Printer Friendly

Correlation between acoustic emission signals and delaminations in carbon fiber-reinforced polymer-matrix composites: A new look at mode I fracture test data.

Abstract

Recently, a rough correlation between cumulative AE signal amplitude and averaged crack area has been derived from in-situ tensile tests on miniature samples of spruce wood combining AE monitoring and simultaneous synchrotron-based X-ray computed micro-tomography. This indicated that, at least for the chosen experimental set-up and equipment, AE tended to be more sensitive than X-ray imaging in detecting formation and propagation of small cracks. An analogous analysis is now presented for quasi-static mode I (tensile opening) delamination tests performed on carbon-fiber poly-ether-ether-ketone composite laminates with simultaneous insitu AE monitoring and X-ray radiography using an imaging system with video recording. The data are further used for estimating the variation in the correlation factor for a given experimental set-up by comparing different approaches for deriving an estimated value of the correlation factor.

Keywords: Carbon fiber-reinforced polymer-matrix composites, mode I tensile-opening fracture, measurement sensitivity, acoustic emission intensity correlation with damage size (crack area or diameter)

1. Introduction

Acoustic Emission (AE) monitoring is a non-destructive method with high sensitivity to microscopic changes caused by, e.g., rapid release of energy from localized sources in materials and structures which makes it different from many other non-destructive test methods (see, e.g., [1,2]). While AE is successfully applied in a wide range of structural health or integrity monitoring using empirical criteria for the evaluation of the AE signals (see, e.g., [3,4]), it has proven more difficult to extract quantitative information on the sources of the AE signals, even though it was recognized a long time ago that the type of source was affecting the recorded AE signals [5].

Recent advances in unsupervised pattern recognition have shown that discrimination between microscopic damage mechanisms in carbon-fiber reinforced polymer-matrix (CFRP) laminates is feasible [6]. The identification of the source mechanism (type and orientation) was achieved by detailed simulations of the whole signal measurement chain from source to signal recording. The main parts of this simulation comprised model sources simulated as dipole or multipoles in different orientations and locations with a range of signal rise times, signal propagation in the material with frequency dependent attenuation and anisotropic wave speeds as well as the frequency-dependent transfer functions of the AE sensors and the signal recording. For CFRP, this procedure identified the source mechanisms of the different clusters resulting from pattern recognition as matrix cracks, fiber-matrix debonding or fiber breaks [7]. However, identification of the type of a localized model AE source does not directly yield quantitative information on the size of the damage created by the respective mechanism, e.g., the area or length of a micro-crack in the matrix polymer.

Combining AE monitoring of load tests on material specimens or structural elements with suitable, complementary non-destructive test methods [8], in principle, can provide an approach for estimating damage size (e.g., crack area or length) for a correlation with the recorded AE signal parameters or their respective source properties. This has been demonstrated by combining in-situ synchrotron X-ray computed micro-tomography with AE monitoring of tensile tests on miniature spruce wood specimens [9]. After detection of significant AE, the tensile loading was stopped and the applied load held constant for performing the X-ray tomography measurement. Consecutive load steps yielded indications of progressive damage, e.g., as growing cracks. Pattern recognition as outlined in [7] essentially yielded two clusters of AE signals for the wood specimens [10,11] that could be associated with slower and faster source mechanisms, i.e., with distinctly different source rise times in the simulations [12]. Tentatively, these were identified as inter-wall cracks between wood cells for the low and with cell wall cracks for the high frequency cluster. Analysing the X-ray tomography image slices and identifying the crack locations, a rough estimate of the crack length increment from one tensile load step to the next could be obtained. This yielded a quantitative, but rough correlation between crack area and AE signal amplitude (measured in mV). Of course, such an estimation depends on the specific set-up (e.g., specimen material and size/shape) and equipment sensitivity (depending mainly on sensor mounting, AE sensor type and data acquisition settings), but holding these constant allows for relative comparison among nominally identical specimens. Comparing the correlation factor indicated that with the chosen signal threshold setting, AE tended to be more sensitive to crack formation than the X-ray computed micro-tomography. This approach is now tried on delamination propagation in CFRP fracture toughness test specimens for an attempt at estimating the micro-crack size due to a single AE signal of given amplitude.

