C–SiC composites with different fiber architectures

Author's Accepted Manuscript

Modal acoustic emission of damage accumulation in C/C-SiC composites with different fiber architectures Fabian Breede, Dietmar Koch, Emmanuel Maillet, Gregory N. Morscher

www.elsevier.com/locate/ceramint

PII: DOI: Reference:

S0272-8842(15)01142-6 http://dx.doi.org/10.1016/j.ceramint.2015.06.026 CERI10765

To appear in:

Ceramics International

Received date: Revised date: Accepted date:

17 April 2015 4 June 2015 4 June 2015

Cite this article as: Fabian Breede, Dietmar Koch, Emmanuel Maillet, Gregory N. Morscher, Modal acoustic emission of damage accumulation in C/C-SiC composites with different fiber architectures, Ceramics International, http://dx.doi.org/10.1016/j.ceramint.2015.06.026 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Modal acoustic emission of damage accumulation in C/C-SiC composites with different fiber architectures Fabian Breedea,∗, Dietmar Kocha , Emmanuel Mailletb , Gregory N. Morscherb a

Institute of Structures and Design, German Aerospace Center (DLR), Pfaffenwaldring 38–40, 70569 Stuttgart, Germany b Mechanical Engineering Department, The University of Akron, Akron, OH 44325-3903, USA

Abstract The stress-dependent material behavior of carbon fiber reinforced ceramic matrix composites (C/C-SiC) was investigated with modal acoustic emission (AE) technique. AE observation is a promising health monitoring method which provides accurate location and identification of the recorded damage-related AE events. For a better understanding and identification of different damage mechanisms C/C-SiC composites with different fiber orientations were studied. Load/unload/reload tensile tests were performed and measurements were made over the entire stress range in order to determine the stressdependence of acoustic activity for increasing damage states. In general it was found that in C/C-SiC composites a significant damage-related increase in AE energy was observed close to the ultimate tensile stress. Also, fiber architecture dependent differences during damage accumulation in terms of total number of AE events and their typical AE energy could be determined and correlated to different damage mechanisms. Keywords: B. Ceramic matrix composites (CMCs), B. Failure analysis, C. Mechanical properties, C. Damage accumulation 1. Introduction Carbon fiber reinforced ceramic matrix composite (CMC) materials are favorable for thermostructural components due to their excellent specific mechanical properties, high thermal conductivity and thermoshock resistance. Carbon fiber reinforced silicon carbide (C/SiC) composites have made it into several high-temperature aerospace applications, like nozzle extensions, combustion chamber components and thermal protection panels for re-entry [1, 2]. The Institute of Structures and Design of the German Aerospace Center (DLR) has been continuously Corresponding author, Tel.: +49 711 6862 618; Fax: +49 711 6862 227 Email address: [email protected] (Fabian Breede) ∗

Preprint submitted to Ceramics International

working on advanced C/C-SiC composites for future thermal protection and space propulsion systems [3, 4]. The damage behavior of these materials must be known to allow a proper design of structural components. The somewhat nonlinear stress-strain characteristic in CMCs when loaded can result from different energy dissipating effects, like laminar- and inter-laminar matrix cracks, crack opening and closure, crack propagation and deflection, interfacial sliding (friction) and single fiber breaks or fiber bundle breaks. All of these effects create acoustic waves which can be digitized and analyzed using the modal acoustic emission (AE) technique. In the recent past the modal acoustic emission approach has been used to investigate the stress-dependent damage accumulation in several ceramic matrix composite June 4, 2015

materials and has proven to correlate individual AE events to specific damage sources. [5–13]. In this study, the stress-dependent damage accumulation of DLR’s C/C-SiC composites was evaluated for the first time applying the acoustic emission technique. Load/unload/reload hysteresis tensile tests were performed on C/C-SiC materials with a fiber orientation of ±15˚, ±45˚ and 0/90˚, respectively. Modal AE was used to record damage-related acoustic events and then to determine their location. Also an attempt was made to correlate waveform properties as well as the energy character of AE events to their probable physical damage sources. Fig. 1. Typical microstructure of an ”as-received” C/C-SiC composite with a 0/90˚ fabric fiber architecture manufactured via liquid silicon infiltration (LSI).

2. Material and testing procedure In this study C/C-SiC composites1 were tested [14, 15]. Two C/C-SiC panels, one with a 0/90˚ and one with a ±15˚ fiber architecture, were manufactured using the liquid silicon infiltration (LSI) process [16]. The 0/90˚ composite was composed of fourteen HTA2 carbon fiber plies of plain weave fabrics (Style 98150). The ±15˚ composite consisted of thirteen layers of plain weave fiber cloth which were fabricated by braiding technique3 using T800HB4 carbon fibers. Each fabric ply was pre-impregnated with a phenolic carbon precursor. In the first processing step the preimpregnated fiber preforms were cured to a CFRP green body panel by warm pressing technique. During the following carbonization process the polymer matrix of the composite was converted to amorphous carbon forming a highly porous carbon/carbon preform. In this state carbon fibers were embedded in characteristic dense carbon matrix blocks. The carbon fibers were thus protected from degradation during the final liquid silicon infiltration. The silicon carbide matrix was then built up by the chemical reaction of the liquid silicon, that is infiltrated by capillary forces, and the

accessible carbon. Here, silicon carbide is formed around each carbon/carbon block. Fig. 1 shows a typical micrograph of an as-received C/C-SiC composite with a 0/90˚ fiber architecture. The fiber volume fraction was about 60 %. The test samples were cut from the 3 mm thick panels into dog-bone and straight-sided tensile test specimens. The length of the dog-bone specimens was 220 mm, the width at the end was 15 mm and the width of the gage section was 10 mm. The gage section length was 60 mm. The straightsided specimens had a length of 150 mm and a width of 10 mm. Load/unload/reload hysteresis tensile tests with 8–9 loops were performed with an electromechanical universal testing machine5 at a rate of 2 kN/min. During the tests glass fiber reinforced epoxy tabs (2 mm thick) were used in the grip section (hydraulic grips). Strain measurements were made using a clip-on extensometer6 with a gage length of 25.4 mm and a range of 2 % strain. Acoustic emission was monitored during the tests with a Fracture Wave Detector7 acquisition system. Three wide-band transducers (Sen-

1

Manufactured at the Institute of Structure and Design (DLR), Stuttgart, Germany. 2 Toho Tenax Co. Ltd., Japan. 3 Fiber preforms fabricated by the Institute of Aircraft Design, The University of Stuttgart, Germany. 4 Toray Carbon Fibers Europe S. A. (CFE), France.

