Acoustic Emission and Durability of Composite Materials

By Nathalie Godin, Pascal Reynaud and Gilbert Fantozzi

In this book, two kinds of analysis based on acoustic emission recorded during mechanical tests are investigated.

In the first, individual, analysis, acoustic signature of each damage mechanism is characterized. So with a clustering method, AE signals that have similar shapes or similar features can be group together into a cluster. Afterwards, each cluster can be linked with a main damage. The second analysis is based on a global AE analysis, on the investigation of liberated energy, with a view to identify a critical point. So beyond this characteristic point, the criticality can be modeled with a power-law in order to evaluate time to failure.

• Language: English
• Publisher: Wiley
• Release date: Feb 14, 2018
• ISBN: 9781119510512

Book preview

Table of Contents

Cover

Title

Copyright

Introduction

1 Acoustic Emission: Definition and Overview

1.1. Overview

1.2. Acoustic waves

1.3. The sensors and acquisition system

1.4. Location of sources

1.5. The extracted descriptors from the AE signal

1.6. The different analyses of AE data

1.7. Added value of quantitative acoustic emission

2 Identification of the Acoustic Signature of Damage Mechanisms

2.1. Selection of signals for analysis

2.2. Acoustic signature of fiber rupture: model materials

2.3. Discrimination using temporal descriptors of damage mechanisms in composites: single-descriptor analysis

2.4. Identification of the acoustic signature of composite damage mechanisms from a frequency descriptor

2.5. Identification of the acoustic signature of composite damage mechanisms using a time/frequency analysis

2.6. Modal acoustic emission

2.7. Unsupervised multivariable statistical analysis

2.8. Supervised multivariable statistical analysis

2.9. The limits of multivariable statistical analysis based on pattern recognition techniques

2.10. Contribution of modeling: towards quantitative acoustic emission

3 Lifetime Estimation

3.1. Prognostic models: physical or data-oriented models

3.2. Generalities on power laws: link with seismology

3.3. Acoustic energy

3.4. Identification of critical times or characteristic times in long-term tests: towards lifetime prediction

3.5. Simulation of the release of energy using a power law: prediction of the rupture time

Conclusion

Bibliography

Index

End User License Agreement

List of Tables

1 Acoustic Emission: Definition and Overview

Table 1.1. The main techniques used for the identification of damage in composites [JAC 00]

Table 1.2. Average wave velocity according to the type of sample for a polyester matrix and glass fiber composite and for different fiber orientations (micro80 sensors, acquisition threshold 30 dB)

Table 1.3. Set of AE signal descriptors composed of nine temporal descriptors determined in real time by the acquisition system and nine descriptors recalculated from the first ones [MOE 07, MOM 08, SIB 11, MAI 12a, RAC 15]

Table 1.4. Descriptors extracted from digitized waveforms [MAI 12a, MOR 16] (A: approximation, low-pass filter; D: detail, high-pass filter)

2 Identification of the Acoustic Signature of Damage Mechanisms

Table 2.1. Mean characteristics of acoustic signatures of fiber ruptures [MOE 07] for equivalent acquisition configurations (micro80 sensor) from MISTRAS group

Table 2.2. Mean characteristics of signals associated with individual fiber ruptures

Table 2.3. Summary of the acoustic signature associated with the different damage modes for organic matrix composites using a single parameter analysis based on the peak signal amplitude

Table 2.4. Summary of the acoustic signature associated with the different modes of damage of the composites using a monoparameter analysis based on a frequency descriptor

Table 2.5. Results of AE data classification by the k-nearest neighbor technique for different tensile tests

Table 2.6. Segmentation results for the extreme dataset according to the validation criteria and for different values of k

Table 2.7. Segmentation results for the dense dataset for the two algorithms according to the choice of input descriptors

3 Lifetime Estimation

Table 3.1. Value of the ratio tm/tR for the different static fatigue tests on SiCf/[Si-B-C] composites

Table 3.2. Criticality entry time values using RAE and the exponent of Benioff’s law

Table 3.3. Estimation of rupture times obtained by extrapolation of Benioff’s law between tc and a time t’ ranging from 60% to 95% of the test duration for two static fatigue tests

List of Figures

1 Acoustic Emission: Definition and Overview

Figure 1.1. AE diagram of the acoustic emission chain, from the generation of the acoustic wave during a mechanical test to the visualization of the AE signal

Figure 1.2. Acoustic emission sources in composite materials: a) matrix cracking in a SiCf/SiC ceramic matrix composite; b) rupture of a carbon fiber in an epoxy resin with fiber/matrix decohesion [LUT 14]; c) damage of a composite with polyamide matrix PA12, carbon fibers [– / + 45 °] subjected to tensile stress after aging in water for 45 days at 110°C. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.3. An acoustic emission signal produced by a crack growth of the same length in a) a brittle material and b) a ductile material