2. Materials and Testing Details

2.1 Carbon-fiber reinforced epoxy and delamination growth

CFRP composites are increasingly used in many engineering applications, including loadbearing elements. However, their susceptibility to impact damage and subsequent interlaminar delamination growth, or more general, their rather weak strength in the through-thickness direction, have spurred intensive research into the characterization of interlaminar delamination growth under quasi-static and fatigue loads [13]. There are standardized fracture test methods for the so-called mode I (tensile opening) [14] and mode II (in-plane shear) [15] interlaminar fracture toughness or delamination resistance, as well as mixed mode I/II [16], all under constant displacement loading. Analogous test methods for cyclic fatigue delamination propagation are currently under development [17]. The evaluation of the quasistatic mode I fracture toughness according to the standard is using so-called double cantilever beam (DCB) specimens and based on measuring the delamination length increments due to specific applied loads at given crosshead displacements. The measurement of the delamination length increments during the load tests is made by visual observation using a travelling microscope following the propagation of the delamination tip along the edge of the standard test specimen. In research, methods are developed to replace this either by digital image recording and analysis or by back-calculating the delamination length from the change in specimen compliance due to the propagation of the delamination. However, with a required resolution of the crack length around 0.05 mm [14] all these approaches effectively average over many microscopic crack events, both on the length, as well as on the time scale.

The specimens used in the present research were provided by NASA Langley for a round robin performed at several laboratories in Europe, the U.S. and Japan. The details of material preparation and the full range of fracture toughness results are summarized in [18,19]. In the study reported here, only CFRP laminates made from AS4 type carbon fiber and thermoplastic poly-ether-ether-ketone (PEEK) matrix are analysed. The starter crack film insert was a polyimide film. Two different film thicknesses (7 and 13 [micro]m, respectively) were used in the four DCB specimens with an average thickness between about 3.08 and 3.15 mm, an average width between 20.05 and 20.09 mm and a total length of 139.5 mm.

2.2 In-situ AE monitoring and X-ray radiography of interlaminar fracture tests

As part of the development of the mode I fracture toughness standard [14] in the framework of the round robin noted above [18,19] delamination tests on the AS4/PEEK CFRP specimens were performed with an equipment that allowed for in-situ X-ray radiography and image recording on video (described in [20]) simultaneously with AE monitoring [21]. For the tests reported here, the loading equipment was set to a displacement speed around 1 mm/min. Load and displacement values were recorded with a separate data acquisition system (not as part of the AE data via external parameter input).

The X-ray radiography used a microfocus X-ray source (focal spot size 5-10 [micro]m at voltages below 40 kV) in projection imaging mode with an imaging system that allowed for video recording of the images (Fig. 1) at a sampling rate (image frequency) of 25 Hz (both from Feinfocus, Germany). X-ray source voltage and current were set constant at 32 kV and 0.116 mA, respectively, for all four tests. Depending on the distance between source and test object compared to that between source and imaging system, image magnifications between a factor of about 2 and 5 are feasible with this set-up. For the tests reported here, the magnification factor was about 2.6. Methylene-di-iodide was used as contrast agent for the X-ray imaging of the delamination growth (Fig. 2) and injected inside the starter crack with a syringe after slightly opening it (displacement of roughly 3 mm).

[FIGURE 1 OMITTED]

AE signal parameters from the quasi-static mode I fracture test on the CFRP laminates were recorded with AE equipment type 3000/SPARTAN with one sensor type R-15 (both from Physical Acoustics Corp.) mounted near the bottom end of the specimen (about 1.5 to 2.0 cm from the edge). Data acquisition settings were a preamplifier gain of 40 dB, a threshold 30 [dB.sub.AE] , a hit definition time of 500 [micro]s, and a rearm time of 1000 [micro]s. The AE analysis will use plots of AE activity, i.e., the number of AE signals recorded per unit time, and AE intensity, i.e., maximum AE signal amplitude distributions, each presented as a function of time (Fig. 3). Video recording of the X-ray images and AE data acquisition were started simultaneously, within roughly 1 second, in order to synchronize the data.

