to a few basic types, namely matrix cracking, delamination, fiber/matrix debonding, and fiber rupture [2]. The rather complex failure mechanism of composites makes it difficult to identify specific damage modes. Therefore, an effective damage monitoring method is needed to ensure the safety and reliability of composites.
One of the most common technologies is monitoring the condition of a composite structure using acoustic emission [3]. The damage signal is transmitted as an elastic wave with a rapid energy release, and the signals collected by the sensors are further analyzed by acoustic emission recording devices to extract characteristic parameters. Traditional acoustic emission time signatures are used to study the failure mechanism and explain the damage process. Due to the complexity and stochastic nature of damage mechanisms in composites, the use of multivariable analysis to process acoustic emission signals is necessary to improve the accuracy of damage model identification.
Typical acoustic emission parameters include amplitude, rise time, duration, pulse count, and energy. In addition, there are some additional acoustic emission characteristics calculated from the basic characteristics. Amplitude and energy are the two most representative characteristics concerning the characterization of damage patterns [4]. In particular, the results of numerous experiments indicate that the amplitude corresponding to matrix cracking is higher than the amplitude of delamination, while the amplitude of fiber rupture is the lowest.
In this study, discrete average clustering methods were used to analyze the acoustic emission data obtained from the three-point bending test of carbon-epoxy composite to obtain clustering centers and characteristic signals. In addition, waveform analysis was performed, including wavelet selection, wavelet packet decomposition, and energy feature extraction, to clarify the characteristics of different damage modes and identify the actual damage patterns of the disputed signals. Further improvement of the methodology involved analyzing the impact of each orthonormal regularization step on damage pattern determination to evaluate the best clustering method.
Identification of damage in the local volume of the composite sample was carried out using general unsupervised clustering methods. An addition to the unsupervised clustering method was the identification of the type of damage in the carbon-epoxy composite, including averaging of the characteristic coefficients. The characteristic coefficients of unsupervised clustering allow the division of the input vector into k classes and approximation to the minimum sum of squared distances of all input vectors to their cluster centers. It should be noted that each element of the set of vectors is in one-to-one correspondence with an orthogonal cluster. The total computation time for the clustering algorithm depends largely on the quality of the random initial cluster centers. The working principle of the fuzzy c-mean algorithm is to assign membership to each input vector corresponding to each cluster center based on the Euclidean distance between the cluster center and the input vector, and the values of the membership function are between 0 and 1.
Wavelet packet analysis requires the choice of a basis function for the transform to accurately extract the energy characteristics of each acoustic emission frequency band corresponding to the localization of mechanical damage in the local volume of the composite sample. The methodology used in this study included working with discrete wavelets for acoustic emission signals and densely branched wavelet bases - these are the Db- wavelet basis, the Symlets wavelet basis and the Coiflets wavelet basis.
Numerical calculations using the considered methodology show that the epoxy composite matrix cracking has the lowest range of peak frequencies, while the fiber rupture has the highest range of frequencies. Considering the existence of the difference between different damage types, delamination should be associated with a lower peak frequency range, due to the fact that delamination is a progression of matrix damage, so its frequency band is close to matrix damage. Therefore, the peak frequency can be considered as a key parameter to characterize the damage pattern.
The energy content of each wavelet envelope should be regarded as a key parameter for analyzing the differences in frequency bands related to the damage pattern. Analysis of continuous wavelet transform algorithms indicated the existence of a set of clustering centers.
References:
1. Pramanik A. et al. Joining a carbon fibre reinforced polymer (CFRP) composites and aluminium alloys-A review. Composites Part A: Applied Science and Manufacturing. 2017. Vol. 101. P. 1-29. https://doi.org/ 10.1016/j.compositesa.2017.06.007
2. Pupurs A. Fiber failure and debonding in composite materials. Modeling damage, fatigue and failure of composite materials. Woodhead Publishing. 2016. Pp. 173-196. https://doi.org/10.1016/B978-1-78242-286-0.00009
3. Eaton M. et al. Characterization of damage in composite structures using acoustic emission. Journal of Physics: conference series. –IOP Publishing. 2011. Vol. 305(1). P. 012086.
https://doi.org/10.1088/1742-6596/305/1/012086
4. Yun H. et al. Nonlinear ultrasonic testing and data analytics for damage characterization: A review. Measurement. 2021. Vol. 186. P. 110155. https://doi.org/10.1016/j.measurement.2021.110155
|
Голосування 
Ця робота ліцензується відповідно до Creative Commons Attribution 4.0 International License
Знайшли помилку? Виділіть помилковий текст мишкою і натисніть Ctrl + Enter