5

Criterion Model C43, MTS Systems Corporation, Eden Prairie, MN, USA. 6 Model 632.27E-30, MTS Systems Corporation 7 Digital Wave Corporation, Centennial, CO, USA.

2

Before loading the tensile specimen boron fiber8 breaks were performed on the face of the fully equipped tensile bar just outside of the top and bottom AE sensors. These manually-generated acoustic events were used to determine the speed of sound of the undamaged sample [5]. The specimens were then mechanically loaded, unloaded to zero stress and reloaded where the maximum stress level was progressively increased until failure. AE activity was continuously recorded during the tensile tests. In order to correlate acoustic emission activity to the composite’s mechanical behavior, only AE events should be considered for the analysis which were generated between the top and the bottom sensors. Only AE events within the tensile bar gage length (60 mm) could be used for location determination and are related to the recorded strain measurements. In this study a typical 3-sensor configuration was used which allowed an easy sub-division into three general location regions according to which sensor was triggered first. Here ”sensor 1 events” are referred to longitudinal locations x≥+15 mm, ”sensor 2 events” x≤-15 mm and ”sensor 3 events” are between -15 mm≤x≤+15 mm, where x=0 is the middle of the gage length. In general an adequate filtering procedure has to be applied to discard false events, e. g. events that were triggered by electro-magnetic interferences. Some of the test samples were observed by scanning electron microscopy (SEM) in order to evaluate if typical damage mechanisms are detectable within the composite comparing ”asreceived” and loaded (damaged) specimens. The SEM specimens were cut from the tensile test bars in longitudinal direction and polished.

F Tensile bar

Acoustic sensor

x=0mm

+15mm AE 3

-30mm

-15mm

Clip-on Extensometer gage length 25.4mm

Tensile bar gage length 60mm

+30mm

AE 1

AE 2 10mm

F

Fig. 2. Schematic illustration of the AE sensor configuration and main tensile test specimen dimensions.

sor Model B1025, 50–2000 kHz, 9.2 mm diameter) were mounted to the face of the tensile specimen (middle, +30 mm, -30 mm) to record acoustic emission events. The tensile test set-up and mounting of the measurement equipment is shown in Fig. 2. The AE sensors were connected to a preamplifier (Model PA-20) which was connected to the acquisition system (Model FM1, 4-channel, 20 MHz, 16 bit). The preamplifier was set at 12 dB for the regular tensile test and 18 dB for the boron fiber breaks. The filter signal and the filter trigger were not amplified. Each AE signal consisted of 1024 points (including 256 points of pretrigger) which were digitized at a sampling rate of 10 MHz. The load and strain were also recorded by the acquisition system. The post-test analysis was conducted with the Wave Explorer software and MS Excel. The test procedure for each test was as follows.

3. Results and discussion 3.1. Speed of sound determination In order to determine the location of the observed AE events the velocity of the extensional wave component Ce must be known. The theoretical speed of sound of an extensional wave Ce,theo 8

Boron fiber source - Specialty Materials, Lowell Massachusetts, USA

3

Table 1 Some C/C-SiC sample properties. Fiber orientation Sample ID Sample shape Fiber type Fiber vol. fraction Density (g/cm3 ) Open porosity Elastic modulus (GPa)

±15˚ HP643 dogbone T800 1.80 103±3

0/90˚ HP717 dogbone HTA ∼ 60 % 1.85 <2 % 74±4

The plot in Fig. 3(a) shows the experimental speed of sound values versus elastic modulus for all three studied C/C-SiC composites. The measured extensional wave velocity is clearly dependent on its initial elastic modulus and fiber architecture, respectively. In C/C-SiC composites the mechanical properties are highly dominated by the orientation of the embedded carbon fibers. The determined speed of sound of unloaded C/C-SiC ranges from 3793 ± 97 m/s (fiber orientation: ±45˚), 6108 ± 137 m/s (0/90˚) to 7731 ± 98 m/s (±15˚) for a total of six boron fiber breaks on each sample. The graph in Fig. 3(b) shows the comparison of experimental and theoretical values, which were calculated using Eq. 1. The experimental results correlate very well to the theoretical speed of sound, derived from the simplified equation Ce = (E/ρ)1/2 . The ratio Ce,exp/Ce,theo for each sample was very close to its theoretical value, which is indicated by the dashed line representing the direct proportion (Ce,exp/Ce,theo = 1) in Fig. 3(b). In general the observed standard deviation within the theoretical speed of sound can be explained by applying an average range of elastic modulus values as well as using the simplified equation without considering the Poisson’s ratio ν.

