An in-process cure monitoring technique based on “guided wave” concept for carbon fiber reinforced polymer (CFRP) composites was developed. Key parameters including physical properties (viscosity and degree of cure) and state transitions (gelation and vitrification) during the cure cycle were clearly identified experimentally from the amplitude and group velocity of guided waves, validated via the semi-empirical cure process modeling software RAVEN. Using the newly developed cure monitoring system, an array of high-temperature piezoelectric transducers acting as an actuator and sensors were employed to excite and sense guided wave signals, in terms of voltage, through unidirectional composite panels fabricated from Hexcel® IM7/8552 prepreg during cure in an oven. Average normalized peak voltage, which pertains to the wave amplitude, was selected as a metric to describe the guided waves phenomena throughout the entire cure cycle. During the transition from rubbery to glassy state, the group velocity of the guided waves was investigated for connection with degree of cure, Tg, and mechanical properties. This work demonstrated the feasibility of in-process cure monitoring and continued progress toward a closed-loop process control to maximize composite part quality and consistency.

## Introduction

Current cure monitoring techniques such as differential scanning calorimetry (DSC), rheology, dynamic mechanical analysis, and thermal gravimetric analysis (TGA) are material characterization tests for thermal, physical, and mechanical properties of a resin during cure. These techniques are usually conducted on a small sample in a controlled laboratory instrument. Thus, they are unlikely to be expanded for in-process cure monitoring of composites without major hardware modification [1].

Another widely used cure monitoring technique, dielectric analysis (DEA), utilizes a loss factor, ε″, to monitor the different stages of cure (e.g., in glass-fiber reinforced epoxy [2]). The degree of cure measured from DEA correlated well with DSC for isothermal cure of epoxy resins [3]. During the resin transfer molding process, the maximum of the ionic conductivity indicated minimum viscosity. The first zero slope of the derivative of log of the ionic conductivity (DLIC), after minimum viscosity, demarked the onset of gelation during cure. Derivative of log of the ionic conductivity also loosely correlated with degree of cure from onset of gelation to full cure and the DLIC plateau estimated vitrification when compared with DSC [4]. Although these tests were performed in a laboratory environment, the principle of DEA has the potential to be implemented for in situ cure monitoring at full scale.

Conventional bulk wave ultrasound has also been implemented as a cure monitoring technique. For thermoset resins, ultrasonic velocity has been used to infer the degree of cure because of its direct relation with the modulus of the resin [57]. Bulk wave contact ultrasound, usually conducted in pulse-echo mode in the low megahertz frequency range, can monitor the completion of resin cure based on when the time delay plateaus/ultrasonic velocity becomes constant value in graphite/epoxy composites [8] and epoxy resins [9]. Other ultrasonic phenomena have also been used for monitoring degree of cure including, attenuation (i.e., amplitude of signal) [7,8,1013], instantaneous phase, and the “mean value of each frequency curve weighted by the maximum corresponding spectral amplitude” [14].

Noncontact air-coupled ultrasonic techniques have been recently employed for cure monitoring without the need for resin-transducer contact. However, this noncontact technique needs to consider the exothermal behavior of the cure process as well as the alteration of acoustic air path. Because the temperature varies greatly throughout the cure process, the ultrasonic velocity in the air varies accordingly. This change in air velocity impacts the measured data since the transducers are not in direct contact with the composite and must be accounted for to determine the correct time of flight in the composite [1].

The guided wave system proposed in this work was designed to complement the current cure monitoring technologies, but not to replace them. This is especially true when comparing with DEA that has been used for cure monitoring of carbon fiber reinforced polymer (CFRP) composites in the laboratory. The guided wave system operates at a higher frequency (in the range of low 100 kHz) than DEA which may provide benefits for potential defect detection. However, defect detection during cure is beyond the scope of this initial study.

The guided wave system has the advantage of being in situ incorporated directly into the standard curing equipment and technique. This system is not a stand-alone bench-top laboratory system where a small amount of material is tested outside its production environment (e.g., DSC, rheology, dynamic mechanical analysis, thermal gravimetric analysis). The system can be scaled up from producing flat composite panels to full-scale complex structures utilized in aerospace and space applications (e.g., cylinders/barrels, wing skins, etc.). The current system is composed of a robust linear array of high-temperature piezoelectric transducers incorporating enhancements identified during its development. Such enhancements include the addition of five more sensors (original system had three), operating with optimized equipment settings (e.g., actuation volts peak-to-peak and center frequency), and more efficient data processing and display [15]. In the future, the linear array of high-temperature piezoelectric transducers could be simply superseded by a single multiplexed optical fiber with phase-shifted fiber Bragg gratings embedded inside the composite panel which can sense wave signals during cure and in-service, enabling life-cycle monitoring.