3. Estimating the Sensitivity of AE Monitoring for Microcrack Formation in AS4/PEEK Specimens Loaded under Mode I Tensile Crack Opening Load

3.1 Approach for the sensitivity analysis

A detailed study [21] of the AE signals recorded during mode I delamination growth of CFRP DCB specimens has shown that by defining empirical delamination initiation criteria from the observed AE activity and intensity, the critical fracture toughness [G.sub.IC] obtained from these is comparable to that obtained from the analysis procedure outlined in the standard [14]. This confirmed that the AE analysis is capable of capturing the essential features of the mode I DCB delamination tests not only in a qualitative manner, but also yields quantitative results. In the previously published analysis of the sensitivity of AE to yield a measure of the crack length induced by tensile loading of miniature wood specimens [9], the AE signal amplitudes were correlated with the crack area measured from synchrotron-based X-ray computed microtomography. The approach presented here hence first estimates the total fracture surface generated during the test from the the X-ray projection images, and then the average AE signal amplitude and the number of AE signals recorded during the mode I delamination test. It is immediately clear that the correlation between AE signal amplitude and crack area obtained for this test may and likely will significantly differ from that obtained in the tests on the miniature wood specimens. Major effects causing differences are, e.g., the material type (CFRP versus wood) and its specific signal attenuation behaviour, the specimen size (30 mm versus about 140 mm) or more important, the distance between AE signal source and sensor, the different microscopic damage mechanisms (cell wall delamination or cracking versus polymer micro-crack formation) as well as the different AE equipment, sensor types (with different frequency transfer functions) and data acquisition settings used for the AE measurement.

3.2 Estimating the fracture surface area

Getting a first estimate of the fracture surface area created during the test is straight-forward. Determining the difference between initial and final delamination length, multiplied by the specimen width yields an approximation which can be considered as a lower bound. The initial delamination length is well defined and straight across the width of the specimen by the tip of the polymer insert film. However, the crack tip during delamination growth (Fig. 2) is not straight and shows small, stochastically distributed protrusions.

[FIGURE 2 OMITTED]

It has to be noted that the image with the contrast agent may not in all cases show the exact location of the tip of the delamination. First, the wettability of the contrast agent may be limited and may hence not penetrate into the microscopic cracks. Second, there is a minimum amount of contrast agent necessary to yield a detectable contrast difference in the radiography image. Looking at the grey-scale levels of the image pixels along the direction of propagation of the delamination yields (coming from the top of the image) a high level (dark) that is slightly decreasing with increasing length showing some scatter, then a roughly linear decrease and finally a roughly constant, but low value (bright), again with some scatter. From this, it is estimated that the total delamination length can be estimated to within a few mm. This will be further considered in the error estimate for the sensitivity correlation.

Fracture surfaces of CFRP laminates in general show some roughness or corrugation that could vary depending on the scale at which this is analysed. This is also true for the delamination fracture of the CFRP DCB specimens. Deducing roughness from fracture surface images, such as, e.g., SEM micrographs is fairly complex in spite of the apparent obvious corrugation or roughness that can be seen (see, e.g., [23] for details). Assuming some type of fractal surface for the CFRP laminate may be misleading [24]. Even though [24] deals with thermoset epoxy rather than thermoplastic PEEK, the fractal dimension seems questionable for estimating the additional crack or delamination area in the thickness direction of the DCB specimens. In the analysis, this will be considered by simply adding a certain, estimated percentage to the lower bound area and assessing the influence of that on the sensitivity.

One specimen showed unstable delamination growth yielding a tip outside the field of view of the imaging system. The other three DCB specimens yield a lower bound of the delamination area increase of 538, 459, and 432 [mm.sup.2], respectively, if the change in slope of the grey-scale image is taken as indication of the position of the crack tip. As discussed above, the tip may extend somewhat further and the total fracture surfaces may be larger than the planar rejection due to surface roughens and corrugation. In the analysis, the lower bounds and area values 20% and 40% larger will be used to estimate lower and upper bounds for the sensitivity.