±45˚ HP717 straightsided HTA 1.85 27±2

can be calculated from the classical plate theory [17]: s s E E Ce,theo = ≈ (1) 2 ρ · (1 − ν ) ρ Ce,theo is a function of the longitudinal elastic modulus E, the material density ρ and the Poisson’s ratio ν, which only has a minor influence, that allows a simplification of the above equation. The theoretical values were compared to an experimentally determined wave velocity Ce,exp . Here, boron fiber breaks were performed outside of the top and bottom AE sensors at the face and the edge of the tensile bar prior to every tensile test. Thus, the sound wave of the manuallygenerated acoustic events travelled across all AE sensors. The experimental speed of sound was then calculated using the distance between the top and the bottom AE sensors ∆x = 60 mm and their difference in times of arrival ∆tx of the first wave peak. Ce,exp =

60 mm ∆x = ∆tx ∆tx

3.2. Acoustic emission 3.2.1. General acoustic emission activity in C/C-SiC composites For the first time the acoustic activity in DLR’s C/C-SiC composites was studied under mechanical loading until failure. This section presents characteristic differences in the degree of damage comparing C/C-SiC composites with different fiber architectures. The total amount of damage-related AE events as well as the number of AE events that were recorded for each AE sensor9 are plotted in Fig. 4 for different fiber orientations. These numbers represent filtered values, where a minimum AE energy filter (0.0005 V2µs) was applied and therefore nondamage-related AE events, i. e. electromagnetic

(2)

The theoretical and the experimental speed of sound were determined for all tensile specimens in the unloaded state. Note that the speed of sound is directly related to any change of the elastic modulus during loading and accumulated damage [5, 6]. Only the ±45˚ composites exhibited significant elastic modulus decrease during the tensile test, see Sec. 3.2. Therefore the speed of sound of the unloaded samples was used to calculated the AE event locations, see Sec. 3.2.3.

9

AE events that hit AE sensor 1, 2 or 3 first which correspond to a designated colored area, see Fig. 2

4

9000 Experimental Ce,exp / m/s

Experimental Ce,exp / m/s

9000

6000

HP643 15°

3000

HP717 0/90°

6000 HP643 15° HP717 0/90° HP717 45° Cexp/Ctheo=1

HP717 45°

3000 3000

0 0

20

40 60 80 Elastic modulus E / GPa

100

120

(a) The experimental speed of sound (calculated from boron fiber breaks) in comparison with the elastic modulus of the unloaded C/C-SiC specimens.

6000 Theoretical Ce,theo = (E/ρ)1/2 / m/s

9000

(b) Correlation of the measured experimental and calculated theoretical wave velocity values.

Fig. 3. Extensional wave velocity Ce (speed of sound) for the tested C/C-SiC composites with different fiber orientations.

interferences, were manually discarded. Major differences were observed for the total number of AE events comparing the different fiber architectures. Within the ±15˚ specimen only 172 AE events were recorded, whereas 2775 AE events were detected within the 0/90˚ specimen and 4670 AE events within the ±45˚ specimen using the same experimental equipment and preamplification settings, indicating significant fiber orientation-dependent damage mechanisms. Note that the failure locations for all specimens were in the blue area close the top AE sensor (AE sensor 1), see Fig. 2. The exact failure location is discussed in Sec. 3.2.3. The number of AE events for each AE sensor was apparently also influenced by the following effects. Within dogbone samples (sepcimens ±15˚ and 0/90˚) a possible radius edge effect was observed, since the top and the bottom AE sensors were close to the end of the gage length (constant cross section), where the sample cross section increased again. The ±15˚ sample showed significantly more AE events for sensor 1 and 2 than the middle sensor. Within the 0/90˚ specimen it is assumed that there was also a possible volumetric effect, since sensor 3 detected almost twice as much AE events than sensor 2. Here, AE sensor 3 (green middle section) covered about twice the volume of the same cross section as the bottom AE sensor. The high amount of AE events of sensor 1 within the ±45˚ specimen resulted from an increased degree

Number of AE events

5000

Total

4670

Sensor 1

4000

Sensor 3 Sensor 2 2775

3000

2490

2000 1397

1196 935

1000

644 172 63

32

783

77

0 HP643 15°

HP717 0/90° C/C-SiC fiber architecture

HP717 45°

Fig. 4. Total number of AE events and the number of AE events for each AE sensor for different C/C-SiC fiber architectures.

of damage due to shear failure. A detailed discussion on comparing differences in AE activity and AE event location within a specific fiber orientation as well as different fiber architectures is presented in the following sections. 3.2.2. Stress-dependent AE activity This section describes the stress-strain curves of the hysteresis tensile tests as well as the stressdependent AE activity. The load/unload/reload stress-strain curves for all three fiber architectures are plotted in Fig. 5(a) to illustrate the fundamental differences in their mechanical behavior. In these C/C-SiC composites the mechanical performance is mainly influenced by the orientation of the carbon fibers, assuming a given appropriate microstructure, as shown in Fig.1. 5

loop until about 15–20 % of their corresponding ultimate tensile strength. With increasing stress more AE events were generated. Note that only very few events occurred during unloading and reloading, which appear as leftward ”spikes” in the cumulative AE energy/events data. The average AE energy of these events were very low, resulting in a horizontal line in the cumulative AE energy plots. Within the ±15˚ specimen a drastic increase in the accumulated AE energy was observed during the last loop after exceeding the previous peak stress of about 190 MPa and up to final failure. These final events corresponded to only 30 % of all AE events but generated more than 90 % of the total accumulated AE energy, see Fig. 5(b). This small stress-strain range just before the final failure was characterized by very ”loud” events with an average AE energy of up to 50 V2 µs. Within the 0/90˚ specimen the curve of cumulative AE events and the curve of cumulative AE energy follow a similar trend, see Fig. 5(c). Only a minor increase in AE activity was found until about 0.25 % strain. By exceeding 130 MPa the AE activity, events and energy, began to increase progressively with increasing stress until final failure. Here the increase of AE activity has a rather smooth transition in contrast to the ±15˚ specimen, which can be explained by the generation of soft and medium loud acoustic events up to 4 V2 µs, resulting in no major AE energy jumps in the cumulative AE energy curve. The ±45˚ specimen showed an even smoother accumulation of AE energy with increasing stress than the 0/90˚ fiber architecture, see Fig. 5(d). Furthermore within the ±45˚ specimen considerably more AE events were recorded at stresses close to the previous peak stress during reloading as well as shortly after unloading for hysteresis loops at high stress levels. The final failure occurred at a lower stress than the peak stress of the previous loop. The AE events that were generated during the final failure were accountable for more than 30 % of the total cumulative AE energy with some loud events up to 16 V2 µs.