In addition, this guided wave approach, unlike conventional bulk wave ultrasound that merely provides information about the panel directly underneath or near the ultrasonic transducer, interrogates a continuous wave path through the thickness of the panel along the line from actuator to sensor. This distinct advantage permits the overall response of a large area of the composite to be monitored using a pair of piezoelectric transducers through the guided wave approach, whereas information is only gathered about a localized area directly underneath the transducer with the traditional bulk wave ultrasonic approach. Using a guided wave approach in CFRP composites, it has been recently demonstrated that the group velocity of guided waves increased as degree of cure of the composite was higher and the porosity level reduced [16,17]. However, this work was done using three cured composite panels. If key cure parameters (e.g., viscosity, degree of cure, Tg, gelation, vitrification, porosity) can be estimated during cure, the process parameters can be dynamically tuned based on the measurements. This would prevent the operator from having to strictly follow a nonoptimized fixed cure cycle.

This work examined the amplitude and group velocity of guided waves in CFRP panels (size: 610 mm × 178 mm) during cure in real-time. The feasibility of using these two time-dependent values to monitor physical properties (viscosity and degree of cure) and identify state transitions (gelation and vitrification) during the cure cycle was investigated. The transformation of the resin from the liquid to rubbery state is typically referred to as gelation. The subsequent transformation from the rubbery state to the glassy state is defined as vitrification [18]. A generalized time-temperature-transformation diagram [1820] for a typical epoxy resin (e.g., Hexcel® 8552) under isothermal cure is shown in Fig. 1 denoting these state transformations. In the figure, Tg0 denotes the glass transition temperature of the initial formulation (completely uncured), Tg∞ represents the glass transition temperature of the cross-linked resin at full cure [20], sol refers to solvent soluble (i.e., ungelled), and gel refers to solvent insoluble [18,19].

To validate experimental results, a semi-empirical cure process model was simulated with the specific cure parameters used in the experiment. The modeling flowchart (Fig. 2) illustrates the modeling procedure used to validate state transitions (gelation and vitrification) and physical parameters from simulation using guided wave-based measurements. As seen in Fig. 2, material characterization tests were used as inputs for a material model, which provides physical parameters for generic cure cycles. These characterization tests are laboratory-scale testing methods (e.g., DSC and rheology) that provide valuable information about the thermal and flow characteristics of the material. Knowledge of these characteristics is useful for cure response modeling [3]. For testing of composite structures, layup characteristics and temperature and pressure values specific to the simulated cure cycle must be specified. The material-specific models can then be utilized to simulate the cure kinetics of the actual composite under the experimental conditions. The RAVEN simulations carried out for this work employed this process.

## Material and Methods

Two 24 ply panels were hand laid up using IM7/8552, 35% resin content, 190 gsm unidirectional prepreg (Hexcel Corporation, Stamford, CT). The panels were 610 mm × 178 mm × 4.6 mm (nominal) and the layup was [024]. The panels were cured in an oven (Fig. 3) following two commonly used cure cycles for this material system. The first cure cycle was a two-stage cure with a B-stage hold. The temperature was ramped to 107 °C at 2.8 °C/min, held 1 h, ramped to 177 °C at 2.8 °C/min, held 2 h, then cooled down. The second cure cycle, commonly used in industry for thin laminates to save time, energy, and thus money removed the B-stage (107 °C) hold and ramped directly to 177 °C at 2.8 °C/min, held 2 h, then cooled down. Both panels were interrogated during the entire cure period through guided waves.

A guided wave was excited into the plate using a five-cycle, Hanning windowed, sinusoidal toneburst signal emitted from a waveform generator (Agilent Technologies, Santa Clara, CA: 81150A) to an amplifier (Krohn-Hite Corporation, Brockton, MA: Model 7602M) to a circular piezoelectric transducer (Physical Acoustics Corporation, Princeton Junction, NJ: Nano-30 (ø7.9 mm, height: 7.1 mm, frequency range: 150–750 kHz, resonant frequency: ∼300 kHz)) that was rated for use up to 177 °C. (Note: The maximum cure temperature of 177 °C is the same as the maximum published operating temperature. Temperature studies of the transducers were performed to isolate any temperature effects. These studies ensured the trends discussed in the results are due to the curing of the composite and not the temperature effect on the transducer. The authors expect some degradation in the transducers if repeatedly used above the operating temperature.)