Figure 1.4. Acoustic emission signal type: a) discrete and b) continuous

Figure 1.5. a) The Kaiser effect and b) the Felicity ratio. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.6. Evolution of the signal peak amplitude as a function of the propagation distance in a 90° unidirectional glass/epoxy composite for three types of sensors (From MISTRAS group). For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.7. a) Instrumentation of a test specimen using two pairs of sensors placed directly on the sample’s surface. b) Instrumentation by means of a waveguide for testing in high temperature or hostile media. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.8. Calibration curve of a micro80 sensor obtained by the reciprocity method for longitudinal waves and surface waves. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.9. a) Wavefront perpendicular to the sensor’s axis, b) wavefront parallel to the sensor’s axis and c) diameter of the sensor and number of wavelengths

Figure 1.10. a) Amplitude recorded by a micro80 sensor for signals of different frequencies and same energy generated by an acousto-ultrasonic card. b) Frequency recorded by a micro80 sensor for signals of different frequencies and energy generated by an acousto-ultrasonic card (propagation distance of 100 mm, composite material propagation medium with SiCf/SiC) [MAI 12a]

Figure 1.11. Localization principle with two sensors

Figure 1.12. a) Spatial and temporal distribution of acoustic activity recorded in a unidirectional composite tensile test with polyamide matrix PA6.6 and glass fibers. b) 2D location for a sandwich structure subjected to bending (carbon epoxy/aluminum honeycomb). c) Spatial and temporal distribution of acoustic activity recorded during a tensile test on a single lap joint (composite/steel). For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.13. Artificial AE source: graphite lead rupture

Figure 1.14. Evolution of the wave propagation velocity as a function of the frequency of the source and comparison with the value obtained using the pencil lead break procedure [MAI 12a] for a SiCf/SiC ceramic matrix composite

Figure 1.15. a) Cyclic traction curve obtained for a SiCf /SiC ceramic matrix composite [MOE 07]. b) Evolution of the coefficient 𝛾 for a SiCf/SiC ceramic matrix composite as a function of strain [MOE 07]

Figure 1.16. Main descriptors measured on an acoustic emission signal

Figure 1.17. Descriptors computed in the time domain on a digitized acoustic emission signal. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.18. Descriptors calculated in the frequency domain on a digitized acoustic emission signal. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.19. a) Signal associated with an individual fracture of a glass fiber. b) Fourier transform of the signal. c) Continuous wavelet transform, Morlet scalogram. d) Spectrogram, smooth Wigner–Ville distribution. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.20. Example of decomposition of a wavelet packet signal up to three levels (A: approximation, low-pass filter; D: detail, high-pass filter) [MAI 12a]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.21. a) Cumulative AE activity in terms of the number of localized signals as a function of the strain and stress/strain curve for a SiCf/[Si-B-C] ceramic matrix composite. b) AE activity and mechanical curve. c) Spatial distribution of AE sources in the position/strain plane. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.22. Cumulative AE activity in terms of cumulative energy associated with the localized signals as a function of the strain and stress/strain curve for a composite SMC (glass fiber and vinylester matrix) subjected to stress. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.23. Detection of the beginning of the significant damage a) from the first signal or the second cascade of activity and b) using a linear regression. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.24. Histogram of the amplitude of the signals localized on the gauge length and recorded during a fatigue test on an aluminum matrix/alumina fiber composite: three signal ranges are shown [JAC 00]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.25. Analysis of the amplitude distribution. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.26. Representation of AE data in the amplitude/average frequency plane for the data recorded in oligocyclic fatigue tests for a 304L-type steel [SHA 06]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.27. Classification of the crack type in the average frequency/rise angle plane for concrete

Figure 1.28. Quantification of the damage in concrete a) in the calm ratio/load ratio plane and b) in the severity/historic index plane

Figure 1.29. Main steps of an unsupervised clustering. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.30. Principles of a supervised classification. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.31. a) Representation of the plot boxes for the descriptor. b) Example of a dendrogram: the numbers from 1 to 18 correspond to the descriptors of Table 1.3; the descriptors 19 and 20 correspond to the frequency barycenter and the peak frequency. c) Passage in the principal component. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.32. Schematic representation of the stages of the learning phase for a Kohonen map. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.33. Principle of classification with a decision tree forest (INBag: in bag data, OOB: out of bag data) [MOR 16]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.34. Principle of the operation of an algorithm based on the k-means method: a) initialization of the cluster centers and calculation of the distance between each signal and cluster centers, b) at the end of the first partition and c) new cluster centers at the end of the first iteration. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.35. a) Evolution of the Davies and Bouldin (DB) coefficient as a function of the number of clusters; the optimal solution is a solution with four clusters. b) Histogram of individual silhouettes for one cluster: the signals circled in red are the badly classified signals; the average silhouette of the cluster is 0.65. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 1.36. Evolution of the amplitude of longitudinal pressure waves and transverse shear waves for a) a mode I and b) a mode II crack. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