C/C-SiC composites with carbon fibers aligned in or near the loading direction, i. e. ±15˚ and 0/90˚, show a distinct linear elastic, hence fiberdominant behavior. As expected, the initial elastic modulus and the ultimate strength increase with the amount of carbon fibers oriented in the loading direction. The stress-strain behavior of the ±45˚ specimen is characterized by a high strain-to-failure value and a clear non-linearity, which results from shear mechanisms that dominate in this load direction. The elastic modulus of the reloading slope decreases with each consecutive unload/reload cycle due to increased damage. The change in modulus is clearly noticeable for the ±45˚ specimen. Here the elastic modulus was reduced by 50 % from 28 GPa (initial tangent modulus) to 14 GPa (reloading tangent modulus, last cycle). The elastic modulus of the fiber-dominant specimens only decreased by 10 % for the ±15˚ and 15 % for the 0/90˚ specimen, respectively. Furthermore the width of the hysteresis loops only increased significantly within the ±45˚ specimen, indicating a much higher degree of matrix debonding effects, frictional sliding and hindrance of crack closure during unloading than within the other specimens. The additional diagrams in Fig. 5 show the stress-strain behavior in more detail as well as their corresponding AE activity. Here, all AE events were considered that were located between the bottom and top AE sensors (AE events between -30, mm and +30 mm). In these graphs the AE events are plotted on top of the stress-strain curve to point out when damage actually occurred throughout the different loading steps. Also the normalized cumulative amount of AE events and the normalized cumulative average AE energy of all events is plotted vs. strain in the same diagram allowing a detailed stress- and strain-dependent AE analysis. The first AE events were detected between 0.032–0.046 % strain, which seems to be a characteristic strain range for these types of C/C-SiC composites where the first very low AE energy events (<0.05 V2 µs) are generated. For all tested fiber architectures no AE events were recorded during the initial linear elastic part of the first 6

250 250

200

45°

200 Stress / MPa

150 ±45°

100 50 0

150

0.6

100

0.4

50

0.2 0.0

0

0.0

0.5

1.0

1.5

0.0

Strain / %

200

0.6 0.4

50

0.2

0

0.0 0.1

0.2 Strain / %

0.3

0.3

1.0

Stress AE events Norm. cum. AE events Norm. cum. AE energy

80 Stress / MPa

Stress AE events Norm. cum. AE events Norm. cum. AE energy

100 Norm. cum. AE energy / events

0.8

150

0.0

0.2

(b) Stress vs. strain and AE activity for ±15˚.

1.0

100

0.1 Strain / %

(a) Stress-strain curves for all three fiber architectures.

Stress / MPa

0.8

0.8

60

0.6

40

0.4

20

0.2

0

0.4

0.0 0.0

(c) Stress vs. strain and AE activity for 0/90˚.

Norm. cum. AE energy / events

Stress / MPa

0/90°

1.0

Stress AE events Norm. cum. AE events Norm. cum. AE energy

0/90°

Norm. cum. AE energy / events

15°

±15°

0.5

Strain / %

1.0

1.5

(d) Stress vs. strain and AE activity for ±45˚.

Fig. 5. Stress-strain behavior during load/unload/reload hysteresis tensile tests and their corresponding AE activity for all three tested fiber architectures. (Considered data: all events between -30 mm and +30 mm.)

amplitude in the 0/90˚ and ±45˚ specimens of this study, the first-threshold-crossing technique led to inaccurate results. Therefore, the Akaike information criterion (AIC)-method [18] was applied. The AIC-method consists in calculating for each point of a waveform the similarity between the signal portions before and after the considered point. The signal onset corresponds to the point where the similarity is minimum between noise prior to onset and highly-correlated signal after the onset. Further details are available in [19]. Recent work [20] showed very promising and most accurate location results using this technique on AE waveforms generated by damage in CMCs. Due to a minimum sampling rate of 10 MHz there is a minimum location error in the range of 0.3–0.8 mm depending on the wave velocity. Additional location error would occur due to inaccurate determination of the signal onset in some waveforms with front end interferences.

3.2.3. Location analysis The longitudinal location of AE events was estimated from the difference in times of arrival of the AE signal onset of the outer sensors (tbottom and ttop ) and the wave velocity Ce using the following equation: xloc =

Ce (tbottom − ttop ) 2

(3)

There are three common ways to determine the required time of arrival at the sensors. The most accurate but very time consuming technique is the manual determination of the first signal peak of the extensional wave. This was done for the ±15˚ specimen, where only some hundred AE events were recorded. The second approach is called first-threshold-crossing technique, in which the arrival time is obtained when the waveform first crosses a predefined amplitude (threshold). Since the recorded waveforms vary significantly in 7

250

20

200

10 0 -10

Ch1 Ch3 Ch2 "Loud">0.5V²µs

-20

150 100 100 50

0

50

100 150 Stress / MPa

200

250

0 0.0

0.1

0.2

0.3

Strain / %

(a) Location of AE events for the ±15˚ specimen. Failure location: +23 mm. 30

(b) Sensor dependent average cumulative AE energy for the ±15˚ specimen. 200

failure location

20 Stress / MPa

150

10 0 Ch1

-10

Ch3 Ch2

-20

200

Stress Ch1 Ch3 Ch2

150

100

100

50

50

Cum. AE energy / V²µs

0

"Loud">0.5V²µs

0

0

-30 0

50

100 Stress / MPa

150

0.0

200

(c) Location of AE events for the 0/90˚ specimen. Failure location: +26 mm.

0.1

0.2 Strain / %

0.3

0.4

(d) Sensor dependent average cumulative AE energy for the 0/90˚ specimen.

30

200

100

20

Stress Ch1 Ch3 Ch2

80 Stress / MPa

10 0 Ch1 Ch3 Ch2 "Loud">0.5V²µs

-10 -20

150

60 100 40 50

20

-30

0

0 0

20

40

60 Stress / MPa

80

100

120

0.0

(e) Location of AE events for the ±45˚ specimen. Failure location: +(40-45) mm (outside of sensors).