3.3 Estimating the average AE amplitude

The AE activity and AE intensity (Fig. 3) typically show significant variation in time. The AE activity first increases roughly exponentially with time, reaches a maximum (close to 1'000 hits per 5 seconds) and then drops. There are three more peaks in AE activity. In the video recording of the delamination growth, these phases of high AE activity correlate with faster delamination growth. The total number of AE signals ("hits") recorded for the delamination growth period for the three DCB specimens analysed amount to about 35'620, 32'790, and 51'970. It has to be noted that these values are estimates that could vary by as much as 30%. The average amplitudes are estimated to be around 70 to 72 [dB.sub.AE] for all three specimens. From Fig. 3 (left) the upper and lower bound for the average AE signal amplitude are estimated as 80 and 70 [dB.sub.AE]. Converted to [micro]V, the average amplitudes of 70 and 72 [dB.sub.AE] correspond to about 3'160 and 3'980 [micro]V, and the upper and lower bounds are 7'080 and 2'240 [micro]V, respectively. The AE signals below the threshold (30 [dB.sub.AE]) are not recorded and hence do not contribute to both, total number of AE signals and average AE signal amplitude.

The majority of AE signals recorded lies in the amplitude range between about 60 and 80 [dB.sub.AE] (Fig. 3).

[FIGURE 3 OMITTED]

3.4 Estimating the average sensitivity of AE analysis

With the estimated lower and upper bounds of the delamination area created during the mode I fracture test in the DCB specimens, the average AE signal amplitude and the respective upper and lower bound for the average crack area created per average AE signal (with respect to signal amplitude) can be estimated. The results are compiled in Table 1. Of course, the assumptions made in the determination of the different quantities are essential, and the result could differ significantly, if these assumptions were proven to be questionable. One important aspect is AE signal attenuation, the AE signal amplitudes were used as recorded and no attempt was made at correcting for attenuation. CFRP composites can cause significant attenuation and the recorded AE signals are possibly affected by reflections from the specimen surfaces, since the width and thickness dimensions (nominally 20 and 3 mm) are relatively small.

Using the lower bound of the delamination area for the results presented in Table 1 yields an estimated average upper bound of the sensitivity from three specimens of 0.00029 mV/[micro][m.sup.2]. The individual values for the three DCB specimens are within the range between about 0.0002 and 0.0004 mV/[micro][m.sup.2] and hence fairly consistent, considering the rough approximations that were used for the estimates. Assuming fracture areas larger than the lower bound by 20% and 40% respectively, will yield lower average sensitivities of about 0.00024 and 0.00021 (again in units of mV/[micro][m.sup.2]). Using an upper bound for the average AE signal amplitude of 77 [dB.sub.AE] or 7'080 [micro]V with the lower bound for the fracture area yields about 0.00061 mV/[micro][m.sup.2], and the same average amplitude with an area that is higher by 40% (as an estimated upper bound) yields 0.00037 mV/[micro][m.sup.2]. If the lower bound area is increased by a factor of two. i.e., 100%, the estimated average sensitivity is 0.0003 mV/[micro][m.sup.2]. This can be compared with the value of 0.0038 mV/[micro][m.sup.2] obtained in [9] for testing the miniature tensile specimens made from spruce wood. The higher sensitivity in the latter case can have several causes as discussed above.

4. Summary and Conclusions

Combining AE measurements on mode I DCB fracture tests with CFRP specimens with insitu projection radiography and highlighting the delamination area with a suitable contrast agent, yields a rough estimate of the sensitivity of the AE measurement with respect to microcrack area that can be detected. From the average sensitivity determined for AE monitoring of the CFRP DCB fracture specimens, it can be estimated that the area of a micro-crack that yields an AE signal with an amplitude of 70 [dB.sub.AE] or 3'160 [micro]V is 10'900 [micro][m.sup.2]. This roughly corresponds (if the crack area is considered to be a square or circular shape) to a crack length of 100 [micro]m. Taking 80 and 60 [dB.sub.AE] as upper and lower AE signal amplitude limit, the crack length range in the test phase with high AE activity and correspondingly fast delamination growth spans roughly 60 to about 190 [micro]m. These values are consistent with an upper bound estimate of a micro-crack size of less than 200 [micro]m from the diameter of the protrusions seen forming stochastically along the crack tip in the X-ray images.

