Cutting Procedures, Tensile Testing, and Ageing of Flexible Unidirectional Composite Laminates

Many iterations of cutting and testing were performed to investigate several different variables. Some variables that were examined include the cutting technique and cutting instrument, the testing rate, the specimen dimension, and the grips. One critical finding was the importance of aligning the specimens with the fiber direction. Data analysis procedures (consistency analysis, Weibull techniques, outlier determination, etc.) are discussed below, as are considerations for ageing.

Cutting technique/instrument

The cutting instrument may influence the measured failure stress because of the various levels of precision associated with each type of cutting instrument. The specimens referenced in Figure 2, Figure 3, and Figure 4 were all cut with an electrically powered fabric cutter. In contrast, all other specimens were cut using the procedure outlined above in section 1 of the protocol, and the results for these specimens are presented in Figure 8 and Figure 10. The specimens cut with the powered fabric cutter had an average failure stress of 872 MPa (standard deviation of 46 MPa, 102 specimens), while similarly sized specimens cut with a medical scalpel had an average failure stress of 909 MPa (standard deviation of 40 MPa, 40 specimens). These results are not surprising, as a closer examination of the edges of the specimens shows that the powered fabric cutter saw creates a much more jagged edge than the scalpel, as seen in Figure 5, effectively narrowing the width of the specimen.

The difference in mechanical performance between specimens cut using these two cutting tools led to a structured investigation of various cutting tools. Specimens were cut using each tool and then imaged. Figure 6, Figure 7, and Supplemental Figure 7 show the resulting edges at high magnification, and Supplemental Figure 8 at lower magnification, for a) an electrically powered fabric cutter, b) a ceramic knife, c) a precision ceramic cutter, d) a rotary blade, e) a utility knife, and f) a medical scalpel.

There appear to be both localized areas of damage and broader regions of damage exhibited in these images. The most localized damage is observed when fibers protrude from the frayed fiber edges or the edge of the fiber is bent and flattened by the blade as in Figure 6a. The broader regions of damage are observed as shearing and potential debonding, which occur in the cross fibers.

Figure 6 and Figure 7 show that the use of the scalpel provides the cleanest cut with the most localized damage, as Figure 6f and Figure 7f depict cleaner cuts than seen in the other panels of Figure 6 and Figure 7. The cross fibers show no evidence of the fibers shearing due to the cut, and the damage at the end of the cross fibers is restricted to approximately half the fiber diameter. The utility knife creates a slightly larger damaged zone; however, the resulting fiber cross sections are cleaner than those utilizing cutting methods other than the scalpel. All the other cutting methods create localized damage to an extent greater than one fiber diameter. Both the scalpel and the utility knife are sharp enough to split a fiber along its length and can result in a slightly ragged edge, as seen in Figure 5f,g. This is in contrast to Supplemental Figure 7d, where the precision ceramic cutter damages the edge fibers by flattening them instead of cutting through them. Slicing through the edge fiber does not result in a large damaged zone in the bulk of the specimen, which would be created if an edge fiber were to be pulled out.

Figure 5, Figure 6a, and Supplemental Figure 7b show typical damage due to the electrically powered fabric cutter. It creates an extremely frayed edge at a variety of length scales. The ceramic utility knife cuts in small sections, causing large-scale delamination and shear in groups of fibers, as can be seen in Figure 6b and Figure 7c. This is less prevalent with the precision ceramic cutter, although those results are not devoid of uneven cuts and frayed fibers, as seen in Supplemental Figure 8e. Cuts made with the rotary blade are not as straight as the other cutting methods (as seen in Supplemental Figure 7e, Supplemental Figure 8f,g, and Figure 7a,b) and can have large-scale fiber pullout (Supplemental Figure 7e). The images of cuts made by the utility knife and medical scalpel show little evidence of large-scale shear, delamination, or fiber pullout, as seen in Figure 6e,f, Figure 7e,f, and Supplemental Figure 7g,h. Comparing Supplemental Figure 8h with Supplemental Figure 8i, the medical scalpel does result in a better edge than the utility knife, with fewer frayed fibers sticking out, although for both methods, such fibers are only observed occasionally.

When cutting precision samples for an examination by SEM, the scalpel gives the best performance. The ceramic utility knife pulls at the fibers at the beginning and ends of cuts, as does the precision ceramic cutter. The metal utility knife introduces maximum fiber pulls at the beginning of a cut. Cutting smaller sample pieces with either the powered fabric cutter or the rotary blades can be challenging and is impractical.

The medical scalpel is the most precise in cutting nearest to the straight edge. The precision ceramic cutter has a large offset from the straight edge, in contrast, leading to more error in cutting a precise width of specimen. The rotary fabric cutter does not always cut the material but, instead, folds it at the point of the blade. The electric fabric cutter cannot be used against a straight edge, so it is difficult to make a perfectly straight cut with this tool. Thus, the medical scalpel tends to give the straightest cut nearest to the straight edge. It is also recommended that the cutting blade is replaced if it becomes nicked or damaged, or if the cut edges on the specimens no longer appear smooth when compared under a microscope to the edges cut with a fresh blade.

Importance of aligning specimens with fiber direction

An early set of tests consisted of 40 specimens that were cut using the electrical fabric cutter and had a width of 25 mm and a gauge length of 150 mm. These specimens were tested at a displacement loading rate of 40 mm/min, using the nonoptimized initial grip design. The testing showed that specimens 1 through 20 were well aligned with the fiber direction, while specimens 21 through 40 were accidentally misaligned by less than 2° (i.e., the fiber direction was not parallel to the main length direction of the specimen). When a specimen is misaligned, a characteristic behavior is observed during the test. One side of the specimen will shear upward while the opposite side shears downward, such that a line that was drawn straight across the specimen before testing will no longer be straight. This is depicted in Supplemental Figure 6 and is due to the edge fibers not being in both capstans.

Due to the misalignment of specimens 21 through 40, there is a distinct difference between the maximum stress (occurring at failure) of specimens 1 through 20 as compared to specimens 21 through 40, as can be seen in Figure 2. Figure 2a presents the maximum stress (occurring at failure) as a function of the specimen number for the misaligned specimens. A homogeneous population of maximum stress would be evenly distributed across the whole area, as in Figure 2b. However, in Figure 2a, there are no data in the first and third quadrants, other than one outlier in quadrant 3, marked as specimen number 13. Figure 2c is a Weibull plot of the two groups and includes the 99% confidence bounds for the associated Weibull distributions. The distributions from the first 20 specimens, group 1, and the second 20 specimens, group 2, are again different, with specimens 1 through 20 exhibiting a higher stress-to-failure than specimens 21 through 40. This observation is further clarified in Figure 2d, where the outlier specimen, number 13, has been removed. In Figure 2d, only one data point barely overlaps with the 99% confidence bounds of the other group; otherwise, there is no overlap in the data.

A misalignment of the specimen with the fiber direction of the material has been shown to give deceptively weaker results, as the misalignment effectively narrows the specimen width. This can be avoided by frequently determining the fiber direction during cutting, taking care to prevent the material from shifting, and measuring from a fixed point on the cutting mat (as compared to the specimen edge) when cutting the specimens. A misalignment can be observed experimentally during testing through its characteristic distortion pattern, as shown in Supplemental Figure 6. If the specimens are all equally misaligned, the effect will be mostly in the Weibull scale parameters. In contrast, if the specimens are randomly misaligned, both the Weibull shape and scale parameters will be affected.

Theory

When tested in tension along the fiber direction, UD laminates can be assumed to behave similarly to a fiber tow, comprised of parallel fibers in a matrix. When a fiber breaks, it will redistribute its load over neighboring fibers over some width and length, and a useful model could be built around the concept of a chain of small bundles of filaments, where the surviving filaments share the load equally. So inevitably, fiber strength properties and strip properties are related, as described by Coleman^19–23. A detailed discussion of applicable theory can also be found in Phoenix and Beyerlein²⁴, and the time-dependent properties of fibers were addressed by Phoenix and Newman^{25, 26}. This theory develops a Weibull failure distribution starting from the assumption that the occurrence of natural, inherent flaws along a fiber is well described by a Poisson-Weibull model. From this, a size effect naturally falls out. Simply put, the larger the volume of material, the lower the failure stress. This is due to the fact that, in a larger volume of material, there is a higher probability that the natural, inherent flaws in the fibers will collocate, creating a weak spot, and thus, lowering the failure stress.

Testing rate

Table 1 shows a comparison of results using three different loading rates. As the loading rate increases, the failure stress also increases. There does not appear to be an effect on the failure strain, so the modulus also appears to increase with an increasing loading rate.

The advantage of testing at different loading rates is that the tests interrogate different aspects of the composite. Slow tests are more reliant on the matrix properties, particularly matrix shear creep, while fast tests primarily explore fiber failure stress^{25, 26}. It is important in choosing a loading rate to pick one that captures the behavior of interest.

Specimen width

Table 2 shows the effect of increasing the specimen width. By increasing the specimen width, the edge effects from cutting should become less important as they take up less of the specimen width. Also, any inaccuracies in measuring the width of the specimens become less important. The increased consistency with increased specimen width is observed in the decrease of the standard deviation of the failure stress. At a width of 10 mm, the mean failure stress is lower, and the standard deviation is higher than that of wider specimens, suggesting that narrow specimens can suffer from significant edge effects. The failure strain decreases with increasing width, perhaps also due to the lessened impact of edge effects.

The wider the specimen width, the smaller the influence will be from edge effects and, therefore, the increased consistency of the specimens. Thus, wider specimens yield better results. However, there is a trade-off in terms of material expense and the cost of grips to test wider, and thus stronger, specimens.

As discussed above, theory predicts a decrease in failure stress with increasing width²⁴. This is noted when comparing the specimens that are 30 mm with the 70 mm-wide specimens. The large decrease in failure stress of the 10 mm-wide specimens is probably due to the increased significance of edge effects at such narrow widths.

Specimen length

As previously discussed, the theory predicts a decrease in failure stress with increasing length²⁴. The results presented in Table 3 show this but are also confounded by the loading rate being constant at 10 mm/min, rather than holding the strain rate constant. Decreasing the strain rate (as happens with a fixed loading rate of 10 mm/min and an increasing gauge length) also causes a decrease in failure stress. The standard deviation for the failure stress increases more than can simply be explained by the different strain rates. This phenomenon could be because longer specimens are more difficult to cut, and edge fibers invariably get cut somewhere along the edge length, effectively reducing the width of the specimen in a random way. Specimens longer than the length of the cutter’s arm are particularly difficult, as it no longer becomes possible to cut them with a single smooth cut with constant velocity. The decrease in the failure strain as the length increases indicates that not all the decrease in failure stress is due to the slower strain rate for longer specimens.

Specimens tested to failure with a gauge length of 100 mm typically show delamination throughout the entire gauge length of the specimen. Specimens tested to failure with a gauge length of 900 mm, exhibit delamination occurs only in a region (typically near the middle) of the gauge, leaving a sizeable portion of the specimen intact, as could be expected from a chain-of-bundles model.

Grips

The grips should be in capstan style. Rotating capstans provide more ease in loading, and only four locking positions for the capstan helps ensure consistency. Capstan grips that close and clamp on the material can be used on exceedingly high-strength slippery materials. However, the fixed opening capstans used in this study work for both UHMMPE and aramids.

A study was done comparing two different types of capstan grips, using a different material. For the first set, the capstan was fixed, and the specimen was not aligned with the load cell but, instead, offset by half the width of the capstan. The second set consisted of rotating capstans with pins to lock them in place during testing. Furthermore, these capstans were offset to align the specimen with the load cell and, thus, prevent a moment on the load cell during loading. The failure load distributions were very similar for these grips, as shown in Figure 8. The rotating grips may give a marginally weaker distribution than the fixed grips, likely due to their wider radius capstan and, thus, longer load transfer length. Furthermore, the fixed grips may have a marginally larger variance than the rotating grips, as there is a higher likelihood of damaging the specimen during loading when the capstans are fixed due to the difficulties in wrapping the specimen around the capstans. The difference between these grips is evident when comparing load vs. extension plots. The results from ten representative specimens are shown in Figure 9 for the fixed and rotating grips. The curves for the rotating grips are smooth and consistent, while in contrast, the fixed grip curves frequently show that the specimens were slipping. When the capstans are fixed in place, it becomes challenging to tighten down on the material, as several wraps are required to prevent the specimen from slipping through the grips entirely.

Data analysis

There is a certain amount of variability inherent in UD laminate materials. The goal of the cutting/testing procedure presented herein is to minimize the additional variability added in specimen preparation and testing. Outlying data points could either be attributed to the inherent distribution of the UD laminates or could be a cutting/testing artifact. The following paragraphs discuss a few techniques to separate the artifacts from the distributions.

Failure stress as a function of specimen number

A plot of the failure stress as a function of specimen number can show general trends in a group of specimens. Unless the material is variable on the macro scale, the inherent variability of the material should not be observed on such a plot. Figure 2b shows an example of a group of self-consistent specimens, in contrast to Figure 2a.

This lack of consistency amongst specimens may not be evident in other analyses. Returning to the example of the misaligned specimens, the difference in failure stress is clear from Figure 2. However, it is not clear from looking at the data for specimens 1 through 40. This is shown in Figure 3, a Weibull plot with 99% confidence bounds for specimens 1 through 40. There is no obvious indication in Figure 3 that the cutting was inconsistent. Furthermore, the failure strains for these same specimens, plotted in Figure 4 as a function of specimen number, also show no evidence of the misalignment/lack of consistency, while the failure stresses do, as shown in Figure 2a.

Weibull distribution and outliers

Given the nature of this UD laminate material, it is expected to have a Weibull failure stress distribution^19–26. This distribution is expected to have a shape parameter that is considerably higher than the associated shape parameter for a single fiber, due to the load-sharing among fibers^24–26. Standard statistical tests can be performed to determine if the failure stress of a batch of specimens is well described by a Weibull distribution.

With the Weibull distribution, a certain number of low-strength specimens are expected. This makes the determination of outliers more difficult than if the data were from a normal distribution. For example, in Figure 9c, the specimen giving a datum in the lower left quadrant appears to be an outlier. Figure 9b presents the same data, only without the potential outlier identified in Figure 9a. Suspect data points should be investigated, particularly those that fall outside the 95% maximum likelihood confidence interval.

Ageing

Table 4 presents the ageing results for specimens 30 mm wide with an effective gauge length of 300 mm, tested at a loading rate of 10 mm/min. These results show no effects of ageing. PPTA has previously been shown to be resistant to degradation caused by temperature and humidity^{1, 2}. Thus, it is not particularly surprising that tensile tests at this strain rate, where the matrix does not play a major role, do not show significant degradation over time, for the period allowed for this ageing experiment.

In summary, the cutting technique can play a big role in the effective width of the specimen, so it is important to choose one that gives consistent results with a minimum of specimen damage. A medical scalpel was found to work best in this study. The type of grips can lead to misleading features in the stress-strain curves; thus, based on this study, rotating capstans are recommended. The loading rate, specimen width, and specimen length all affect the final strength value and must be chosen with care. In particular, the specimen width must be wide enough so that any fluctuations in cutting do not have an undue influence on the results, and the specimen length must be long enough that the specimen fails between the grips but not so long as to make it hard to cut. By holding all of the above constant, scientists can identify the effects of ageing.

Figure 1: SEM image of UD material, with red and blue lines following individual surface fibers to highlight nonparallel fibers. Please click here to view a larger version of this figure.

Figure 2: Plots of failure stress for aligned and misaligned specimens. (a and b) Plots of the failure stress of each specimen as a function of its specimen number. Panel a consists of 40 specimens of which group 1, specimens 1–20 and circled in red, are well aligned and group 2, specimens 21–40 and circled in blue, are misaligned with the fiber direction. Panel b consists of 40 well-aligned specimens. (c and d) Plots of the Weibull distributions of the two groups with 99% confidence bounds, showing a minimal overlap of the data points from group 2 with the bounds of group 1. Panel c shows an outlier. Panel d does not show specimen 13, which is an outlier as it is far away from the maximum likelihood estimate for the distribution. The specimens were about 25 mm wide, tested at nominally 40 mm/min, and cut with an electric fabric cutter. Please click here to view a larger version of this figure.

Figure 3: A Weibull plot of both group 1 and 2 (as described in Figure 2) together, showing 99% confidence bounds. Please click here to view a larger version of this figure.

Figure 4: A plot of the failure strain of each specimen as a function of its specimen number, for the same set of specimens as shown in Figure 2 and Figure 3. The specimens were about 25 mm wide, tested at a tensile displacement loading rate of approximately 40 mm/min, and cut with an electric fabric cutter. Please click here to view a larger version of this figure.

Figure 5: A jagged edge, typical of a cut made with the electrically powered fabric cutter. Please click here to view a larger version of this figure.

Figure 6: SEM images of the edges of the cross-cut fibers with insets of stereomicroscope images. The cut was made with (a) an electrically powered fabric cutter, (b) a ceramic knife, (c) a precision ceramic cutter, (d) a rotary blade, (e) a utility knife, and (f) a medical scalpel. Please click here to view a larger version of this figure.

Figure 7: Overview of the cut, produced by SEM images of the corners. SEM images of the corners, giving an overview of the cut produced by (a) an electrically powered fabric cutter, (b) a ceramic knife, (c) a precision ceramic cutter, (d) a rotary blade, (e) a utility knife, and (f) a medical scalpel. Please click here to view a larger version of this figure.

Figure 8: Weibull plot comparing the failure load for two different sets of capstan grips. Please click here to view a larger version of this figure.

Figure 9: Load vs. extension plots of 10 representative specimens. Testing performed using (a) fixed and (b) rotating capstan grips Please click here to view a larger version of this figure.

Figure 10: Failure stress distributions. Failure stress distributions plotted using Weibull scaling, for specimens with a gauge length of 300 mm, a width of 30 mm, loaded at 10 mm/min, and cut along the ‘warp’ direction, (a) including an outlier and (b) without outlier. Please click here to view a larger version of this figure.

Table 1: Mean values, with standard deviations in parenthesis, showing the effects of varying the loading rate on specimens with a gauge length of 300 mm, 30 mm wide, and cut along the ‘warp’ direction, where each batch is at least 35 specimens.

Table 2: Mean values, with standard deviations in parenthesis, showing the effects of varying the width on specimens with a gauge length of 300 mm, a loading rate of 10 mm/min, and cut along the ‘warp’ direction, where each batch is at least 35 specimens.

Table 3: Mean values, with standard deviations in parenthesis, showing the effects of varying the length on specimens with a width of 30 mm, a loading rate of 10 mm/min, and cut along the ‘warp’ direction, where each batch is at least 35 specimens.

Table 4: Mean values, with standard deviations in parenthesis, showing the effects of ageing at 70 °C with 76% RH on specimens with a gauge length of 300 mm, a width of 30 mm, a loading rate of 10 mm/min, and cut along the ‘warp’ direction, where each batch is at least 35 specimens.

Supplemental Figure 1: Schematic of UD laminates. (a) Fiber (cylinders) orientation in two unidirectional (UD) layers, one with a 0° orientation and the other with a 90° orientation. (b) Schematic for cutting a piece of UD material from its bolt. The bolt’s width is measured along the red dotted line. For the piece of material cut off, the length is measured along the red dotted line, and the width is measured perpendicular to the length. The ‘warp’ direction is indicated by the blue arrow, and the ‘weft’ direction is indicated by the red arrow. The principal fiber direction is defined as the direction of the uppermost layer (i.e., along the red arrow/weft direction). Since the principal fiber direction refers to the layer that is being viewed (the top layer), turning the material over will change the principal fiber direction from weft to warp. Note that there is no warp and weft in the traditional textile sense, as the material used here is not woven. (c) Schematic showing a small tab of material, cut in preparation for separation. (d) UD laminate after separating the top layer from the unidirectional material. The green dashed line indicates where to cut to separate the precursor material from the roll. Please click here to download this file.

Supplemental Figure 2: SEM comparison. SEM comparison was performed between (a) a side view of a new, sharp scalpel blade with an unnotched edge, (b) an edge-on view of a new scalpel blade showing how the blade comes to a fine point, (c) a side view of a used scalpel blade with a defect in the edge and scratches along the edge, and (d) an edge-on view of a used scalpel blade showing that the blade no longer has as fine an edge and is now dull. Arrows mark the blade’s edge. Please click here to download this file.

Supplemental Figure 3: A used scalpel blade, with the arrow pointing to scratches along the length of the blade. Please click here to download this file.

Supplemental Figure 4: Cutting layout. Specimens are cut along the weft direction, where the red arrow indicates both the principal fiber direction and the weft direction, while the blue arrow indicates the warp direction. The terms weft and warp are used to reference standard textile directions, although they are not strictly applicable as the UD material is not woven. Please click here to download this file.

Supplemental Figure 5: Photographs of the specimen at various stages of preparation. (a) Marking video extensometer points using a template. (b) Loading the specimen, specifically positioning the end of the specimen at the grip line. Take care to center the specimen on the capstan grips by aligning the center of the specimen within approximately 1 mm of the center of the capstan grips. (c) Specimens in the environmental chamber. Please click here to download this file.

Supplemental Figure 6: Schematic of characteristic behavior during loading of a misaligned specimen. A horizontal line is drawn across it. (a) Schematic of the unloaded specimen. In (b), the specimen is loaded. (c) Actual misaligned specimen. The red arrows show the direction of the applied stress. Please click here to download this file.

Supplemental Figure 7: SEM images focusing on typical cutting damage on material cut. The cuts were made with (a) a dull utility knife; (b) an electrically powered fabric cutter, showing large amounts of damage parallel to the cut fibers; (c) a ceramic knife, showing how the knife cuts in sections, as well as the large sheared region that extends well into the material; (d) a precision ceramic cutter, showing how the ceramic blade does not cut through the fibers themselves; (e) a rotary blade, showing fiber pullout as well as a wavy cutting edge; (f) a utility knife, showing how a utility knife cuts through the fibers and can have a hairy edge; (g) a medical scalpel, showing how the scalpel can cleanly slice through fibers; (h) a medical scalpel, showing that the damage from the cut is localized with no larger-scale shear, delamination, or fiber pullout. Please click here to download this file.

Supplemental Figure 8: Stereomicroscope images of typical edge defects. The cut was made with (a) an electrically powered fabric cutter, showing large-scale frayed edges; (b) an electrically powered fabric cutter, showing small-scale frayed edges; (c) a ceramic knife, showing uneven cutting; (d) a ceramic knife, showing frequently frayed fibers; (e) a precision ceramic cutter, showing uneven cutting and frayed fibers; (f) a rotary blade, showing a cleaner yet less straight edge; (g) a rotary blade, showing a fairly common defect; (h) a utility knife, (i) a medical scalpel. Please click here to download this file.