Cum. AE Energy

AE event location / mm

150

50

-30

AE event location / mm

200 Stress Ch1 Ch3 Ch2

Cum. AE energy

failure location

Stress / MPa

AE event location / mm

30

0.5

Strain / %

1.0

1.5

(f ) Sensor dependent average cumulative AE energy for the ±45˚ specimen.

Fig. 6. Location of AE events which occurred between the top and bottom sensors using the AIC-technique (left) and comparison of the average cumulative AE energy for AE each sensor (right).

8

entation no significant differences were observed for the sensor-dependent cumulative AE energy curves, see Fig. 6(d). The highest cumulative AE energy was recorded for sensor 3 (middle). It had about 2 times the AE energy of sensor 2 (bottom) which seems to be an area effect since the middle sensor covers twice the area of the gage length compared to the outer sensors. Surprisingly there was no concentration of very loud AE events during final failure at +26 mm. The location plot of the ±45˚ specimen in Fig. 6(e) shows results that are similar to the 0/90˚ fiber architecture. However main differences were the total amount of AE events and a smaller fraction of loud AE events, see Fig. 4 and Tab. 2. A distinct increase in AE activity was observed above 90 MPa. For this fiber orientation a straight sided specimen was used. The concentration of loud AE events at +30 mm likely corresponded to events that were generated from the actual failure location outside of the top AE sensor at 40–45 mm. The location of AE events that occur outside of the AE sensors cannot be calculated correctly using Eq. 3. Therefore those AE events are wrongly located close to one of the outer sensors, here +30 mm, since the difference in times of arrival of those events is always the same and consequently independent from their actual point of creation. The significant increase in cumulative AE energy of sensor 1 (top) represents the damage accumulation during final failure, see Fig. 6(f).

The longitudinal location of all AE events that occurred between the top and bottom sensors are plotted vs. stress for each specimen on the left in Fig. 6. The failure location of each specimen is depicted by a dashed line. The AE events are marked by different colored labels depending on which sensor was triggered first. Furthermore loud AE events that exceeded an average AE energy of 0.5 V2 µs are highlighted by red dots. The graphs on the right in Fig. 6 show the sensordependent average cumulative AE energy which gives additional information on the damage accumulation in the designated gage areas. In general it is expected that soft AE events, that were clearly recorded close to one of the AE sensors, also occurred throughout the entire gage length but were not detected due to attenuation effects [20, 21]. This results in areas which are typically in between two sensors where no or only a few AE events were detected, see Fig. 6(a) and Fig. 6(c). For the ±15˚ specimen soft events started to occur predominantly just above the bottom sensor after exceeding about 50 MPa. Loud events were only recorded more frequently above 200 MPa just prior to final failure. Two concentrated areas of loud AE events were identified at about -25 mm and +23 mm. The latter was the final failure location. Fig. 6(b) clearly shows that damage was accumulated at -25 mm just before the final failure occurred close to the opposite AE sensor at +23 mm. This is represented by an increase in cumulative AE energy of sensor 2 (bottom). In the end the cumulative AE energy curve of sensor 1 surpassed sensor 2 due to very high AE energy events during final failure which were associated with multiple fiber breaks and fiber-bundle breaks. Fig. 6(c) shows that a lot more AE activity was recorded for the 0/90˚ specimen, also see Fig. 4. From 25 MPa to about 130 MPa only very soft AE events occured throughout the entire gage length. Note that blank areas in between sensors were due to attenuation effects. With increasing stress AE events noticeably spread out from the middle of each AE sensor since the AE energy of events also increased. Within this fiber ori-

3.2.4. AE energy and waveform analysis The main goal of the AE energy and waveform analysis is to distinguish between several damage mechanisms/sources during loading by identifying characteristic differences in their recorded waveform (amplitude, energy, frequency). The analysis of AE events showed that the recorded waveforms vary significantly in amplitude and energy, respectively. The AE energy is the area under the waveform. The average AE energy of the three AE sensors was determined for each event. The magnitude of the average AE energy ranged from 0.001 V2 µs, for the smallest detectable damagerelated events, up to very high energy events with 9

1.0

1.0 “soft“

0.8

0/90°

45°

Norm. number of AE events exceeding energy

Norm. number of AE events

15°

“loud“

0.6 0.4 0.2 0.0 <0.01

0.01-0.1 0.1-0.5 0.5-1.0 1-5 Average AE energy / V²µs

(a) Histogram of normalized AE events vs. average AE energy intervals.

0/90°

0.8

45°

0.6

“soft“

“loud“

0.4 0.2 0.0 0.001

>5

15°

0.01

0.1 1 Average AE energy / V²µs

10

(b) Normalized number of AE events which exceed a certain average AE energy on a logarithmic scale.

Fig. 7. Comparison of the number of AE events with different AE energy levels using a histogram diagram (left) and a logarithmic style (right) for all three tested fiber architectures. Table 2 Fraction of loud and soft AE events for different fiber orientations of C/C-SiC composites. Fiber orientation Loud AE events (>0.5 V2 µs) Soft AE events (<0.5 V2 µs)

±15˚ 18 % 82 %

0/90˚ 4% 96 %

see Fig. 7 and Tab. 2. This indicated an additional different characteristic damage source which particularly appeared close to ultimate strength and was responsible for the final failure. A waveform and frequency analysis was performed to determine if certain differences could be found for characteristic AE events that would also help to identify different damage sources. Typical waveforms of soft AE events within the average AE energy range of 0.01–0.1 V2µs as well as loud AE events that occurred close to the final failure are depicted in Fig. 8 and Fig. 9. Here characteristic AE events were selected that occurred in a similar specimen location which also happened to be the failure location. In this case the AE events hit the top AE sensor first allowing a proper comparison between different fiber architectures. The figures show the waveform of sensor 1 (top) and also the waveform of sensor 2 (bottom) to illustrate the degree of wave attenuation. The waveforms of soft AE events which are presented here were very typical and could be found in all fiber orientations throughout all stress levels. A typical waveform of a loud AE event which was common for fiber dominant composites, such as 15˚ or 0/90˚, is shown in Fig. 8(b). These loud AE events were only generated at high stress levels close to the ultimate strength. The frequency content of recorded AE waveforms can be described by determining the weighted frequency average, also called frequency