The amplifier magnified the input signal to a peak-to-peak voltage of approximately 120 V. The panel response along the fiber direction was recorded by identical piezoelectric transducers in a pitch-catch configuration on two oscilloscopes (Agilent Technologies: MSO9064A and Tektronix, Beaverton, OR: MSO3014). Experimental setup for the automated guided wave system for in-process cure monitoring is shown in Fig. 3.

Fluorinated ethylene propylene release film was placed on the top and bottom of the composite panel. The sensors (Physical Acoustics Corporation: Nano-30) were bonded using high temperature red room-temperature-vulcanizing silicone (VersaChem® (Hartford, CT) Part No. 65300) to a thin (0.1 mm) sheet of steel (“caul” plate) that was placed on top of the panel which prevented the sensors from being pressed into the panel during cure while still allowing the guided wave which propagates in the composite to be measured. The eight sensors were located at a distance, x, from the actuator of 57, 76, 95, 114, 133, 152, 171, and 190 mm, respectively. A breather cloth and vacuum bag covered the panel and full vacuum was applied using a vacuum pump (Fig. 3). An oven was used to cure the panel as part of the building-block approach to developing the guided wave system for use in an autoclave where the recommended 690 kPa pressure can be used. By following this approach, the port in the back of the oven could be utilized for the ingress/egress of all cables during development of the system prior to making modifications to the autoclave. It should be noted that the goal of the current work was not to produce a high-quality composite panel but rather to develop and test the guided wave system at elevated temperatures during cure and determine what meaningful information can be derived from the results.

The automation code for the sensing system was written in MATLAB and utilized the instrument control toolbox to control both the waveform generator and the oscilloscopes. The general procedure of the algorithm is outlined in Fig. 4. The center frequency of the five-cycle, Hanning windowed, sinusoidal toneburst signal was set on the waveform generator. During each iteration, the center frequency was cycled through 14 frequencies (100, 110, 120, 130, 140, 150, 175, 200, 225, 250, 275, 300, 325, and 350 kHz). The range of voltages to be measured by the oscilloscope was set based on the peak voltage recorded at that center frequency on the previous iteration. Dynamically scaling the range on the oscilloscope based on the previous iteration ensured that the range was minimized to increase signal to noise ratio, while keeping it large enough to prevent the recorded voltage from being cutoff. After the equipment was set, 16 measurements were averaged on the oscilloscope and transferred to the computer. These data were processed through a bandpass filter and analyzed in real-time using MATLAB. The instantaneous recorded waveforms, the full time history waveforms, and key metrics such as peak voltage were all displayed on-screen during cure. This process, which was completely automated by a personal computer running MATLAB scripts and functions, was iterated throughout the cure of the part.

## Results

### Data Collection.

The panel response at sensor 4 (Fig. 3) for five-cycle, Hanning windowed, sinusoidal toneburst actuation with center frequency 140 kHz at cure time around 270 min is shown in Fig. 5. At this center frequency, A0 was the dominant wave mode.

By assembling each individual waveform data at a particular sensor and center frequency of actuation, the full time history of the panel response can be viewed as a three-dimensional (3D) surface and contour plot. This is shown for sensor 4 with an actuation center frequency of 140 kHz in Fig. 6. Figure 5 is the last “slice” of Fig. 6.

Similar to Fig. 6, the full time history of the panel response at sensor 4 excited by a five-cycle, Hanning windowed, sinusoidal toneburst actuation with a center frequency of 300 kHz is shown as a three-dimensional surface and contour plot in Fig. 7.

At a center frequency of 140 kHz (Fig. 6), the A0 wave mode was dominant in the glassy state, whereas the S0 wave mode was dominant in the glassy state at a center frequency of 300 kHz (Fig. 7). This can be seen in the surface/contour plots from an oven time of about 170 min to the end of cure. Trends in wave amplitude and time of arrival (TOA) for these plots are discussed later.

### Cure Kinetic Simulations.

In order to validate the experimental results, a simulation of the cure response (Figs. 8 and 9) was performed using RAVEN composite process simulation software. The material model utilized in RAVEN is a semi-empirical material model [21] based on lab-scale tests (e.g., DSC and rheology). The average temperature of the two part thermocouples (Fig. 3) was modeled as the temperature of the panel. The degree of cure (Eq. (1)), cure rate (Eq. (2)), and resin viscosity (Eq. (3)) were direct outputs of the simulation
$σ(t)=∫tit(q˙−q˙baseline)dtHT$
(1)

$dσdt=σt−σt−ΔtΔt$
(2)

$η={η01eE1RT+η02eE2RT(σgσg−σ)A+Bσ+Cσ2 η<ηmaxηmax η≥ηmax}$
(3)

In Eq. (1), σ is degree of cure, $q˙$ is specific heat flow, $q˙baseline$ is the baseline heat flow, and HT is total reaction heat. Equation (2) is simply a time derivative of Eq. (1). In Eq. (3), $η$ is the viscosity, R is the universal gas constant (8.314 J/K mol), T is temperature, σ is degree of cure, and the remaining variables are model fit parameters to rheology tests ($η01$ = 7.5 × 10−11 Pa·s, $η02$ = 4.81 × 10−2 Pa·s, E1 = 81,908 J/mol, E1 = 13,228 J/mol, σg = 0.545, A = 2.466, B = 0.0, C = 0.0, $ηmax$ = 1.0 × 106 Pa·s) [21].

A one-dimensional (drill-through) analysis was also performed in RAVEN, which accounted for temperature variation throughout the thickness. An insignificant difference in the composite curing response (viscosity, degree of cure, cure rate, gelation time, and vitrification point) was observed compared to the material-only simulation, which assumed constant temperature through the thickness. This is primarily because the composite panel is thin; therefore, the temperature gradient across the thickness was negligible. For the one-dimensional comparison analysis, the recorded air temperature of the oven and the average temperature of the two part thermocouples (modeled as the temperature of the vacuum bag) were utilized as inputs to the model. Heat transfer coefficients were applied at the boundaries of the model, which included each material layer listed in the materials and methods section. A heat transfer coefficient of 30 W/m2 K was used throughout all simulations.

During the first temperature ramp (to 107 °C), the viscosity (Eq. (3)) drops and the resin becomes a liquid. The resin reaches minimum viscosity during the second temperature ramp (to 177 °C) prior to gelation. Upon gelation, the viscosity spikes (Fig. 8).

The degree of cure (Fig. 9) follows an S-shaped curve with the majority of cure occurring during the second temperature ramp (to 177 °C) and early part of the 177 °C temperature hold (from around oven time 130–190 min). The cure rate is the time derivative of the degree of cure and is a maximum just before gelation at an oven time around 145 min. The glass transition temperature (Tg) was calculated using the DeBenedetto equation
$Tg=Tg0+λσ1−(1−λ)σ(Tg∞−Tg0)$
(4)

where σ is the degree of cure and λ = 0.78, Tg0 = −7 °C, and Tg∞ = 250 °C which are model parameters for Hexcel 8552 resin that are fit to experimental data during material characterization [21]. Vitrification is the process in which the curing composite transitions to a glassy state. The onset of vitrification occurs as the Tg approaches the temperature of the part being cured (Fig. 9) [1]. The rubbery state and glassy state in the following figures are analogous to sol/gel rubber and sol/gel glass in Fig. 1, respectively.

### Data Analysis and Correlation to Cure Kinetic Simulations.

The peak voltage, Vpeak, from every measurement was determined as the maximum of the measured signal (Eq. (5))
$Vpeak(OTi,xj,fck)=maxl(V(OTi,xj,fck,tl))$
(5)

where OTi, xj, and $fck$ denote the discrete oven time, location (i.e., sensor), and center frequency of actuation, respectively, at which the measurement was taken. For reference, the panel response in Fig. 5 was taken at OT = 272 min, x = 114 mm (sensor 4), and fc = 140 kHz and the resulting peak voltage, Vpeak, of the signal was approximately 1.7 mV.

The peak voltages were then normalized by dividing the peak voltage value from every measurement by the maximum voltage observed during the entire cure cycle by that sensor and actuation frequency (Eq. (6)). This normalized every peak voltage value to a number between zero and one. The maximum value of one occurs at the peak voltage and does not necessarily have to be at the end of cure cycle
$Vnorm,peak(OTi,xj,fck)=Vpeak(OTi,xj,fck)maxi(Vpeak(OTi,xj,fck))$
(6)
All sensors measure the guided waves at an identical oven time in one measurement; however, since the system loops through the actuation center frequencies, measurements at different actuation center frequencies occur at slightly different oven times (approximately 30 s increments). Thus, Vnorm,peak was first interpolated at defined oven times. Then, Vnorm,peak was averaged across sensor and center actuation frequency at defined oven times (Eq. (7))
$Vavg,norm,peak(OT)=1np∑k=1p∑j=1nVnorm,peak(OT,xj,fck)$
(7)

The average normalized peak voltages, Vavg,norm,peak, are the averaged results from every sensor and actuation frequency throughout the entire cure cycle. This was done to remove sensor and frequency variation in the results allowing for cleaner interpretation and communication of the effect of curing on the amplitude of the guided waves. It is expected that there is information of value in the frequency dependence of the results, which was beyond the scope of this initial study. Figure 10 displays the air temperature and the average part thermocouple temperature recorded by the oven as well as the average normalized peak voltages of the guided waves throughout the cure cycle.

Figure 11 shows part temperature, resin viscosity, and average normalized peak voltage for the liquid and rubbery states during cure. As is apparent in this plot, average normalized peak voltage and resin viscosity have approximately an inverse relationship. As a result, average normalized peak voltage can be considered a good indicator of resin viscosity during these stages of cure.

In Fig. 12, the average normalized group velocities of the A0 lamb wave mode for six excitation frequencies (120, 130, 140, 150, 175, and 200 kHz) as well as the air temperature and the average part thermocouple temperature are shown. The authors consider the data presented as the “apparent” group velocity since only wave envelope velocity was investigated without considering the detail of the dissipative attenuation. The A0 lamb wave mode was the dominant wave mode at these frequencies when the resin of the composite was in the glassy state or in the transition from the rubbery to the glassy state (oven time approximately 175–270 min). This can be visualized more clearly in Fig. 6. To determine the group velocity, the TOA was identified by finding the measurement time at which Vpeak occurred. A linear fit was then made to the sensor location, x, and TOA at each frequency and oven time. The slope of this fit is the group velocity, cg (Eq. (8))
$x=cg×TOA+B$
(8)

where B is a constant of the linear fit.

The group velocity was then averaged across center actuation frequency (Eq. (9)) at defined oven times and then normalized (Eq. (10)). The group velocity values were first interpolated at defined oven times since measurements at different actuation center frequencies occur at slightly different oven times (approximately 30 s increments). As with peak voltage, the averaging across center frequency allows for simplified interpretation of the effect of curing on the group velocity of the guided waves. Again, it is expected that there is information of value in the dispersive nature of the results, but this was beyond the scope of this initial study
$cg,avg(OT)=1p∑k=1pcg(OT,fck)$
(9)

$cg,norm,avg(OT)=cg,avg(OT)max(cg,avg(OT))$
(10)

Figures 13(a)13(f) are analogous to Figs. 611 for the modified cure cycle without the B-stage hold.

## Discussion

The key transition points during the cure cycle can be identified from the developed cure monitoring system. First, during the initial temperature ramp the viscosity drops and the resin begins to flow and the composite begins to consolidate. During this time, an increase in the average normalized peak voltage of the guided waves was observed (Figs. 11 and 13(f)). During the ramp to 177 °C, the viscosity dropped to a minimum and this corresponded to the maximum average normalized peak voltage (Figs. 11 and 13(f)).

The authors are interested in further study into which type of disturbance occurs when the resin is in liquid state. A deeper analysis, capable of modeling the ultrasonic propagation in all states of resin (liquid, rubbery, and glassy), is being investigated but was beyond the scope of this initial work.

When the resin began to gel (during the ramp to 177 °C), the average normalized peak voltage started to decrease and continued this trend throughout gelation (Figs. 10 and 13(e)). Although denoted as a single point in time, it should be understood that gelation is a process and does not occur instantaneously. The gelation “point” is simply the time prior to which one would describe the resin as a liquid and after which one would describe the resin as a rubber. The decrease of normalized amplitude during gelation is best understood by the analogy of dropping a pebble in a lake compared to honey. The wave generated would have a higher amplitude (lower attenuation) in water than in the thicker, higher viscosity honey.

Lastly, the average normalized peak voltage increased throughout vitrification (transition from rubbery to glassy state) before reaching a plateau near the end of cure (Figs. 10 and 13(f)). The vitrification point denoted in both the simulation and experimental results was calculated by simulation from the semi-empirical model. To identify the vitrification point solely from the experimental results would be difficult, and unnecessary. One can easily identify that vitrification is occurring/has occurred by the increase in amplitude and subsequent plateau during the transition from rubbery to glassy state. Similar to gelation, vitrification is a process and the vitrification point is simply the time prior to which one would describe the resin as rubbery and after which one would describe the resin as a glassy. It does not occur instantaneously at a single point in time. The time of vitrification from simulation was consistent with previously published works on composites fabricated with Hexcel 8552 resin [2224].

These trends were observed in both cure cycles. An interesting difference between the two cure cycles was the ratio of the average normalized peak voltage when the resin was in the liquid state (including at minimum viscosity) compared with the glassy state near end of cure. The average normalized peak voltage was higher in the liquid state than the glassy state for the two-stage cure cycle (Fig. 10) when more cure time was spent in the liquid state, whereas the average normalized peak voltage was higher in the glassy state than in the liquid state for the cure cycle without the B-stage hold (Fig. 13(e)). More investigation would be required to verify this trend as repeatable and understand its cause.

Next, as the degree of cure and Tg (Fig. 8) from the Hexcel 8552® material model [21] simulated by RAVEN increased, the group velocity (Fig. 12) of the guided waves increased (i.e., degree of cure was directly proportional to group velocity). In addition, the group velocity curve resembled the upper half of the S-shape degree of cure and Tg curves. Although a quantitative equation for degree of cure and Tg based solely on group velocity was not presented in this work, the group velocity curve could be used as a qualitative measure to predict the degree of cure of the composite. The average normalized group velocity was reported in this preliminary work and further investigation is needed to understand the frequency dependence of the group velocity in this complex environment. Factors such as temperature, modulus of the composite, and material structure could all contribute to the frequency dependence. Each of these trends was consistent over the frequency range investigated when the A0 wave mode was dominant. The A0 wave mode became dominant near vitrification as the resin transitioned from the rubbery to glassy state. The development of the A0 wave mode can be clearly seen in Fig. 6 beginning at an oven time of approximately 175 min.

As mentioned previously, the goal of this work was not to produce a high quality composite panel but rather to develop and test the system and quality of the signals recorded during cure at elevated temperatures inside an oven. Additional modifications will be required to ensure the panels fabricated have no sign of mark-off in the regions where the sensors were placed above the panel on the caul.

Possible future research with regards to this preliminary system includes developing dispersion curves for the composites during cure by understanding the wavenumber, phase velocity, and group velocity frequency dependence throughout the cure cycle. This will require denser spacing of the sensors to measure higher wavenumbers (smaller wavelengths) as with the A0 wave mode. In addition, future work includes investigating the connection between the physical, viscoelastic, and mechanical properties of the composite during cure with the group velocity and attenuation.

## Conclusions

In summary, an automated cure monitoring system employing an array of high-temperature piezoelectric transducers was developed to interrogate 24 ply unidirectional composite panels. The experimental results (identification of flow, minimum viscosity, gelation, and vitrification) based on guided wave concept correlate quite well with the predictions of the semi-empirical material model from a cure process modeling software RAVEN. The average normalized peak voltage and resin viscosity have approximately an inverse relationship during the liquid and rubbery states of cure. As a result, the average normalized peak voltage can be considered a good indicator of resin viscosity during these stages of cure and can be used to identify gelation. The averaged normalized peak voltage is an averaged metric across all frequencies and sensor array that is directly proportional to wave amplitude (inversely proportional to wave attenuation between actuator and sensor). In addition, degree of cure and Tg were directly proportional to group velocity during the glassy state.

The system and signal processing initially developed in this work has the potential to be used in the future to dynamically control the cure cycle in a closed-loop process to maximize composite panel quality and consistency. This is possible because the data acquisition and subsequent analysis of the guided waves can be performed almost simultaneously in near real-time during cure. The guided wave system is incorporated directly into standard curing equipment and technique and could be scaled from producing flat composite panels, as in this work, to full-scale complex structures (e.g., cylinders/barrels, wing skins, etc.). Using guided waves throughout the entire cure cycle (i.e., in situ measurements during liquid, rubbery, and glassy states) pushes the technology envelope forward as guided waves are typically applied to solid medium. It should be noted that the definition of guided waves traditionally used for solid media is broadened to encompass the liquid and rubbery states of the resin during cure. Under this context, the so called guided waves are defined by the measure of voltages from the sensors a distance from the actuator. In this complex liquid state, the fibers interact with the pressure waves excited by the piezoelectric transducer. To the authors' knowledge, this is the first attempt of measuring guided waves in composites (or any other material system) as they are propagating in the panel during cure.

## Acknowledgment

The authors would like to acknowledge the financial support from a Graduate Research Assistantship at National Institute of Aerospace (NIA) through the Advanced Composites Project at NASA Langley Research Center. The system is associated with a nonprovisional application filed at U.S. Patent and Trademark Office in September 2017 [25].

## Funding Data

• Langley Research Center (Grant Nos. AERX22017D and NNL09AA00A).