2 Identification of the Acoustic Signature of Damage Mechanisms

Figure 2.1. Schematic representation of the acoustic activities expected for the signal clusters corresponding to matrix cracking and fiber breaks during a static fatigue test on a ceramic matrix composite

Figure 2.2. Evolution of the number of a) recorded signals and b) localized signals for a tensile test on a composite SMC (glass fiber and vinylester matrix) in the strain–stress plane. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.3. Typical amplitude/duration correlation diagram showing signal-related noise, saturated signals and poorly defined signals

Figure 2.4. Load--strain curves obtained for different fiber bundles. Gauge length 100 mm for Nicalon fibers, 60 mm for Hi-Nicalon fibers and glass fibers [MOE 07]

Figure 2.5. a) Evolution of the location of AE signals (red dots) in a glass fiber bundle as a function of the strain and evolution of the applied load. b) Load and cumulative number of localized AE signals as a function of strain for a bundle containing 2000 glass fibers (60 mm gauge length). c) Typical signal of an individual glass fiber rupture. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.6. Evolution of the acoustic energy of AE signals as a function of the strain of fibers in a bundle during a tensile test [MOE 07]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.7. Stress/strain curve and monitoring of acoustic activity in a fragmentation test for a glass fiber in a polyester matrix: acoustic emission signals are represented by their amplitude [HUG 02a]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.8. Stress/strain curve and monitoring of acoustic activity in terms of the number of localized signals in a fragmentation test (T = 70 °C) for glass fiber in a polyester matrix [HUG 02a]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.9. AE signal associated with fiber rupture in a multifragmentation test for glass fiber in a polyester matrix

Figure 2.10. a) Overview of a minicomposite composed of 500 SiC fibers and SiC matrix and b) observation of matrix cracks and extracted fibers

Figure 2.11. Comparison between the tensile mechanical behavior of minicomposite and Hi-Nicalon fiber bundle [MOE 07]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.12. Comparison between the distributions of different acoustic emission descriptors recorded during tensile tests on Hi-Nicalon fibers (in gray) and minicomposites (in black): a) energy histogram and b) histogram of the rise time/duration ratio

Figure 2.13. a) Topology of the Kohonen map and b) characteristics of the AE signals from each zone of the map shown in the amplitude/log(rise time) plane. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.14. Activities of the different clusters identified for stratified composites/hemp fibers during a cyclic fatigue test, with each cluster being represented by its uncertainty envelope (figure communicated by authors E. Ramasso and V. Placet). For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.15. a) Radar diagram showing the characteristics of the four clusters. b) Activity of the different clusters as a function of the deformation for a test at 500°C on a CMC, SiCf/[Si-B-C] composite. c) Diagram representing the various damage mechanisms of the CMC composite. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.16. Correlation diagrams between the rise time and the amplitude of AE signals obtained during a) static fatigue tests at 1200°C and 150 MPa, b) static fatigue tests at 1200°C and 230 MPa and c) cyclic fatigue tests performed at 700°C and 0/130 MPa at a frequency of 0.25Hz [MOM 08]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.17. a) Tensile test device for X-ray tomograph and instrumentation of specimens. b) Correlation of tomographic images with acoustic activity recorded during in situ tensile tests on polyurethane foam [DES 05]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.18. a) Cumulative number of localized cluster 1 sources as a function of the damage parameter D and b) cumulative number of cluster 2 localized sources as a function of the hysteresis loop area, for tensile tests on SiCf/[Si-B-C] minicomposites at different room temperatures [MAI 12a]. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.19. Activation of the Kohonen map for a pure polyester resin test on a unidirectional fiberglass/polyester resin composite off-axis at 90° and oriented at 45°. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.20. Activation chronologies of matrix cracking and fiber/matrix decohesion for a unidirectional glass/polyester composite test in a healthy state, oriented at a) 90° and b) 45°. For a color version of this figure, see www.iste.co.uk/godin/acoustic.zip

Figure 2.21. Activation chronologies of matrix cracking and fiber/matrix decohesion for a unidirectional composite test oriented at 90° a) and oriented