±45˚ 1% 99 %

a magnitude of more than 5 V2 µs. In order to compare the different composites in this study the average AE energy was used to classify the AE events into different energy levels. Here, AE events that had an average energy value lower than 0.5 V2 µs are considered soft events and those that were larger than 0.5 V2 µs are defined as loud events. Soft AE events dominated for all tested fiber orientations. The fraction of loud AE events increased within composites where carbon fibers were oriented in loading direction, i. e. 0/90˚ and ±15˚ specimen, see Tab. 2. Two plots in Fig. 7 illustrate the distribution of AE events in more detail depending on their average AE energy for all three fiber orientations. The detailed subdivision indicated that in all composites AE events with an average AE energy between 0.01–0.1 V2 µs were recorded most frequently, see Fig. 7(a). These AE events represent a characteristic minor damage source in C/C-SiC composites.Within the ±15˚ specimen the fraction of loud AE events significantly increased in contrast to the other fiber orientations, 10

0.5

1.0

Ch. 1 FC: 770 kHz AE energy: 0.12 V²µs

Ch. 1 FC: 710 kHz AE energy: 8.9 V²µs

Amplitude / V

Amplitude / V

0.5

0.0

0.0

-0.5

-0.5

-1.0 0

20

40

60

80

100

0

20

40

Time / µs

60

80

100

Time / µs

0.5

1.0

Ch. 2 FC: 650 kHz AE energy: 0.03 V²µs

Ch. 2 FC: 530 kHz AE energy: 7.2 V²µs

Amplitude / V

Amplitude / V

0.5

0.0

0.0

-0.5

-0.5

-1.0 0

20

40

60

80

100

0

Time / µs

(a) Soft AE event: no. 68, average AE energy: 0.06 V2 µs, stress: 166 MPa.

20

40

60

80

100

Time / µs

(b) Loud AE event: no. 207, average AE energy: 6 V2 µs, stress: 218 MPa.

Fig. 8. Typical AE waveform of a soft (left) and a loud (right) AE event for the ±15˚ specimen. (Note: top sensor-Ch1, bottom sensor-Ch2)

centroid (FC) value. For all fiber orientations as well as for soft and loud AE events, the average frequency centroid value was found to be 720±120 kHz with one exception. For the ±45˚ specimen, a noticeable decrease in the average frequency centroid value to about 500±80 kHz was observed for the last 400 AE events during final failure (last 1.5 s of the test). A characteristic waveform with a lower frequency centroid value is shown in Fig. 9(b). C/C-SiC composites with a ±45˚ fiber orientation fracture under shear mode in 45˚ with respect to the loading direction. The lower frequency content during final failure is due to shear mechanisms or due to the increased damage state in the composite which does not allow higher frequencies to propagate. Also a rather squarish specimen geometry may have led to a stable frequency content within all fiber orientations which would promote an extensional wave and minimize the flexural part of the waveform.

3.3. Microscopic analysis SEM micrograph images of ”as-received” and tested C/C-SiC specimens are described and discussed in this section. Within C/C-SiC composites micro-cracks are formed during cool-down after siliconization due to the CTE mismatch of carbon fibers and the C-, SiC- and Si-matrix. Here the ”as-received” microstructure is characterized by micro-cracks in the Si-SiC matrix, which sometimes also propagate into C/C-blocks where the carbon matrix locally debonds from the carbon fibers, see Fig. 10(a). In general these cracks are deflected and stopped in front of a perpendicular carbon fiber tow. In areas where fibers are undulated carbon matrix rich regions are typically formed which are usually also pre-cracked, see Fig. 10(b). This typical pre-cracked microstructure applies to all fiber architectures. Only minor differences were observed comparing the microstructure of ”as-received” and tested 11

1.0

0.5

Ch. 1 FC: 430 kHz AE energy: 13.2 V²µs

Ch. 1 FC: 800 kHz AE energy: 0.1 V²µs

Amplitude / V

Amplitude / V

0.5

0.0

0.0

-0.5

-1.0

-0.5 0

20

40

60

80

0

100

20

40

80

100

1.0

0.5

Ch. 2 FC: 320 kHz AE energy: 1.2 V²µs

Ch. 2 FC: 610 kHz AE energy: 0.03 V²µs

0.5 Amplitude / V

Amplitude / V

60 Time / µs

Time / µs

0.0

0.0

-0.5

-1.0

-0.5 0

20

40

60

80

0

100

(a) Soft AE event: no. 51, average AE energy: 0.07 V2 µs, stress: 23 MPa.

20

40

60

80

100

Time / µs

Time / µs

(b) Loud AE event: no. 5652, average AE energy: 5 V2 µs, stress: 96 MPa.

Fig. 9. Typical AE waveform of a soft (left) and a loud (right) AE event for the ±45˚ specimen. (Note: top sensor-Ch1, bottom sensor-Ch2)

tion increased from 10 % for the ±15˚ sample to about 50 % for the 0/90˚ specimen which consequently led to a significant increase in soft AE events, which supports the mechanism assumption and the correlation to soft AE events. Within the ±45˚ specimen most likely all fibers were subjected to debonding effects due to their fiber orientation and in combination with high strain values. Furthermore additional cracks were observed in matrix rich areas which could have been created by increased shear mechanisms. Both effects are responsible for the highest amount of AE events within the ±45˚ specimen. The fracture surfaces of all specimens showed a high degree of fiber pull-out. The fracture surface of a 0/90˚ specimen is shown in Fig. 11. Multiple fiber breaks, the rupture of entire fiberbundles and the involved pull-out mechanisms generated very loud AE events during the final failure. The ±15˚ and 0/90˚ specimens failed

specimens. The analyzed areas close to the failure location exhibited only some bigger cracks that propagated through matrix rich regions, see Fig. 10(c). Fig. 10(d) shows a crack which propagated from a 90˚ fiber tow into a 0˚ fiber tow with local fiber breaks. But still most fiber tows in loading direction showed no visible fiber breaks except the failure location itself. This indirectly inferred that the majority of soft AE events (0.01– 0.1 V2 µs) were created by the propagation and deflection of existing matrix cracks and/or by interfacial debonding mechanisms between the carbon matrix and carbon fibers within fiber tows that were not oriented in loading direction (90˚ and 45˚). This proposed assumption could not be verified by SEM images. It is assumed that these very small debonding cracks are closed again after the tensile load has faded and therefore were not detected by SEM microscopy. The fraction of fibers that were not oriented in loading direc12

(a) Typical pre-cracked mircostructure in an unloaded, ”as-received” 0/90˚ specimen (edge view).