## References

References
1.
Lionetto
,
F.
, and
Maffezzoli
,
A.
,
2013
, “
Monitoring the Cure State of Thermosetting Resins by Ultrasound
,”
Materials
,
6
(
9
), pp.
3783
3804
.
2.
Nixdorf
,
K.
, and
Busse
,
G.
,
2001
, “
The Dielectric Properties of Glass-Fibre-Reinforced Epoxy Resin During Polymerisation
,”
Compos. Sci. Technol.
,
61
(
6
), pp.
889
894
.
3.
Hardis
,
R.
,
Jessop
,
J. L.
,
Peters
,
F. E.
, and
Kessler
,
M. R.
,
2013
, “
Cure Kinetics Characterization and Monitoring of an Epoxy Resin Using DSC, Raman Spectroscopy, and DEA
,”
Composites, Part A
,
49
, pp.
100
108
.
4.
McIlhagger
,
A.
,
Brown
,
D.
, and
Hill
,
B.
,
2000
, “
The Development of a Dielectric System for the On-Line Cure Monitoring of the Resin Transfer Moulding Process
,”
Composites, Part A
,
31
(
12
), pp.
1373
1381
.
5.
Lindrose
,
A. M.
,
1978
, “
Ultrasonic Wave and Moduli Changes in a Curing Epoxy Resin
,”
Exp. Mech.
,
18
(
6
), pp.
227
232
.
6.
Speake
,
J.
,
Arridge
,
R.
, and
Curtis
,
G.
,
1974
, “
Measurement of the Cure of Resins by Ultrasonic Techniques
,”
J. Phys. D: Appl. Phys.
,
7
(
3
), pp.
412
424
.
7.
,
R.
, and
Cawley
,
P.
,
1988
, “
A Review of Defect Types and Nondestructive Testing Techniques for Composites and Bonded Joints
,”
NDT Int.
,
21
(
4
), pp.
208
222
.
8.
Chen
,
J.
,
Hoa
,
S.
,
Jen
,
C.
, and
Wang
,
H.
,
1999
, “
Fiber-Optic and Ultrasonic Measurements for In-Situ Cure Monitoring of Graphite/Epoxy Composites
,”
J. Compos. Mater.
,
33
(
20
), pp.
1860
1881
.
9.
Maffezzoli
,
A.
,
Quarta
,
E.
,
Luprano
,
V.
,
Montagna
,
G.
, and
Nicolais
,
L.
,
1999
, “
Cure Monitoring of Epoxy Matrices for Composites by Ultrasonic Wave Propagation
,”
J. Appl. Polym. Sci.
,
73
(
10
), pp.
1969
1977
.
10.
Birt
,
E.
, and
Smith
,
R.
,
2004
, “
A Review of NDE Methods for Porosity Measurement in Fibre-Reinforced Polymer Composites
,”
Insight-Non-Destr. Test. Cond. Monit.
,
46
(
11
), pp.
681
686
.
11.
Stone
,
D.
, and
Clarke
,
B.
,
1975
, “
Ultrasonic Attenuation as a Measure of Void Content in Carbon-Fibre Reinforced Plastics
,”
Non-Destr. Test.
,
8
(
3
), pp.
137
145
.
12.
Jeong
,
H.
, and
Hsu
,
D.
,
1995
, “
Experimental Analysis of Porosity-Induced Ultrasonic Attenuation and Velocity Change in Carbon Composites
,”
Ultrasonics
,
33
(
3
), pp.
195
203
.
13.
Ghodhbani
,
N.
,
Maréchal
,
P.
, and
Duflo
,
H.
,
2016
, “
Ultrasound Monitoring of the Cure Kinetics of an Epoxy Resin: Identification, Frequency and Temperature Dependence
,”
Polym. Test.
,
56
, pp.
156
166
.
14.
Pavlopoulou
,
S.
,
Soutis
,
C.
, and
Staszewski
,
W.
,
2012
, “
Cure Monitoring Through Time-Frequency Analysis of Guided Ultrasonic Waves
,”
Plast., Rubber Compos.
,
41
(
4–5
), pp. 180–186.
15.
Hudson
,
T. B.
,
Grimsley
,
B. W.
, and
Yuan
,
F. G.
,
2016
, “
Development of a Fully Automated Guided Wave System for In-Process Cure Monitoring of CFRP Composite Laminates
,”
American Society of Composites Thirty-First Technical Conference
, Williamsburg, VA, Sept. 19–22, Paper No. 2920.
16.
Hudson
,
T. B.
,
Hou
,
T. H.
,
Grimsley
,
B. W.
, and
Yuan
,
F. G.
,
2015
, “
Detection of CFRP Composite Manufacturing Defects Using a Guided Wave Approach
,”
Society for the Advancement of Materials and Process Engineering Technical Conference
, Baltimore, MD, May 18–21, pp.
2096
2111
.
17.
Hudson
,
T. B.
,
Grimsley
,
B. W.
, and
Yuan
,
F. G.
,
2017
, “
Detection and Quantification of Composite Manufacturing Defects Using Guided Waves
,”
Struct. Health Monit.
(submitted).
18.
Farquharson
,
S.
,
Carignan
,
J.
,
Khitrov
,
V.
,
,
A.
, and
Shaw
,
M.
,
2004
, “
Development of a Phase Diagram to Control Composite Manufacturing Using Raman Spectroscopy
,”
Proc. SPIE
,
5272
, pp.
19
29
.
19.
Gillham
,
J. K.
,
1987
, “
Formation and Properties of Thermosetting and High Tg Polymeric Materials
,”
Makromol. Chem., Macromol. Symp.
,
7
(
1
), pp.
67
74
.
20.
Aronhime
,
M. T.
, and
Gillham
,
J. K.
,
1986
, “
Time-Temperature-Transformation (TTT) Cure Diagram of Thermosetting Polymeric Systems
,”
Epoxy Resins and Composites III
,
Springer
, Berlin, pp.
83
113
.
21.
Ee
,
D. V.
, and
Poursartip
,
A.
,
2009
, “
HexPly 8552 Material Properties Database for Use With COMPRO CCA and RAVEN
,” National Institute for Aviation Research, Wichita, KS, accessed Oct. 12, 2016, http://www.niar.wichita.edu/coe/ncamp_documents/Hexcel%208552/Hexcel8552MaterialCharacterizationBinderV0-9.pdf
22.
Ersoy
,
N.
,
Potter
,
K.
,
Wisnom
,
M. R.
, and
Clegg
,
M. J.
,
2005
, “
An Experimental Method to Study the Frictional Processes During Composites Manufacturing
,”
Composites, Part A
,
36
(
11
), pp.
1536
1544
.
23.
Ersoy
,
N.
,
Potter
,
K.
,
Wisnom
,
M. R.
, and
Clegg
,
M. J.
,
2005
, “
Development of Spring-In Angle During Cure of a Thermosetting Composite
,”
Composites, Part A
,
36
(
12
), pp.
1700
1706
.
24.
Ersoy
,
N.
,
Garstka
,
T.
,
Potter
,
K.
,
Wisnom
,
M. R.
,
Porter
,
D.
,
Clegg
,
M.
, and
Stringer, G.
,
2010
, “
Development of the Properties of a Carbon Fibre Reinforced Thermosetting Composite Through Cure
,”
Composites, Part A
,
41
(
3
), pp.
401
409
.
25.
Hudson
,
T. B.
,
Yuan
,
F. G.
, and
Grimsley
,
B. W.
,
2017
, “
Automated Guided Wave System for In-Process Cure Monitoring of CFRP Composite Laminates
,” Nonprovisional Patent Applications No. 15/710,211 filed on Sept. 20, 2017.