DETECTION OF MOISTURE IN COMPOSITES USING AN ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY SENSOR
G.D. Davis, L.A. Krebs, and
C.M.
Dacres
10260
Old Columbia Road
Columbia,
MD 21046
back to DACCO SCI News and Highlights
dacres@daccosci.com
(Chester Dacres)
davis@daccosci.com
(Guy Davis)
Abstract
An in-situ corrosion sensor was adapted to detect moisture in composites. The sensor, based on electrochemical impedance spectroscopy (EIS), is sensitive to moisture in the range of 0-1% for graphite/epoxy composites and up to 3% for glass/polyimide composites. It detects moisture in graphite/epoxy honeycomb structures and has the potential to be applicable to other bonded structures as well. Coupled with a commercial portable computer, the hand-held sensor will be able to detect moisture in composite components of the F/A-18, F-14, EA-6B, and V-22 aircraft, among others. Such moisture detection capability is critical for repair operations, since assurance of dry material prior to the elevated temperatures used during curing of a repair will prevent the formation of steam and subsequent voids, delaminations, and disbonds. It can also warn of possible moisture-induced degradation resulting from hydration of aluminum bonding surfaces or corrosion of core material.
Introduction
Concerns over absorbed moisture and its removal from composite materials during structural repair are prevalent in several industries.[1]-[3] In addition to the “expected” types of structural damage often caused by exposure to moisture, the presence of undetected moisture absorbed in composites has been repeatedly shown to cause more damage during repair procedures than was present prior to the repair attempt. Specific moisture-induced effects include:[4]
· Bondline voids resulting from moisture in the laminate skins migrating to the bondline during the curing process
· Blistering of laminate plies if the vapor pressure of the moisture exceeds the strength of the polymeric matrix
· Skin-to-core disbonds (blown skins) or honeycomb core node failures if the vapor pressure in the core exceeds the strength of these bonds
· Plasticization of the adhesive or polymeric matrix
· Hydration of aluminum bonding surfaces
· Corrosion of thin honeycomb material.
The third failure mode listed above becomes more likely if either of the last two phenomena are present. These occur after extended exposure of the adhesive or honeycomb cells to moisture.[5]-[7][8] Such hydration can reduce interfacial strengths by up to 90%[9] and can induce structural failure under flight loadings and conditions. Moisture induced failures have been reported on the horizontal stabilizer, rudder, and trailing edge flap of the F/A-18[10],[11] and on the radomes of EA-6B and F-14.
Figure 1. Left: As-received starboard rudder of F/A-18A which failed in-flight. Right: Ultrasonic C-scan of F/A-18 rudder after detection of water in the mid-hinge region. Also shown are Porta Pull values showing moisture-induced reduction in strengths.
|
In at least two cases, rudders on the F/A-18 have failed in-flight with loss of 80-90% of one of the rudders (Figure 1, left). Failure analysis of the remaining section indicated extensive honeycomb corrosion near the mid-hinge point due to moisture ingress via the grounding terminal. Fillet bond failure between the core and the graphite/epoxy skin leading to skin delamination was also noted in the lower recovered section.11
In another example, x-rays detected water in an F/A-18 rudder. The engineering investigation showed aluminum corrosion products in the drained fluid. Ultrasonic C-scans showed a large area of attenuation around the mid-hinge point (Figure 1, right). Porta Pull plugs from that area showed very low strengths (48-143 psi) with higher strengths in the surrounding areas (286-674 psi). All plugs showed varying degrees of adhesive failure mode.
Moisture in the composite, in the adhesive between the composite and the core, or in the core itself can cause significant damage during repair when the moisture turns to steam and internal pressure can exceed bond strengths. The repair failures are more pronounced with the use of 350°F-curing adhesives, such as FM-300, and have led to change orders authorizing the use of 250°F-curing adhesives during Depot level composite repairs.[12] Figure 2 shows an example of such repair-induced damage. During the curing of the repair patch on the left, additional delamination, marked by the black line surrounding the patch, occurred as residual moisture turned into steam.
Figure 2. F/A-18 trailing edge flap following repair showing additional moisture-induced delamination during the repair operation.
To compensate for the lack of real-time data, the drying schedules for parts requiring repair have been set to prevent entrapped moisture especially in large parts. Drying temperature and time are based primarily on historical experience and success for each material system and component configuration, with a typical schedule of several days at 160-180°F (70-82°C) at ambient pressure. Although this procedure presents difficulties in keeping repair costs down and reducing aircraft down time, little advancement has been made on the practical side of assuring adequate drying. The only means currently available to inspect composite moisture content is to monitor the humidity level of the drying air, much like the humidistat in a household clothes drier. Technologies that have been tried include thermography, and x-ray and neutron radiography. Thermography appears useful in detecting but not quantifying trapped moisture in honeycomb materials. X-ray and neutron radiography are both quite sensitive in detecting trapped moisture, but their safety restrictions and costs have been the major deterrent to date. Portability, a previous disadvantage to these radioscopic techniques, appears to be overcome by several commercial entities.
In order to address the problem of determining moisture content in composite materials in a reliable, cost-efficient, and portable way, DACCO SCI, INC., (DSI) has adapted its in-situ corrosion sensor to monitor absorbed moisture. The original in-situ corrosion monitor, [13]-[15][16] used to study the degradation of coated metals, is based on the well-accepted laboratory technique of electrochemical impedance spectroscopy (EIS). This sensor has proven to be very sensitive to the presence of moisture in paints, adhesives, and other polymers. More recent results have shown that it is also sensitive to moisture levels ranging from 0.3% to 3.0 % in selected composite materials of interest to the Navy and Air Force.
Experimental
EIS is an established laboratory technique for the investigation of coating deterioration and substrate corrosion.[17]-[21] Very good correlation has been reported between short-term EIS data (low frequency impedance and break-point frequency) and long-term coating performance in sea water immersion, demonstrating the technique's predictive capabilities.
Figure 3. Equivalent circuit model of a coated metal. RW represents the solution resistance. For a coated metal specimen Rpo and Cc correspond to the resistance and capacitance of the coating while Rt and Cdl correspond to the resistance and capacitance of the metal interface.
|
Initially, the coating resistance Rpo is very high so that the system acts as a capacitor with a log impedance-frequency slope of -1 (Figure 4, left). As the coating absorbs moisture and substrate corrosion occurs, Rpo decreases and the lower circuit branch becomes important at low frequencies. Under these conditions, the impedance becomes independent of frequency. It is possible to also model composite materials using equivalent circuit modeling (see, e.g., Figure 8 and Figure 12), although the circuit, and the physical meaning of the components in the circuit, may differ from the coated metal example.
A plot of the low-frequency impedance versus exposure time can be interpreted to determine the stage of degradation of a coating or an adhesive joint as illustrated in Figure 4, right. In this experiment, an epoxy-coated aluminum specimen was immersed in hot water.[22] The behavior of the data is very similar to that of other painted metals. There are definite degradation stages corresponding to water uptake by the adhesive, incubation of hydration, and intense hydration of the adherend. For this program, we concentrated on the first stage – moisture absorption.
|
Traditionally, the EIS technique requires a specimen to be immersed into an electrolyte along with counter and reference electrodes. Such procedures are not suitable for detecting corrosion on structures in the field. The sensor developed by DACCO SCI enables the EIS measurements to be obtained in situ in the field.13-16 It allows paint deterioration or substrate degradation to be detected in its early stages with identical measurements to those obtained with conventional EIS. A similar approach is suitable for detecting moisture in composites. The sensor electrode acts as both the reference and counter electrodes with the substrate or, in the case of composite inspection, a second sensor electrode being the working electrode. The complex impedance across the paint, adhesive film, or composite material is measured as a function of frequency.
Approach
Several glass/polyimide and graphite/epoxy materials, provided by the Navy and the Air Force, were examined, and are listed in Table 1. These materials are relevant to the F/A-18C/D, F/A-18E/F, V-22, and Harrier aircraft and H46 and H53 helicopters, among others. The procedure for inspection was designed to meet the following criteria:
· provide accurate reproducible data relevant to the moisture content within a material,
· suitable for use under both laboratory and field service conditions.
Specimens were dried at 100-120°C (212-250°F) until a constant weight was achieved. This weight was defined to be the “dry” weight, or 0.00% moisture content (MC). Impedance spectra were collected from each dry specimen with the sensor electrodes in various orientations, including electrodes on opposite sides of a specimen, on the same side and along a fiber direction, and same-sided but diagonal to the fiber directions. Additional orientations for the honeycomb composites included placing one electrode on the skin surface while using the other electrode to make electrical contact to the core, or to any exposed adhesive backing. All specimens except the honeycomb composites had their edges masked with tape in order to minimize moisture absorption through the edges. Specimens were then exposed to a constant relative humidity of 98% at 50°C (122°F), and removed periodically to be weighed and have impedance spectra collected. This procedure was performed for MC values ranging from 0.1% to 3.0%.
Two versions of the DSI corrosion sensor have been developed: a hand-held probe and an electrode permanently attached to the structure. The moisture sensor being developed in this program is based on the use of pairs of hand-held probes. The location of the probe electrodes determined what portion of the composite material was being inspected. Electrodes on the same side of the composite probed only a surface region of undetermined depth. Electrodes on opposite sides of the composite probed the entire thickness. Signals collected using electrodes located on the front composite surface and on the honeycomb core contained information from the composite plus adhesive (and core, in the case of Nomex).
Table 1. List of Phase I specimens.
|
At the time of each measurement, sensor electrodes were pressed against the specimen surface using an ultrasonic couplant, Exosen, to assure good conductivity between the electrode and the specimen. After each measurement, the specimen surfaces were rinsed with deionized water before being returned to high humidity. The exceptions to this procedure were the honeycomb specimen measurements involving contact to the adhesive backing or the core material. Contacts in these cases were made without Exosen. Previous studies have shown that the Exosen is useful in assuring reliable electrical contact, but is not necessary.
Results and Discussion
Glass
Polyimide
Composites
EIS spectra collected from the glass polyimide specimens are shown in Figure 5. The spectra showed a mainly capacitive impedance behavior although a resistive low-frequency region is seen following moisture adsorption. The same-side measurement on the S2/PMR-15 specimen (Figure 5 (a)) clearly had a lower overall impedance than the opposite-side measurement (Figure 5 (b)), as well as both electrode arrangements for the S2/AFR700 material. This was attributed to the conductive coating on the S2/PMR-15 specimen, which provided a low impedance path for current to flow between electrodes on the same side. Electrodes placed on opposite sides measure through the thickness of the specimen, and were not affected by the conductive coating. The S2/AFR700 was not coated.
Figure 5. Selected EIS spectra for S2/PMR-15 (a) same-sided measurements, (b) opposite sided measurements, and S2/AFR700 (c) same-sided measurements, (d) opposite sided measurements. Notice that the S2/PMR-15 same-sided measurements indicate a significantly lower impedance than the other measurements pictured, due to the presence of a conductive coating.
|
The low frequency impedance versus MC relationships are given in Figure 6. Figure 6(a) shows the S2/ AFR700 data for both sensor orientations. Two points can be immediately derived from these spectra:
· Sensors oriented to make measurements from the same surface (same-sided) or through the material thickness (opposite-sided) give equivalent impedance results,
· A clear logarithmic decrease in impedance is seen for MC levels in the range of 0.6% to greater than 2.0%.
Figure 6 (b) and (c) show the spectra for the coated S2/PMR-15 specimen. The impedance behavior of the opposite-sided measurement (Figure 6 (b)) is similar to that of S2/AFR700. However, the same-sided measurement (Figure 6 (c)), which is plotted on a linear scale, is dominated by the conductive surface coating. With this coating, the range of greatest sensitivity appears to be below 1.0% MC, although without the coating it is reasonable to assume that the behavior would more closely resemble that shown in Figure 6 (b).
Selected EIS spectra for one of the Navy graphite-epoxy materials are shown in Figure 7. Same-side and opposite-side measurements gave similar results as did measurements with the electrodes oriented along the outer fiber direction and at 45° to this direction. The overall impedance values for these composites are generally lower then the impedance values for the glass/polyimide composites just discussed. Also, the change in impedance as a function of MC is less pronounced in the graphite-epoxy composites as compared with the glass-polyimide specimens. We attribute both of these effects to the presence of the conductive graphite fibers in the composite structure, which serve to both lower the overall impedance and to affect the effects of moisture absorption. In most cases, the impedance behavior remained capacitive over the entire frequency range and did not have a low-frequency resistive component. Nonetheless, spectra information could be correlated with MC levels. To better understand and track the changes these composites underwent as a result of absorbing moisture, equivalent circuit modeling was used. This approach uses the data collected at all frequencies but is dependent on the choice of circuit model.
Figure 6. Low Frequency impedance versus MC for the glass polyimide specimens; (a) S2/AFR700, both sensor orientations, (b) S2/PMR-15, opposite-side, and (c) S2/PMR-15, same-sides (plotted on a linear scale).
|
Equivalent circuit analysis focused on identifying circuits that were physically significant and evaluating the impedance data collected, with the intention of identifying circuit elements particularly sensitive to changes in the quantity of moisture absorbed. Although data were collected on all available materials, data from one monolithic (IM7/8552, 7 ply) and one honeycomb (AS4/3501-6, FM300, Al-core) were extensively studied for the purpose of developing the model fitting procedure. The method was based on tracking the changing values of several moisture-sensitive circuit elements as a function of moisture content. Identifying and monitoring the trends of more than one circuit element increases confidence in this method of moisture content prediction. Representative spectra shown in the Bode Magnitude format are given in Figure 7 for the monolithic sample. During analysis, the spectra were also viewed in the Bode Phase Angle and Nyquist formats, in order to enhance understanding of the effects of increasing moisture content.
Figure 7. EIS spectra of the IM7/8552, 7 ply specimen following different periods of humidity exposure. The moisture uptake (in percent) is given in the legend. Plots are shown in Bode magnitude (left) and phase (right) formats.
|
A danger of this type of modeling lies in the fact that there are no unique mathematical solutions to fit spectra such as those of Figure 7. It is also true that as more parameters are added to any model, it becomes easier to fit the data mathematically. Therefore it is necessary to design an equivalent circuit model that has some relationship to the physical reality of the material under investigation, using as few parameters as possible. The model selected for the monolithic samples is shown in Figure 8. The model consists of three parallel R/CPE combinations in series with a “solution” resistance. This circuit was designed to describe the expected behavior of same-side contact measurements. The first and third R/CPE loops were constrained to be equal. Each was intended to describe the signal response of the resin-rich layer at the composite surface. The middle loop, R2/CPE2, was intended to model the behavior of the composite interior, where the signal is governed with the graphite fiber system. Although exact material values such as resistivity and dielectric constant were not known for the composite material, reasonable estimates were applied during model fitting in order to further constrain the parameters in favor of a physically realistic result.
Equivalent circuit modeling of the data shown in
Figure
7 using the circuit model shown in
Figure
8 revealed several parameters with sensitivity to
moisture content. Examples of
correlations between moisture content and particular circuit elements modeled
for the 7-ply IM7/8552 sample spectra are shown in
Figure
9. The
parameters shown in Figure
9 vary with moisture in a regular
way. By fitting models to data
in multiple formats and by tracking several moisture-sensitive parameters,
it becomes possible to predict values of moisture content for these types
of materials more reliably than by tracking only one
parameter.
Figure 8. Equivalent Circuit Modes for monolithic composite. The circuit should be able to model similarly constructed materials by varying the values of the individual circuit elements.
Figure 9. Correlation of moisture-sensitive circuit elements with moisture content in the 7-ply IM7/8552 monolithic composite sample.
 |
 |
 |
Graphite-Epoxy Honeycomb Composites
The honeycomb composites presented the most challenging measurement situations of the materials studied. Several arrangements of electrode placement were devised in order to test as many of the component materials as possible. These electrode arrangements included placement on opposite sides (on the surfaces of the two graphite-epoxy sheets sandwiching the core material) and on the same side, as well as placing one electrode on the graphite-epoxy surface and making contact to the core material or to the adhesive backing with the other electrode lead.
The most straight-forward results were collected from the IM7/8552 composite with the Nomex core, locating the sensor electrodes at the graphite/epoxy surface and at the core. Figure 10 shows the low frequency impedance values versus MC. The useful range of sensitivity in this case is roughly 0.15 % to 0.50%, in which the impedance steadily decreases as MC increases. It is believed that the Nomex core absorbed moisture more readily than the graphite-epoxy composite and controlled the impedance measurements.
|
Figure 10 Impedance versus MC for honeycomb composite with IM7/8552 skin and Nomex core. Electrodes located on skin and on core.
|
For other specimens, equivalent-circuit analysis was needed, as reflected in the capacitive behavior of Figure 11. The aluminum core honeycomb composite model of Figure 12 consists of two loops in series and a third imbedded loop. The R1/CPE1 parallel combination is a collapsed version of the monolithic composite circuit of Figure 8, and is included to account for moisture absorption in the skin of the honeycomb. The R2/C2 parallel combination with the imbedded R3/C3 parallel loop is based on a standard circuit model often used for the analysis of coated metals. In this case, R2/C2 represents moisture absorption in the adhesive joining the core and the skin. The R3/C3 portion represents activity at the adhesive/core interface. A final resistor is again placed in series with the circuit to represent the “solution” resistance.
Figure 13 shows some moisture–sensitive parameters from equivalent circuit modeling of the AS4/3501-6 honeycomb. Again, data were fit in multiple formats, and several parameters were tracked. The situation of the honeycomb composite is complicated by the variety of materials present in the system, and the differing rates at which they absorb moisture. Ideally, the model parameters in this case will reveal not only how much moisture is in the system, but also in which materials it has accumulated (i.e. graphite/epoxy composite, adhesive). Continued evaluation of this complex combination of materials is necessary to develop the moisture sensor system as a useful tool. Continued analyses of the skin materials will aid in the understanding of these more complicated structures.
Figure 11. EIS spectra of the AS4/3501-6, FM300, Al-core honeycomb specimen following different periods of humidity exposure. The moisture uptake (in percent) is given in the legend. Plots are shown in Bode magnitude (left) and phase (right) formats.
|
|
Figure 12. Equivalent Circuit Models the AS4/3501-6, FM300, Al core honeycomb composite. The circuit should be able to model similarly constructed materials by varying the values of the individual circuit elements.
Figure 13. Correlation of moisture-sensitive circuit elements with moisture content in the AS4/3501-6 honeycomb composite sample.
 |
 |
Summary and Conclusions
Equivalent circuit modeling is a viable way to measure and calibrate the amount of moisture present in graphite-epoxy, and in more complicated honeycomb composite materials. By concentrating first on a few selected materials, the circuit model designs and the fitting procedure can be developed and refined for reliable prediction of moisture content. Examination of impedance data in multiple presentation formats together with concurrent tracking of several moisture-sensitive components in the circuit model help to increase confidence in the prediction.
Acknowledgements
The authors wish to acknowledge the funding provided by the US Navy (NAVAIR) under contracts N00421-97-C-1371 and N00421-99-C-1318 for research into the composite moisture-sensor probe concept. The US Air Force (AFOSR) provided the funding for the research and development of the original in-situ corrosion sensor, from which the moisture sensor has been adapted. The authors wish to thank Craig Bevans for his efforts and input throughout this project. They would also like to thank Paul Mehrkam and Doug Perl for the photographs showing the detrimental effects of moisture in composite and honeycomb structures.
References
[1] J.M. Augl, G.T. Sivy, “Reduction of Moisture Effects During the Cure of Epoxy Adhesives uses in Composite Repair,” in NASA Langely Research Center Welding, Bonding, and Fastening, p.439 (1984)