(b) Typical micro-cracks in an unloaded, ”as-received” 0/90˚ specimen (face view).

90°

(c) Increased cracking and crack deflection in matrix rich areas close to the failure location of a loaded 0/90˚ specimen (face view).

0°

(d) Crack propagation from a 90˚ fiber bundle into a 0˚ bundle including fiber breaks of a loaded 0/90˚ specimen close to the failure location (face view).

Fig. 10. Typical polished SEM images of the microstructure of unloaded and loaded C/C-SiC specimens.

along a 90˚ fiber tow within 1–2 mm (longitudinal length) whereas the ±45˚ failed under 45˚ in a shear mode.

results were in good agreement with the theoretical values and were found to be between 3800–7800 m/s depending on their fiber orientation and elastic modulus, respectively. 2. Major differences in the amount of recorded acoustic emission events were found comparing different fiber architectures. It was found that the overall AE activity drastically increased within fiber architectures with increasing fiber angles with respect to the loading direction (off-axis fiber orientations).

4. Conclusion C/C-SiC composites with different fiber orientations were tested in load/unload/reload tensile tests and modal acoustic emission technique was applied to study the damage accumulation during loading. 1. The wave velocity was determined for different fiber orientations. The experimental 13

that C/C-SiC composites can be loaded to 80-90 % of their representative ultimate tensile strength without any major damage effects. 5. Microscopy analysis showed that there was only little difference between ”as-processed” and tested micrographs. Note that the matrix of C/C-SiC composites are already precracked after the manufacturing process. This indirectly inferred that the majority of soft AE events (0.01–0.1 V2µs) were created by the propagation and deflection of existing matrix cracks and/or by interfacial debonding mechanisms between the carbon matrix and carbon fibers within fiber tows that were not oriented in loading direction (90˚ and 45˚). These damage sources only had a minor influence on the mechanical performance of fiber dominant composites, such as ±15˚ and 0/90˚. For a ±45˚ test orientation, the increased matrix cracking led to a noticeable decrease in the elastic modulus and in combination with shear effects high strain to failure values were observed. Only few fiber breaks were found in loaded micrographs which implies that loud AE events were created by multiple fiber and fiber-bundle breaks close to the ultimate strength which eventually led to the catastrophic failure with typical fiber pullout fracture surfaces. 6. Rather high attenuation effects were observed where soft AE events could not be detected in between AE sensors. The frequency analysis exhibited a stable frequency content within all tested fiber architectures.

Fig. 11. Fracture surface of a 0/90˚ C/C-SiC specimens with typical fiber pull-out.

3. The stress-dependent damage accumulation was studied by analyzing the cumulative AE energy curves. First damage-related AE events were detected at about 0.03– 0.04 % strain. The AE energy distribution showed that soft AE events with an average AE energy of 0.01–0.1 V2µs were recorded most and represented about 70– 80 % of all damage-related events. They occurred throughout all stress levels. The amount of AE events increased with increasing stress. The majority of AE energy was generated at high stress levels including loud AE events (>0.5 V2 µs) which were only created close to the ultimate stress. Only very few AE events were detected during unloading and reloading which indicates that crack opening and closure as well as frictional sliding mechanisms take place without a detectable energy release. 4. The location analysis and AE sensor dependent AE energy evaluation demonstrated the damage evolution throughout the specimen. However, in terms of health monitoring aspects a precise prediction of the actual failure location before the final rupture of the specimen is difficult to determine. Loud AE events which eventually lead to failure are typically generated just before or during final failure. This on the other hand means

It is important to mention that the investigated C/C-SiC composites showed some significant differences compared to recently studied SiC/SiC composites [5, 7, 9, 11, 12] using the same AE monitoring technique. Within CVI-10 and MI11 -SiC/SiC composites high energy events occur early in the test which correspond to matrix 10 11

14

CVI: chemical vapor infiltration MI: melt infiltration

cracks. Note, ”as-processed” SiC/SiC materials usually exhibit no or only minor matrix cracks, since there is no significant mismatch in the thermal expansion coefficient of SiC fibers and SiC matrix during high temperature processing. Also, matrix cracking accumulates most of the AE energy during loading and significantly reduces the modulus for fiber dominant architectures, which is not the case for C/C-SiC composites. Attenuation effects are also different [6] which makes it important to study C/C-SiC composites and gain experience when interpreting AE results. Another promising accurate method for identification of different damage sources could be accomplished using in-situ nano-CT imagery during loading of representative mini-composites [22]. Next to the acoustic emission technique the insitu measurement of electrical resistivity can be used for health monitoring of CMCs [23–25]. The electrical resistivity measurement was also applied in this study. The results will be presented in a separate publication.

[4]

[5]

[6]

[7]

[8]

[9]

[10]

Acknowledgements The authors gratefully acknowledge the German Research Foundation (DFG) for supporting this work within the framework of the SFB-TRR 40. The main author also highly appreciated the research stay at the Mechanical Engineering Department of the University of Akron and Prof. Dr. G. N. Morscher for making this collaborative research effort possible as well as Emmanuel Maillet’s scientific support.

[11]

[12]

References [13]

[1] F. Christin, Design, Fabrication, and Application of Thermostructural Composites (TSC) like C/C, C/SiC, and SiC/SiC Composites, Advanced Engineering Materials 4 (12) (2002) 903–912. [2] S. Schmidt, S. Beyer, H. Immich, H. Knabe, R. Meistring, A. Gessler, Ceramic Matrix Composites: A Challenge in Space-Propulsion Technology Applications, International Journal of Applied Ceramic Technology 2 (2) (2005) 85–96. [3] F. Breede, R. Jemmali, H. Voggenreiter, D. Koch, Design and Testing of a C/C-SiC Nozzle Extension Manufactured Via Filament Winding Technique

[14]

[15]

15

and Liquid Silicon Infiltration, Design, Development, and Applications of Structural Ceramics, Composites, and Nanomaterials: Ceramic Transactions 244 (2014) 3–14. W. Hendrik, J. Longo, J. Turner, The Sharp Edge Flight Experiment SHEFEX II, a Mission Overview and Status, International Space Planes and Hypersonic Systems and Technologies Conferences, American Institute of Aeronautics and Astronautics, 2008. G. N. Morscher, Modal acoustic emission of damage accumulation in a woven SiC/SiC composite, Composites Science and Technology 59 (5) (1999) 687– 697. G. N. Morscher, A. L. Gyekenyesi, The velocity and attenuation of acoustic emission waves in SiC/SiC composites loaded in tension, Composites Science and Technology 62 (9) (2002) 1171–1180. G. N. Morscher, Stress-dependent matrix cracking in 2D woven SiC-fiber reinforced melt-infiltrated SiC matrix composites, Composites Science and Technology 64 (9) (2004) 1311–1319. G. N. Morscher, Modeling the elastic modulus of 2D woven CVI SiC composites, Composites Science and Technology 66 (15) (2006) 2804–2814. A. L. Gyekenyesi, G. N. Morscher, L. M. Cosgriff, In situ monitoring of damage in SiC/SiC composites using acousto-ultrasonics, Composites Part B: Engineering 37 (1) (2006) 47–53. G. N. Morscher, M. Singh, J. D. Kiser, M. Freedman, R. Bhatt, Modeling stress-dependent matrix cracking and stress-strain behavior in 2D woven SiC fiber reinforced CVI SiC composites, Composites Science and Technology 67 (6) (2007) 1009–1017. G. N. Morscher, H. M. Yun, J. A. DiCarlo, In-Plane Cracking Behavior and Ultimate Strength for 2D Woven and Braided Melt-Infiltrated SiC/SiC Composites Tensile Loaded in Off-Axis Fiber Directions, Journal of the American Ceramic Society 90 (10) (2007) 3185–3193. G. N. Morscher, J. A. DiCarlo, J. D. Kiser, H. M. Yun, Effects of Fiber Architecture on Matrix Cracking for Melt-Infiltrated SiC/SiC Composites, International Journal of Applied Ceramic Technology 7 (3) (2010) 276–290. G. N. Morscher, N. Godin, Use of Acoustic Emission for Ceramic Matrix Composites, John Wiley & Sons, Inc., 569–590, 2014. F. Breede, S. Hofmann, E. Klatt, S. Denis, Influence of Fiber Orientation on the Mechanical Properties and Microstructure of C/C-SiC Composite Plates Produced by Wet Filament Winding Technique, Processing and Properties of Advanced Ceramics and Composites III (2011) 3–10. F. Breede, Mechanical and Microstructural Characterization of C/C-SiC Manufactured Via Triaxial and Biaxial Braided Fiber Preforms, Mechanical Proper-

[16]

[17]

[18]

[19]

[20]

[21]

[22]

[23]

[24]

[25]

ties and Performance of Engineering Ceramics and Composites VII: Ceramic Engineering and Science Proceedings 33 (2) (2012) 183–194. B. Heidenreich, Manufacture and Applications of C/C-SiC and C/SiC Composites, Processing and Properties of Advanced Ceramics and Composites IV (2012) 183–198. M. R. Gorman, Plate wave acoustic emission, The Journal of the Acoustical Society of America 90 (1) (1991) 358–364. H. Akaike, Markovian representation of stochastic processes and its application to the analysis of autoregressive moving average processes, Annals of the Institute of Statistical Mathematics 26 (1) (1974) 363– 387. P. Sedlak, Y. Hirose, M. Enoki, Acoustic emission localization in thin multi-layer plates using first-arrival determination, Mechanical Systems and Signal Processing 36 (2) (2013) 636–649. E. Maillet, G. N. Morscher, Waveform-based selection of acoustic emission events generated by damage in composite materials, Mechanical Systems and Signal Processing 5253 (0) (2015) 217–227. E. Maillet, N. Godin, M. RMili, P. Reynaud, G. Fantozzi, J. Lamon, Real-time evaluation of energy attenuation: A novel approach to acoustic emission analysis for damage monitoring of ceramic matrix composites, Journal of the European Ceramic Society 34 (7) (2014) 1673–1679. H. A. Bale, A. Haboub, A. A. MacDowell, J. R. Nasiatka, D. Y. Parkinson, B. N. Cox, D. B. Marshall, R. O. Ritchie, Real-time quantitative imaging of failure events in materials under load at temperatures above 1,600C, Nat Mater 12 (1) (2013) 40–46. G. N. Morscher, C. Baker, C. Smith, Electrical resistance of SiC fiber reinforced SiC/Si matrix composites at room temperature during tensile testing, International Journal of Applied Ceramic Technology 11 (2) (2014) 263–272. G. N. Morscher, C. Smith, E. Maillet, C. Baker, R. Mansour, Electrical resistance monitoring of damage and crack growth in advanced SiC-based ceramic composites, American Ceramic Society Bulletin 93 (7) (2014) 4. H. Boehrk, P. Leschinski, T. Reimer, Electrical resistivity measurement of carbon-fiber-reinforced ceramic matrix composite under thermo-mechanical load, Composites Science and Technology 76 (0) (2013) 1–7.

16