With threshold values of 40 and 30 [dB.sub.AE], i.e., 100 [micro]V and about 32 [micro]V, the minimal detectable crack length, estimated from the average sensitivity of 0.00029 mV/[micro][m.sup.2] amount to about 20 and 10 [micro]m, respectively. With complementary information on the total defect size, e.g., the delamination area or fracture surface from other non-destructive test methods, AE analysis can yield rough quantitative estimates of the damage size caused by the microscopic source mechanism. It is, however, not straight-forward to compare sensitivities estimated in different tests on different materials directly, since the main effects (material type, specimen size/shape, AE sensor location relative to the source, and transfer function of the measurement chain) may have acted in different ways. Nevertheless, this confirms that AE monitoring easily achieves sensitivity to defects on the scale of a few tens of microns in CFRP laminates, at least in laboratory-scale applications.

Acknowledgment

The contributions of Mr R Jaggi and of Dr R A Nordstrom for the in-situ AE monitoring and simultaneous radiography measurements and their analysis as well as discussions with members of the committee on acoustic emission of the German Society for Non-Destructive Testing, especially Dr. J Bohse, are gratefully acknowledged.

References

[1.] C R Rios-Soberanis, Acoustic Emission Technique, an Overview as a Characterization Tool in Materials Science, Journal of Applied Research and Technology, Vol:9 No. 3 pp. 367-379, 2011.

[2.] M E Ibrahim, Nondestructive evaluation of thick-section composites and sandwich structures: A review, Composites: Part A: Applied Science and Manufacturing, Vol. 64, pp. 36-48, 2014.

[3.] E Proverbio, Evaluation of deterioration in reinforced concrete structures by AE technique, Materials and Corrosion, Vol. 62, No. 2, pp. 161-169, 2011.

[4.] S C Lovejoy, Acoustic Emission Testing of Beams to Simulate SHM of Vintage Reinforced Concrete Deck Girder Highway Bridges, Structural Health Monitoring - An International Journal, Vol. 7, No. 4, pp. 329-346, 2008.

[5.] H N G Wadley, C B Scruby and G Shrimpton, Quantitative acoustic emission source characterisation during low temperature cleavage and intergranular fracture, Acta Metallurgica, Vol. 29, pp. 399-414, 1981.

[6.] M G R Sause, A Gribov, A R Unwin and S Horn, Pattern recognition approach to identify natural clusters of acoustic emission signals. Pattern Recognition Letters, Vol. 33, pp. 17-23, 2012.

[7.] M G R Sause, T Muller, A Horoschenkoff and S Horn, Quantification of failure mechanisms in mode-I loading of fiber reinforced plastics utilizing acoustic emission analysis. Composites Science and Technology, Vol. 72, pp. 167-174, 2012.

[8.] A J Brunner, E Hack and J Neuenschwander, Nondestructive Testing of Polymers and Polymer-Matrix Composites, Wiley Encyclopedia of Polymer Science & Technology (Ed.Arza Seidel), J. Wiley, pp. 1-39, 2015.

[9.] F Baensch, M Zauner, S J Sanabria, B R Pinzer, M G R Sause, A J Brunner, M Stampanoni and P Niemz, Damage Evolution in Wood--Synchrotron based microtomography as complementary evidence for interpreting acoustic emission behavior, Holzforschung, Vol. 68, No 8, pp 1015-1025, 2015.

[10.] F Baensch, M G R Sause, A J Brunner and P Niemz, Damage Evolution in Wood--Pattern Recognition based on Acoustic Emission Frequency Spectra, Holzforschung, Vol. 69, No. 3, pp. 357-365, 2015.

[11.] F Baensch, Damage evolution in wood and layered wood composites monitored in situ by acoustic emission, digital image correlation and synchrotron based tomographic microscopy, PhD Thesis, ETH Zurich, No. 22372, pp 1-186, 2015.

[12.] L Vergeynst, M G R Sause, F Ritschel, A J Brunner, P Niemz and K Steppe, Finite element modelling used to support wood failure identification based on acoustic emission signals, Proceedings COST--Timber Bridge Conference (Eds S Franke, B Franke, R Widmann), pp. 141-146, 2014.

[13.] A J Brunner, B R K Blackman and P Davies, A status report on delamination resistance testing of polymer-matrix composites, Engineering Fracture Mechanics, Vol. 75, No. 9, pp. 2779-2794, 2008.

[14.] International Organisation for Standardisation, ISO, Fibre-reinforced plastic composites--determination of mode I interlaminar fracture toughness, [G.sub.IC], for unidirectionally reinforced materials, 15024, pp. 1-24, 2001.

[15.] International Organisation for Standardisation, ISO, Fibre-reinforced plastic composites--Determination of the mode II fracture resistance for unidirectionally reinforced materials using the calibrated end-loaded split (C-ELS) test and an effective crack length approach, 15114, pp. 1-18, 2014.

[16.] American Society for Materials and Testing, International, ASTM, Standard Test Method for Mixed Mode I-Mode II Interlaminar Fracture Toughness of Unidirectional Fiber Reinforced Polymer Matrix Composites, D6671/D6671M-13e1, pp. 1-15, 2013.

[17.] A J Brunner, Fracture mechanics of polymer composites for aerospace applications, Chapter 8 in 'Polymer Composites in the Aerospace Industry' (Eds P E Irving, C Soutis), Woodhead Publishing, Series in Composites Science and Engineering: Number 50, pp. 191-230, 2015.

[18.] T K O'Brien and R H Martin, Results of ASTM round robin testing for mode I interlaminar fracture toughness of composite materials, NASA Technical Memorandum 104222, pp. 1-120, 1992.

[19.] T K O'Brien and R H Martin, Results of ASTM round robin testing for mode I interlaminar fracture toughness of composite materials, Journal of Composites Technology and Research, Vol. 15, No. 4, pp. 269-281, 1993.

[20.] T de Kalbermatten, R Jaggi, P Flueler, H H Kausch and P Davies, Microfocus radiography studies during mode I interlaminar fracture tests on composites, Journal of Materials Science Letters, Vol. 11, pp 543-546, 1992.

[21.] P Flueler and A J Brunner, Crack Propagation in Fiber-Reinforced Composite Materials Analysed with In-Situ Microfocal X-Ray Radiography and Simultaneous Acoustic Emission Monitoring, Proceedings European Conference on Composites Testing and Standardisation ECCM-CTS, (Eds. P J Hogg, G D Sims, F L Matthews, A R Bunsell and A Massiah), European Association for Composite Materials EACM (ISBN 29506577-1-0), pp 385-394, 1992.

[22.] J Bohse, T Krietsch, J Chen and A J Brunner, Acoustic Emission Analysis and Micromechanical Interpretation of Mode I Fracture Toughness Tests on Composite Materials', Proceedings ESIS Conference on Fracture of Polymers, Composites and Adhesives (Ed. J G Williams and A Pavan), ESIS Publication 27, Elsevier, pp 15-26, 2000.

[23.] L R O Hein, K A Campos, P C R O Caltabiano and K G Kostov, A Brief Discussion About Image Quality and SEM Methods for Quantitative Fractography of Polymer Composites, SCANNING, Vol. 35, pp. 196-204, 2013.

[24.] K A de Campos, J Augusto, T A Pereira, L R de Oliveira Hein, 3-D reconstruction by extended depth-of-field in failure analysis--Case study II: Fractal analysis of interlaminar fracture in carbon/epoxy composites, Engineering Failure Analysis, Vol. 25, pp. 271-279, 2012.

Andreas J. BRUNNER

(1) Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Mechanical Systems Engineering; Dubendorf, Switzerland

Phone: +41 58 765 44 93, Fax: not available; e-mail: andreas.brunner@empa.ch
Table 1: Estimated average and upper and lower bounds for the crack
area created in the AS4/PEEK CFRP laminate by an AE signal of average
amplitude

Specimen No.  Estimated         Estimated lower    Estimated
              average AE        bound              number of AE
              signal amplitude  delamination       signals [-]
              [[micro]V]        area [[mm.sup.2]]

1             -                 -                  -
2             3.160             538                35.620
3             3.980             459                32.790
4             3.160             432                51.970

Specimen No.  Estimated upper
              bound for
              sensitivity per
              average AE
              signal [mV/[micro][m.sup.2]]

1             -
2             0.00021
3             0.00028
4             0.00038
COPYRIGHT 2016 Acoustic Emission Group
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Brunner, Andreas J.
Publication:Journal of Acoustic Emission
Article Type:Report
Date:Jan 1, 2016
Words:4376
Previous Article:Comparison of localization strategies of damage in concrete by acoustic emission.
Next Article:Advanced signal analysis applied to discriminate different corrosion forms by acoustic emission data.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters