1.

20 Strength Analysis of Unidirectional Composites and Laminates

C. T. SUN Purdue University, West Lafayette, IN, USA
1.20.1 INTRODUCTION 1.20.2 LAMINA FAILURE ANALYSIS 1.20.2.1 Lamina Strength Criteria 1.20.2.2 Comparison Among Lamina Strength Criteria 1.20.2.2.1 Bidirectional stress plane 1.20.2.2.2 Off-axis loading 1.20.2.2.3 Pure shear 1.20.2.3 Comparison with Experimental Data 1.20.2.3.1 Lamina strength criteria comparison with off-axis tension data 1.20.2.3.2 Lamina strength criteria comparison with tubular specimens 1.20.2.3.3 Discussion 1.20.2.3.4 Concluding remarks 1.20.3 LAMINATE STRENGTH ANALYSIS 1.20.3.1 Stiffness Reduction 1.20.3.1.1 Parallel spring model 1.20.3.1.2 Incremental stiffness reduction model 1.20.3.2 Laminate Failure Analysis Methods 1.20.3.2.1 Ply-by-ply discount method 1.20.3.2.2 Sudden failure method 1.20.3.3 Laminate Failure Analysis for Biaxial Loading 1.20.3.3.1 Comparison with data for biaxial loading 1.20.3.3.2 Biaxial failure in the strain plane 1.20.3.3.3 Biaxial testing data for glass woven fabric composite 1.20.3.4 Laminate Strength Analysis for Unidirectional Loading 1.20.3.4.1 Selection of laminates and off-axis loading angles 1.20.3.4.2 Laminate coupon specimens and oblique end tabs 1.20.3.4.3 Results and data analysis 1.20.3.4.4 Comparison with test data 1.20.3.5 Discussion 1.20.4 CONCLUSIONS 1.20.5 REFERENCES 2 2 2 3 3 5 7 8 9 9 10 13 15 15 15 16 16 16 17 17 17 18 19 19 19 20 20 21 23 25 26

2 1.20.1

Strength Analysis of Unidirectional Composites and Laminates INTRODUCTION stresses control the failure mechanisms. Free edge delamination and failure in certain types of laminate belong to this category. The purpose of this study is to evaluate and compare these strength criteria based on available experimental data. Such an exercise may be useful for revealing the characteristics of each criterion considered and providing some useful guides in using these criteria in practical applications.

Prediction of the load-carrying capability of a structure is one of the most important tasks in structural design. In the last four decades, the subject of strength (failure) criteria for composite materials has attracted numerous researchers who have developed and are continuously developing criteria for predicting the strength of composites and their laminates (Nahas, 1986; Labossiere and Neale, 1987). Undeniably, progress has been made. However, until now it does not appear that there is any criterion that is universally accepted by composites researchers as adequate under general loading conditions. In fact, there is a growing consensus that the strength analysis of composite structures is far from being mature. Evidence of this is the publication of a recent special edition of Composites Science and Technology, 1998, 58(7) which was entirely dedicated to the subject of failure theories for continuous fiber-reinforced composites. Although none have been rigorously verified experimentally, many strength criteria for composites have been formally introduced in almost all existing textbooks on composite materials. Often, without giving the reader proper warning about the inadequacy of these strength theories, these textbooks have helped create the impression that accurate strength analysis tools are available. For this reason, there seems to be a need to evaluate the accuracy of these strength criteria that are presented in textbooks. Indeed, a recent report by Sun et al. (1996) attempted to perform some comparative evaluations of the conventionally adopted failure analysis methods. In this chapter, six representative and most popular strength criteria are chosen for study. These criteria are commonly referred to as Maximum Stress, Maximum Strain, HillTsai, TsaiWu, HashinRotem, and Hashin criteria. All are based on the assumption that fiber composites are orthotropic continua, and macro (averaged) stresses are used in the criteria. These criteria are grouped into noninteractive, fully interactive, and partially interactive theories. These criteria are basically phenomenological in which detailed failure processes are not described. Moreover, they are all based on linear elastic analysis. In this study, only in-plane failure of fiber composites is considered. Thus, the strength criteria are expressed in the state of plane stress. The analysis of laminate strength is performed using the classical laminated plate theory in conjunction with a stiffness reduction procedure. Such a procedure is recognized to be inadequate in laminates where interlaminar

1.20.2

LAMINA FAILURE ANALYSIS

The purpose of a lamina strength criterion is to determine the strength and mode of failure of a unidirectional composite or lamina in a state of combined stress. Many lamina strength criteria were developed for three-dimensional orthotropic materials. In this study, only the reduced two-dimensional criteria are included. In these criteria, the in-plane principal strengths of the composite are needed. X & X': tensile and compressive strengths in fiber direction; Y & Y: tensile and compressive strengths in transverse direction; S: in-plane shear strength. For a strain-based analysis, the corresponding ultimate strains are denoted by Xe,Xe',Ye,Ye', and Se, respectively. The ability of a lamina failure criterion to determine mode of failure is essential in bringing this analysis tool to the laminate level. In general, the following modes of failure are identified. Fiber breakage (mode 1): longitudinal stress (s11) or longitudinal strain (e11) dominates lamina failure. Transverse matrix cracking (mode 2): transverse stress (s22) or transverse strain (e22) dominates lamina failure. Shear matrix cracking (mode 3): shear stress (t12) or shear strain (g12) dominates lamina failure. It is noted that both mode 2 and mode 3 are matrix failures. The two modes are sometimes separated because they are caused by different stress components.

1.20.2.1

Lamina Strength Criteria

Lamina failure criteria can be categorized into three groups. (i) Limit criteria: these criteria predict the failure load for each mode of failure by

Lamina Failure Analysis comparing lamina stresses s11, s22 and t12, (or strains e11, e22, and g12) with corresponding strengths separately. Interaction among the stresses (or strains) is not considered. (ii) Interactive criteria: these criteria assume that all the stress components simultaneously contribute to the failure of the composite. They are usually expressed in the form of a single equation. (iii) Separate mode criteria: these criteria separate into a criterion for matrix failure mode and a criterion for fiber failure mode. The criterion for each failure mode may involve stress interactions. The six lamina strength criteria are given as follows. (i) Limit criteria Maximum stress:
s11 X s22 Y t12 S

Hashin (1980):
t12 2 11 2 s X S 1 s11 X0

fiber failure (tension) fiber failure (compression) matrix failure 6

t12 2 22 2 s Y x 1

1 1 1

fiber failure transverse matrix cracking shear matrix cracking 1

Maximum strain:
e11 Xe e22 Ye g12 Se

1 1 1

fiber failure transverse matrix cracking shear matrix cracking 2

(ii) Interactive criteria HillTsai (Azzi and Tsai, 1965; Tsai, 1965):
s 2
11

s 2
22

Y t 2
12

s s  11 22 X X

For Maximum Stress, Maximum Strain, HillTsai, and HashinRotem criteria, the longitudinal and transverse strengths must be chosen based on the sign of the applied stress. The TsaiWu criterion is designed for use in all stress quadrants of the stress plane. The Tsai Wu criterion requires a biaxial test to experimentally determine the interaction term F12. It has been suggested to use F12 = 1/(2XX'). Narayanaswami and Adelman (1977) found this term to be insignificant for the most part, and suggested setting it equal to zero. Cui et al. (1992) also found that F12 = 0 was within the range satisfying current engineering requirements. Thus, to avoid ambiguity, F12 is set equal to zero in the present study. The Hashin criterion listed here is a slight modification of the two-dimensional criterion presented in his 1980 paper. In that paper, Hashin suggested using a combination of both axial and transverse shear strengths SA and ST for the compressive matrix failure mode when s22 5 0. Since it is difficult to find transverse shear strength values in the literature, the tensile matrix failure equation given above is used as the compressive equation by simply replacing Y with Y'.

1.20.2.2 1.20.2.2.1
4

Comparison Among Lamina Strength Criteria Bidirectional stress plane

TsaiWu (Tsai and Wu, 1971):

F1 s11 F2 s22 F11 s11 2 F22 s22 2 2F12 s11 s22 F66 t12 2 1

where
F1 F11 F12 1 1 1 1 ; F2 0 ; X X0 Y Y 1 1 1 ; F22 F66 2 0 0 XX YY S experimentally determined

(iii) Separate mode criteria HashinRotem (Hashin and Rotem, 1973):

s11 X

fiber failure 5 matrix failure

t12 2 22 2 s Y S 1

A series of failure envelopes for combined stresses is presented to graphically show the characteristics of the six selected lamina strength criteria. These envelopes are composed of failure stresses normalized by the lamina's respective tensile strengths X and Y, or shear strength S. For these graphs, the AS4/3501-6 graphite/epoxy system tested by Sun and Zhou (1988) is used (see Table 1 for elastic and strength constants). Figure 1 is a plot of the selected criteria in a s11 7 s22 stress plane (t12 = 0). The Maximum Stress envelope is a simple rectangle bounded by the failure loads +s11 and +s22. For the analysis using the Maximum Strain criterion, failure strains are calculated from the strengths using the linear relationship:

4
Xe

Strength Analysis of Unidirectional Composites and Laminates

Table 1 7 Moduli and strength values of the AS4/ 3501-6 graphiteepoxy system. 138.90 GPa 9.86 GPa 5.24 GPa 0.30 0.132 mm X X' Y Y' S 2206.0 MPa 72013.0 MPa 56.5 MPa 7206.8 MPa 110.3 MPa

X X0 Y ; X0e ; Ye E1 E1 E2 Y0 S Y0e ; Se G12 E2

The Maximum Strain envelope is close to that of the Maximum Stress but is slightly skewed due to the effect of Poisson's ratio. There is considerably more skewing in the vertical (s22) direction because v12 > v21 in unidirectional fiber composites. Both HillTsai and TsaiWu criteria allow quadratic stress interactions; therefore, each has a curved failure envelope. Both criteria match up with the two limit criteria for all four unidirectional loading cases (+s11 with s22 = 0 and +s22 with s11 = 0) as expected. The TsaiWu criterion is a continuous curve throughout all four quadrants. The TsaiWu criterion includes linear stress terms while the HillTsai criterion, on the other hand, is a purely quadratic criterion. In order to account for differences in tensile and compressive strengths commonly found in fiber composites, this criterion uses the appropriate strength values in each quadrant (X or X' and Y or Y' accordingly). Though both are interactive,

E1 E2 G12 v12 Ply thickness

Source: Sun and Zhou (1988).

TsaiWu and HillTsai produce different failure envelopes in the stress plane. In the compressive s22 quadrants, the TsaiWu failure envelope extends beyond the longitudinal strengths X and X'. Finally, both the Hashin and HashinRotem criteria reduce to the Maximum Stress criterion in the s11 7 s22 plane since t12 = 0. A plot of the selected criteria in a s11 7 t12 stress plane (s22 = 0) is shown in Figure 2. The Maximum Stress envelope in this stress plane is again a rectangle, bounded by the failure loads +s11 normalized by X and +t12 normalized by S. The Maximum Strain criterion predicts exactly the same loads as the Maximum Stress

Figure 1 Comparison of lamina failure criteria under s11 7 s22 biaxial stress.

Lamina Failure Analysis

Figure 2 Comparison of lamina failure criteria under s11 7 t12 biaxial stress.

criterion. The HashinRotem criterion also reduces to the Maximum Stress criterion in this stress plane. Again, HillTsai and Tsai Wu failure envelopes intersect the other three criteria for the four unidirectional loading cases (+s11 with t12 = 0 and +t12 with s11 = 0). In the biaxial loading regions, the two interactive criteria are nearly identical. The linear stress term s11 in TsaiWu produces a slightly higher or lower failure load than in HillTsai, depending on the quadrant. Interestingly, the Hashin criterion reduces to Maximum Stress with a compressive s11, and reduces to HillTsai with tensile s11. Figure 3 contains a plot of the six criteria in the s227t12 stress plane (s11 = 0). Maximum Stress and Maximum Strain are identical rectangles showing +s22 normalized by Y and +t12 normalized by S. TsaiWu and Hill Tsai produce curved envelopes due to s22 7 t12 interaction. Again, the linear s22 term in TsaiWu produces a different shape than HillTsai, pushing the failure envelope beyond the lamina shear strength S. Hashin and HashinRotem in this stress plane match HillTsai exactly considering s11 = 0. As

expected, all six criteria intersect at the four unidirectional loading cases (+s22 with t12 = 0 and +t12 with s22 = 0).

1.20.2.2.2

Off-axis loading

The off-axis tension test is a simple way to determine the failure of a composite under combined stresses. Figure 4 shows predictions of the six criteria for the AS4/3501-6 composite whose properties are found in Table 1. In Figure 4, the loading angle y is defined as the angle between the fiber and loading directions. It is evident that all six criteria predict very similar failure stress sxx over the entire off-axis range. All six criteria predict a failure load of X at y = 08 and Y at y = 908. The Maximum Stress criterion predicts three separate failure regions representing the three possible modes of failure; i.e., fiber breakage, transverse matrix cracking, and shear matrix cracking. Between y = 08 and 2.98, fiber breakage is predicted (mode 1). At y = 08, the failure load is simply X. As y increases in this region, the predicted failure

Strength Analysis of Unidirectional Composites and Laminates

Figure 3 Comparison of lamina failure criteria under s22 7 t12 biaxial stress.

Figure 4 Comparison of lamina failure criteria for off-axis loading.

Lamina Failure Analysis

Figure 5 Comparison of lamina failure criteria for positive pure shear.

load actually increases just slightly. This is because the s11/X ratio remains dominant even though the s11 component in the lamina goes down; thus, a larger applied load is necessary to satisfy the dominant equation. At the critical angle y = 2.98, the t12/S ratio becomes dominant; therefore, the failure mode switches to shear matrix cracking (mode 3). This region continues until y = 278 where the failure mode switches again to transverse matrix cracking (mode 2) with the s22/Y ratio becoming dominant. This region continues through 908 where the failure load is simply Y. The Maximum Strain criterion produces a similar failure curve to that of Maximum Stress. From y = 08 to 2.98, mode 1 failure occurs. Maximum Strain predicts a slightly higher failure load than Maximum Stress in this region. In the shear region (mode 3), Maximum Strain results are identical to the Maximum Stress results. Maximum Strain's shear region extends to 298, where failure switches to mode 2. Note that the use of measured ultimate strains that include nonlinear effects would insignificantly alter the character of this failure curve. The two interactive criteria, HillTsai and TsaiWu, predict the exact three failure regions as the Maximum Stress criterion, with critical angles at 2.98 and 278 for mode 13 and mode 32 transitions, respectively. Since these criteria are completely interactive, their failure curves remain smooth throughout the entire off-axis loading case. Because X >> S, a criterion which couples these stress components (Hill Tsai and TsaiWu) will predict a noticeably

lower value in this area than a limit criterion (Maximum Stress and Strain). For the region where both t12 and s22 have significant contributions, the difference between the limit and interactive criteria diminishes because Y and S are of similar magnitude. The two interactive criteria eventually converge with the two limit criteria at y = 908 (only s22 exists). The two separate mode criteria exhibit characteristics of both the limit and interactive criteria. They yield the same three failure regions as the other criteria. Due to its ability to separate modes, HashinRotem's failure prediction is identical to Maximum Stress in the fiber (s11) dominated region y = 02.98. After the mode 13 transition at 2.98, HashinRotem's failure prediction begins to move away from Maximum Stress and towards the prediction of HillTsai.

1.20.2.2.3

Pure shear

The six lamina strength criteria are compared by using their failure predictions in a pure shear loading situation. The angle between the applied shear load and the lamina fibers is denoted by y as in the unidirectional off-axis loading previously discussed. The shear loading is given by +txy(sxx = syy = 0). A lamina composed of the material in Table 1 is used for this analysis. Figure 5 shows a +txy loading case for a single lamina rotated at an angle y. All six criteria predict a failure load of S at y = 08

Strength Analysis of Unidirectional Composites and Laminates

Figure 6 Comparison of lamina failure criteria for negative pure shear.

and 908, as expected. The criteria are all symmetric about y = 458 where a failure load of approximately 7Y' (compressive) is predicted. The criteria all predict shear matrix cracking from 08 to 30.88 and transverse matrix cracking from 30.88 to 458. For the Maximum Stress criterion, the mode 32 transition point dramatically alters the failure prediction. The s22/Y' term becomes dominant after the transition at y = 30.88. Because s22 is smaller at 30.88 than at 458, it actually requires a larger ultimate txy to satisfy the failure criterion. This explains the drastic change in the failure curve. Maximum Stress and Maximum Strain differ only in this transverse failure region due to the Poisson effect, and are plotted as one curve. They both predict a maximum failure load at the sheartransverse failure transition. The interactive criteria produce smooth curves for this loading case, even though they differentiate between modes of failure. The linear terms in TsaiWu gives its failure curve a different character than HillTsai, though both clearly eliminate the cusp formed by the limit criteria. Both separate mode criteria Hashin and HashinRotem closely match HillTsai. The separate mode criteria deviate slightly due to the small s11/X contribution found in the HillTsai criterion near y = 458, where s11 becomes a maximum. All interactive and separate mode criteria reach a maximum at 458 where s22 is dominant and at its maximum (txy = 7Y'). Figure 6 shows a 7txy loading case for a

single lamina rotated to an angle y. Like the +txy loading case, all criteria are symmetric about y = 458, and switch from a shear matrix cracking failure to transverse matrix cracking at y = 13.48. A strength of approximately txy = Y (tensile) is predicted at y = 458. The limit criteria predict an increasing failure load in the shear failure region. This is due to a decrease in the magnitude of the component t12, even though the t12/S ratio is dominant; higher loads are predicted in order to satisfy the dominant equation. Maximum Stress and Maximum Strain differ slightly in the transverse failure mode region. HillTsai, Hashin Rotem, and Hashin are nearly identical with only a small s11/X contribution in the HillTsai criterion separating them. TsaiWu's failure curve varies from the other interactive criteria due to its linear terms, primarily the s22 term. Again, because the shear loading is negative, the lamina is weakest at y = 458, in contrast with the positive shear case in which it is strongest at 458. In contrast to off-axis loading, in the case of pure shear loading these criteria predict quite different lamina strengths.

1.20.2.3

Comparison with Experimental Data

The accuracy of lamina strength criteria depends on reliable material strength data, i.e., X, X', Y, Y', and S, or the corresponding ultimate

Lamina Failure Analysis

Figure 7 Comparison of lamina failure criteria to off-axis data.

strains. Except for X and Y, good measurement of compressive strength and shear strength are not easy to obtain, which makes an objective assessment of the lamina failure criteria all the more difficult. 1.20.2.3.1 Lamina strength criteria comparison with off-axis tension data

Many authors have performed off-axis unidirectional lamina tensile tests. Because all the lamina strength criteria predict very similar failure loads, correlation with experimental data cannot be used as a means of ranking these criteria. Tests of a boronepoxy system by Pipes and Cole (1973) illustrate this point. The strength properties of this material system are provided in Table 2. Using those values, the theoretical predictions of the six lamina strength criteria along with the experimental data are plotted in Figure 7. For this range of off-axis angles, matrix failure dominates the strength. It is seen that the interactive and separate mode criteria yielded better predictions than the limit criteria. Other sets of offaxis data yield similar conclusions. 1.20.2.3.2 Lamina strength criteria comparison with tubular specimens

Wu and Scheublein (1974) generated biaxial lamina data using a graphiteepoxy (Morganite II) system with material constants listed in Table 3. Figure 8 shows the predictions for the material system vs. experimental data (s11 7 s22 plane). Clearly all criteria match at the four axis intercepts, considering that those data points are used to generate the failure envelopes. The data in the fourth quadrant is not sufficient to distinguish any one criterion. Swanson et al. (1987) obtained strength data for an AS4/55A unidirectional composite in the s22 7 t12 plane. Table 4 lists the strength properties. Figure 9 shows theoretical predictions of the six lamina strength criteria compared with the experimental data. It is evident that Hill Tsai, TsaiWu, HashinRotem, and Hashin (all including s22 7 t12 stress interaction) predict the data very well for tensile s22. However, for the combination of 7s22 and t12, only TsaiWu performs well. In fact, the test data indicate that lamina shear strength increases as the s22 component becomes compressive. In order to verify the aforementioned phenomenon, two additional independent sets of
Table 2 X X' Y Y' S
Source: Pipes and Cole (1973).

Strength values for the boronepoxy material system. 1296.2 MPa 72489.0 MPa 62.1 MPa 7310.3 MPa 68.5 MPa

Use of a tubular specimen allows a biaxial state of stress to be applied to a composite laminate. Use of such specimens also eliminates the free edge effect found in flat coupon specimens (Colvin and Swanson, 1990).

10

Strength Analysis of Unidirectional Composites and Laminates

Figure 8 Comparison of lamina failure criteria to s11 7 s22 data from Wu and Scheublein (1974).

s22 7 t12 biaxial data are analyzed. The first set is a T800/3900-2 graphiteepoxy tested by Swanson and Qian (1992). The second set is from tests by Voloshin and Arcan (1980) using glassepoxy (Scotch-Ply Type 1002). Both sets of material strength constants are given in Table 4. Figures 10 and 11 show the predictions from the six lamina failure criteria vs. experimental data. The trend of an increasing shear
Table 3 X X' Y Y' S
Source: Wu and Scheublein (1974).

strength as the s22 term becomes more compressive is seen again. Further discussion on this phenomenon is given in Section 1.20.2.3.3.

1.20.2.3.3

Discussion

Strength values for the graphiteepoxy material system. 1027.3 MPa 7710.2 MPa 43.4 MPa 7125.5 MPa 72.4 MPa

All six lamina strength criteria considered here are phenomenological or macromechanical in approach. In other words, they are more or less curve-fitting techniques. Except for the TsaiWu criterion, these criteria are developed based on certain assumptions of failure mechanisms. Thus, the accuracy of each criterion depends on whether the assumed dominant failure mechanism is included and properly described by the stresses or strains in the criterion.

Table 4 Strength values for material systems used in s22 7 t biaxial failure comparisons. AS4/55A Swanson et al. (1987) X X' Y Y' S not provided not provided 26.7 MPa 94.7 MPa 51.8 MPa X X' Y Y' S Scotch-Ply (Type 1002) Voloshin and Arcan (1980) 1108.0 MPa 617.8 MPa 19.61 MPa 137.30 MPa 36.92 MPa X X' Y Y' S T800/3900-2 Swanson and Qian (1992) not provided not provided 65.0 MPa 200.0 MPa 100.0 MPa

Lamina Failure Analysis

11

Figure 9 Comparison of lamina failure criteria to s22 7 t12 AS4/55A data from Swanson et al. (1987).

Figure 10 Comparison of lamina failure criteria to s22 7 t12 T800 data from Swanson and Qian (1992).

12

Strength Analysis of Unidirectional Composites and Laminates

Figure 11 Comparison of lamina failure criteria to s22 7 t12 glassepoxy data from Voloshin and Arcan (1980).

(i)

On the TsaiWu criterion

(ii)

On the maximum strain criterion

The TsaiWu criterion is the only one among the six criteria that strictly follows a curvefitting approach. Unlike the HillTsai criterion, the polynomial function used in the TsaiWu criterion cannot be interpreted as part of deformation energy because of the presence of linear terms. The simplicity in using a single equation to predict failure in a lamina under general loading as offered by TsaiWu is very attractive. However, in taking such an approach, all the failure mechanisms must be included simultaneously for any loading. This may present some conceptual difficulty. It is difficult to argue that, for example, failure of a composite under biaxial tension should depend on its compressive strength properties and vice versa. Although mathematically more convenient, a practice as adopted by TsaiWu may cause unreasonable failure predictions. As shown in Figure 1, TsaiWu suggests that a compressive stress s22 could increase the longitudinal strength of the composite. There are no known mechanistic reasons to support this. In fact, this phenomenon is produced by the fact that |Y '|>|Y| which causes translation of the failure ellipse to the said position.

On the s11 7 t12 and s22 7 t12 planes (see Figures 3 and 4), the Maximum Stress and Maximum Strain criteria predict identical results. However, for biaxial loading in the s11 7 s22 plane (see Figure 1), these two criteria differ significantly. The Maximum Strain criterion predicts that for a tensile longitudinal stress s11, the tensile transverse stress s22 would be greater than Y in order to fail the composite. Specifically, for s11 near X, the s22 required to cause failure approaches 2Y. If the transverse strength of the composite is controlled by the fiber/matrix interfacial strength, then this is not realistic. It is concluded that the Maximum Strain criterion is not adequate for predicting the transverse matrix cracking failure mode where a significant s11 is present.

(iii) Dependence of shear strength on compressive normal stress s22 In all the existing lamina failure criteria, the lamina strengths X, Y, and S are assumed to be constants. However, from the three s22 7 t12

Lamina Failure Analysis

13

Figure 12

Comparison of lamina failure criteria and the Modified Matrix Criterion to s22 7 t12 AS4/55A data from Swanson et al. (1987).

biaxial plots in Section 1.20.2.3.2 (Figures 9 11), there is strong evidence that when the composite is subjected to a combined s22 7 t12 loading, it becomes stronger when s22 is compressive. More specifically, for a given s22 = +s0, the shear stress t12 at failure corresponding to s22 = 7s0 is appreciably greater than the shear stress t12 corresponding to s22 = +s0. This behavior indicates that a compressive fiber/matrix interfacial normal stress (which is proportional to s22) could create a greater fiber/matrix interfacial shear strength. To reflect this behavior, the matrix failure criterion of Equation (5) may be modified to
s22 2 t12 2 1 Y S ms22  8

In the absence of s11, the linear stress terms in TsaiWu produce a failure envelope in the s227t12 plane that exhibits a characteristic similar to that of Equation (8). However, this effect would disappear in the TsaiWu criterion if |Y'|5|Y|. Figure 12 shows AS4/55A data from Figure 9 plotted with the TsaiWu criterion and the modified matrix failure criterion, Equation (8). It is clear from the comparison that the modified matrix criterion fits the data as well as TsaiWu if m = 0.6 is chosen. Figures 13 and 14 are simply replots of Figures 10 and 11, respectively, with the modified matrix failure criterion added. These plots also show that this modified matrix failure criterion can improve the strength prediction in the s22 7 t12 plane.

where
m m0 0 if s22 40 if s22 50 9

1.20.2.3.4

Concluding remarks

The term m plays a role similar to friction coefficients. Equation (8), denoted henceforth as the modified failure matrix criterion, still yields the expected values of s22 = Y at t12 = 0 and t12 = S at s22 = 0.

From the comparisons with experimental data, we conclude that a Separate Mode Failure criterion in the form
s11 X 22 2 s Y

1 2 t12 S ms22 1

for fiber failure for matrix failure

10

14

Strength Analysis of Unidirectional Composites and Laminates

Figure 13

Comparison of lamina failure criteria and the Modified Matrix Criterion to s22 7 t12 T800 data from Swanson and Qian (1992).

Figure 14 Comparison of lamina failure criteria and the Modified Matrix Criterion to s22 7 t12 glass epoxy data from Voloshin and Arcan (1980).

Laminate Strength Analysis is reasonably accurate for lamina failure prediction. A modified TsaiWu failure criterion,
s11 1 for fiber failure X F2 s22 F22 s22 2 2F12 s11 s22 F66 t12 1
2

15

11

for matrix failure

be accounted for in the laminate analysis. However, the local stress concentration effect due to matrix cracks is usually neglected except for laminates with thick laminae such as a [0/908/0] laminate. This local stress concentration effect on laminate strength was discussed by Sun and Jen (1987).

are also adequate for data fitting for most advanced composites (for which |Y'|>|Y|). 1.20.3.1 1.20.3 LAMINATE STRENGTH ANALYSIS Stiffness Reduction

Classical laminate strength analysis is based on the two-dimensional plane stress field in the laminate. Laminate failure is the eventual result of progressive failure processes taking place in the constituent laminae under loading. Conceptually, a ply-by-ply failure analysis should yield the desired failure load for the laminate. In reality, however, the failure mechanisms in laminates are a great deal more complicated than those in a unidirectional composite under inplane loading. New failure mechanisms are added to the three intralamina failure modes occurring at the lamina level. The most notable three-dimensional failure modes include delamination and failure induced by free edge singular stresses. Classical laminate strength analysis is restricted to those laminates whose failure is not dominated by three-dimensional failure modes. Another complication encountered in classical laminate failure analysis is that the lamina strengths in the laminate may be quite different from those obtained from the unidirectional composite panel. Such variations are attributed to the constraining effect from the adjacent laminae in the laminate. For better results from laminate failure analysis, the in situ lamina strengths should be used. The effects of free edge stresses are usually treated separately from classical laminate failure analysis. Thus, it is generally assumed that the laminate is either free of free edge stresses, or laminate failure does not initiate from the free edge. Some authors have utilized tubular specimens to avoid the effect of free edge stresses. The use of layers of film adhesive at the interlayers can also toughen the interface, forcing failure to occur in in-plane modes. This latter approach is taken in this study to enable the use of laminate coupon specimens for testing laminate strength. As lamina failure is progressive in nature, the progressive loss of lamina stiffness must also

Some of the laminate failure analysis methods consider a laminate capable of load bearing after an individual ply within the laminate has failed. These methods require a procedure for discounting the failed ply and reducing the laminate stiffness. Two methods for achieving this are as the parallel spring model and the incremental stiffness reduction model.

1.20.3.1.1

Parallel spring model

Each lamina is modeled with a pair of springs representing fiber (longitudinal) and matrix (shear and transverse) deformation modes. The entire laminate is modeled by grouping together a number of parallel lamina spring sets as shown in Figure 15. When fiber breakage occurs, the longitudinal modulus is reduced. When matrix cracking occurs, the shear and transverse moduli are reduced. Various approaches and models have been proposed to investigate stiffness reduction caused by matrix cracking in a laminate (for references, see Tao and Sun, 1996). In fiber dominated laminates, such laminate stiffness reductions are usually small, and, for the sake of simplicity, the matrix dominated moduli, E2 and G12, are usually set equal to zero. This model is also capable of differentiating between types of matrix failure if desired; i.e., the transverse and shear moduli can be reduced separately depending on the specific type of matrix failure mode. The model which reduces E1 for fiber failure and E2 and G12 for either transverse or shear matrix failure is denoted the PSM. The model which reduces E1 for fiber failure, E2 for transverse matrix failure, and E2 and G12 for shear matrix failure is denoted the PSMs. The idea behind the PSMs is that a transverse matrix failure does not necessarily inhibit the ability of the lamina to carry significant shear loads. Creating these two different reduction models has little micromechanical basis, and is done mainly for curve fitting purposes.

16

Strength Analysis of Unidirectional Composites and Laminates

Figure 15 Schematic of the parallel stiffness model.

1.20.3.1.2

Incremental stiffness reduction model

To avoid the sudden jump in strain at ply failure seen in the parallel spring model, a model resembling the bilinear hardening in classical plasticity can be formulated. Laminate stiffness reduction is achieved similar to that in the parallel spring model. However, it is assumed that the reduced laminate stiffness governs only the incremental loaddeformation relations beyond immediate ply failure. Both of these stiffness reduction models have flexibility. Instead of reducing the appropriate moduli suddenly after a ply failure, a nonlinear function such as an exponential function may be used to gradually reduce these values. This progressive softening approach may model certain laminates better than others, i.e., those laminates whose failure is dominated by matrix cracking. For most fiber-dominated composites, setting the stiffness constants directly to zero after the corresponding mode of failure occurs is simple and unambiguous. The use of such reduction can be justified by regarding the laminate analysis to be at the location where matrix cracking occurs. Consider a 908 lamina (within a laminate) containing a number of transverse matrix cracks, as shown in Figure 16. The 908 ply still retains some stiffness in the loading direction (E2 direction locally). However, the assumption is made that ensuing 08 fiber failure will occur at the weakest point. This point is where matrix cracking has occurred in the 908 plies, or where locally E2 = 0. Thus, it is acceptable in the ultimate

strength analysis to reduce E2 directly to zero after transverse matrix cracking. Since matrix cracks are discrete, between two cracks a failed lamina would still contribute substantially to laminate stiffness. It is obvious that such drastic lamina stiffness reduction, if assumed to be true over the entire laminate, would overestimate the ultimate strains of the laminate. In fiber-dominated laminates, the effect of matrix cracks on overall laminate stiffnesses is usually very small. It is reasonable to estimate the laminate ultimate strains by using the virgin laminate stressstrain relations and the laminate failure stresses obtained from the laminate failure analysis.

1.20.3.2

Laminate Failure Analysis Methods

As with lamina failure analysis, a variety of laminate failure analysis methods have been proposed. The following is a description of each methodology.

1.20.3.2.1

Ply-by-ply discount method

This is a very common method for laminate failure analysis. Laminate is treated as a homogeneous material and is analyzed with a lamina strength criterion. The laminated plate theory is used to initially calculate stresses and strains in each lamina. A lamina strength criterion is then used to determine the particular ply which will fail first and the mode of that failure. A stiffness reduction model is used to reduce the stiffness

Laminate Strength Analysis

17

Figure 16 Schematic of laminate with matrix cracks.

of the laminate due to that individual ply failure. The laminate with reduced stiffnesses is again analyzed for stresses and strains. The lamina failure criterion predicts the next ply failure, and laminate stiffness is accordingly reduced again. This cycle continues until ultimate laminate failure is reached. A number of definitions have been proposed on how to determine ultimate laminate failure. One common way is to assume ultimate laminate failure when fiber breakage occurs in any lamina. Another way is to check if excessive strains occur (i.e., yielding of the laminate stiffness matrix). Matrix-dominated laminates such as [+45]s may fail without fiber breakage. Others have suggested a last ply definition in which the laminate is considered failed if every ply has been damaged. For this study, laminate failure is defined as fiber breakage in any ply or the reduced stiffness matrix becomes singular.

1.20.3.3

Laminate Failure Analysis for Biaxial Loading

The six lamina strength criteria are used in conjunction with the above laminate failure analysis methods to predict laminate strength for biaxial loading. In view of the limitations of the present method, laminates whose failure is caused by free edge stresses or delamination are avoided. The ply-by-ply discount method is used in conjunction with the parallel spring model (PSM) for laminate stiffness reduction. In this analysis, the appropriate lamina moduli reduce to zero at the individual ply failures. The laminate is assumed to reach ultimate failure when any ply within the laminate fails by fiber breakage.

1.20.3.3.1

Comparison with data for biaxial loading

1.20.3.2.2

Sudden failure method

In highly fiber-dominated composite laminates, laminate stiffness reduction due to progressive matrix failures insignificantly affects laminate ultimate strength. This suggests that in such laminates the progressive stiffness reduction seen in the previous method may be unnecessary, and laminate failure may be taken to coincide with fiber failure of the load-carrying ply. To perform this analysis, a lamina strength criterion is chosen, and the failure load is determined by calculating the load required for fiber failure in the dominant lamina. No stiffness reductions are included in the process. The laminate strength predicted by the sudden failure method is usually higher than the laminate strength predicted by the ply-byply discount method.

Recently, Swanson and Trask (1989) performed biaxial testing on an AS4/3501-6 graphiteepoxy [0/+45/90]s laminate using tubular specimens. The ply properties given by Swanson and Trask (1989) are listed in Table 5. Note that they are slightly different from those given in Table 1. Figure 17 shows the failure envelopes in the sxx 7 syy plane. The predicted envelopes are
Table 5 Moduli and strength values AS4/3501-6 graphiteepoxy system. 134.60 GPa 11.03 GPa 5.52 GPa 0.28 0.13 mm X X' Y Y' S of

E1 E2 G12 v12 Ply thickness

1986.0 MPa 1193.0 MPa 47.9 MPa 168.0 MPa 95.7 MPa

Source: Swanson and Trask (1988).

18

Strength Analysis of Unidirectional Composites and Laminates

Figure 17

Comparison of ultimate stress envelopes with experimental data, Swanson and Trask (1989) for a [0/+45/90]s laminate under biaxial loads.

generated with the six lamina failure criteria in conjunction with the ply-by-ply discount method. Comparing with the data, it is evident that the predictions of all six lamina failure criteria agree with the data very well. However, the fully interactive criteria such as HillTsai and TsaiWu are very sensitive to the sequence of lamina matrix failures and can be significantly affected by the sudden reduction of matrix stiffness resulting in jumps in strength. These jumps are evident in Figure 17. Swanson and Qian (1992) also tested laminate tubes in the sxx 7 syy plane. The [0/+45/90]s, [03/+45/90]s, and [0/(+45)2/90]s strength data (I and IV quadrants only) were found to match well with the Maximum Strain analysis. Considering that all lamina failure criteria are in close agreement in these quadrants for the p/4 laminate, it is clear that they should all be adequate for these laminates.

1.20.3.3.2

Biaxial failure in the strain plane

The failure strains corresponding to the failure stress envelopes of Figure 17 are plotted in

Figure 18. It is interesting to note that the failure strain envelopes predicted by Maximum Stress, Maximum Strain, Hashin, and Hashin Rotem criteria essentially coincide with the limits set by the ultimate tensile strain (Xe) and compressive strain (Xe') of the unidirectional composite. However, these limit strains should not be automatically taken as the ultimate strains of the laminate at failure. For the p/4 quasi-isotropic laminate under biaxial loading in the third stress quadrant (compressive sxx and syy), there are no ply matrix failures before ultimate laminate failure. Thus, there is no stiffness reduction before laminate failure, and the calculated failure strain is the laminate failure strain. Under tensile biaxial loading (tensile sxx and syy), all plies in the laminate have suffered matrix failure before laminate final failure. The drastic reduction (to zero) of ply transverse and shear stiffnesses tends to slightly overestimate the ultimate strains. Therefore, the strain failure envelope as shown in Figure 18 should not be used directly in laminate strength design without accounting for the stiffness reductions.

Laminate Strength Analysis

19

Figure 18 Comparison of ultimate strain envelopes with experimental data, Swanson and Trask (1989) for a [0/+45/90]s laminate under biaxial loads.

1.20.3.3.3

Biaxial testing data for glass woven fabric composite

Recently, Wang and Socie (1992, 1994) performed biaxial testing on NEMA G-10 E-glass plain woven fabric composite laminates. They used both tubular specimens and flat square specimens. Strength data were obtained for all four biaxial loading quadrants. Their test results clearly indicate that the laminate failure strain envelope is bounded by the uniaxial ultimate strains of the laminate. Specifically, they found that failure in one direction was not affected by loading in the transverse direction. Hence, the failure strain envelope appears to be rectangular as predicted by the Maximum Strain and Maximum Stress criteria.

interfaces of laminate coupon specimens, it is possible to suppress these three-dimensional effects. This type of coupon specimen is used to generate additional laminate strength data for the purpose of evaluating the strength criteria.

1.20.3.4.1

Selection of laminates and off-axis loading angles

1.20.3.4

Laminate Strength Analysis for Unidirectional Loading

Laminate strength data are available for coupon specimens under uniaxial loading. In most cases though, free edge stresses control the initiation and final failure of these laminates. Data as such are not suitable for evaluating the laminate failure analysis methods as attempted here. However, by placing film adhesive at the

A variety of laminates was theoretically examined in advance of actual strength testing. Only the ply-by-ply discount method was used to determine ultimate laminate strength. Laminate layups and off-axis loading angles were chosen to provide a comparison among the six lamina strength criteria. The focus of the experiment was to examine whether these strength criteria could predict the correct trend of laminate strength vs. loading angle. Table 6 shows the laminates and off-axis loading angles selected. Included in these tests were unidirectional laminate specimens for principal material properties. Five or more tests were performed at each off-axis loading angle to provide accurate mean results. In the table, A indicates the location of an adhesive film in the laminate.

20

Strength Analysis of Unidirectional Composites and Laminates

Table 6 Laminates and off-axis loading angles tested. Laminate [0]8 [0/A/+45/A/745/A/90]s [90/A/0/A/90/A/0]s [0/A/+45/A/745]s [90/A/+30/A/730]s Off-axis angles tested 0 8 and 90 8 0 8 22.5 8 every 7.5 8 0 8 7.5 8 every 1.5 8, 15 8, 22.5 8 0 8 30 8 every 7.5 8, 26 8, 458 0 8 22.5 8 every 7.5 8

1.20.3.4.2

Laminate coupon specimens and oblique end tabs

wide with a 171 mm gage section. This yielded a 9:1 aspect ratio. 1.20.3.4.3 Results and data analysis

The material used for testing was AS4/3501-6 graphiteepoxy from Hercules. A 0.13 mm thick ply of film adhesive was added at each lamina interface except the middle interface because of symmetry. The adhesive film was FM 1000 marketed by American Cyanamid. The elastic and strength properties of this material (FM 1000) were determined by Sun and Zhou (1988) and are listed in Table 7. For those laminate specimens with anisotropic stiffness, it was important not to overlook the effect of shearextension coupling. Highly anisotropic laminate coupon specimens can fail prematurely due to stress concentrations in the tab region if rectangular tabs are used. In order to accommodate the deformation induced by the extensionshear coupling, oblique end tabs suggested by Sun and Chung (1993) were used. The oblique angle f (see Figure 19) was derived using extensional laminate stiffnesses A11 and A16 as
cotf A16 =A11 12

Tests were performed in the Composite Materials Laboratory (CML) at Purdue University. The first set of coupons tested were 08 and 908 unidirectional laminates. All specimens were strain gauged to determine the appropriate elastic constants. Table 8 lists the results. Only the moduli E1, E2, n12, and the strength values X and Y were determined in the present test. The strengths S, X', and Y' were taken from a similar AS4/3501-6 material system given by Sun and Zhou (1988). The shear modulus G12 in Sun and Zhou (1988) was obtained from testing [+45]s laminate and is lower than 6.9 GPa given by Daniel and Ishai (1994). The latter value is listed in Table 8. For laminates containing layers of film adhesive, the calculation of ultimate stress must account for the thickness and stiffness of the

It was demonstrated by Sun and Chung (1993) that, using oblique end tabs, an almost uniform state of stress corresponding to uniaxial loading could be generated in off-axis laminate coupon specimens rigidly gripped during loading. In quasi-isotropic laminates, extensionshear coupling is absent, and standard rectangular tabs were used. Glassepoxy and graphiteepoxy quasi-isotropic laminates were used for oblique end tabs. For the rectangular tabs, material stiffness is not as critical, and chopped fiberglass circuit board was used. All specimens were 19 mm
Table 7 Material properties for FM 1000 film adhesive. sult E G v Ply thickness 38 MPa 1.724 GPa 0.648 GPa 0.33 0.127 mm

Figure 19 Example of oblique end tabs.

Laminate Strength Analysis

Table 8 Moduli and strength values for the tested AS4/3501-6 graphiteepoxy system. E1 E2 G12 v12 Ply thickness 153.7 GPa 11.0 GPa 6.9 GPa 0.32 0.13 mm X X' Y Y' S 2171.0 MPa 2013.0 MPa 67.0 MPa 206.8 MPa 110.3 MPa

21

Table 9 Ultimate laminate stresses (MPa). Off-axis loading angle [0/A/+45/A/45/A/90]s [0/A/+45/A/45]s [90/A/+30/A/30]s Off-axis loading angle [90/A/+30/A/30]s 08 765 883 966 08 1126 7.5 8 752 843 908 1.5 8 1140 15 8 774 929 837 38 1074 22.5 8 832 1028 807 4.5 8 1018 26 8 1129 68 861 30 8 1074 7.5 8 713 45 8 818 15 8 394 22.5 8 288

film adhesive in determining the true graphite epoxy laminate strength. First, it is assumed that the total measured ultimate load is the summation of the load carried by the composite, PC, and the load carried by the adhesive, PA, i.e.,
Pexp PC PA 13

reaching a strain of approximately 3% at which the test was stopped. As seen in Figure 20, a [90/A/0/A/90/A/0]s laminate loaded at 22.58 exemplifies this behavior. For comparison, the stress7strain curve for a [0/A/+45/A/45]s laminate loaded at 458 is also plotted, which demonstrates the typical behavior involving fiber failure. 1.20.3.4.4 Comparison with test data

For such laminates under uniaxial loads, it is reasonable to assume that the composite and adhesive loads can be separated by the rule of mixtures:
 PA  PC  EA hA Pexp A11 EA hA 14

 A11 Pexp A11 EA hA

15

where EA is the adhesive modulus, hA is total thickness of all adhesive layers in the laminate, and A11 is the total extensional stiffness calculated just for the graphiteepoxy plies. Equation (15) is used to obtain the true composite ultimate load from the experimental load. The laminate strength is determined by dividing PC by the total cross-sectional area of the composite plies. Table 9 lists all the averaged ultimate stress data for the tested laminates. Though the majority of the laminates failed by fiber breakage, a few laminates in which fiber failure did not occur were matrix dominated. In the matrix dominated coupon tests, the specimen never fractured into two (or more) pieces. Instead, the deformation continued until

Theoretical predictions using the six lamina failure criteria in conjunction with the ply-byply discount method for laminates under offaxis loading are compared with the experimental data. Only the results obtained with the PSM stiffness reduction method are reported here. Results with the PSMS method can be found in the report by Sun et al. (1996). Figure 21 shows theoretical predictions for the [0/A/+45/A/745/A/90]s laminate obtained using the PSM stiffness reduction procedure. The experimental data (CML data) for the [0/A/+45/A/745/A/90]s laminate shows an increase in strength as the off-axis loading angle rotates from 08 to 22.58. Maximum Stress, Maximum Strain, and HashinRotem all predict this increase, supporting the idea of separating fiber failure (governed by fiber stress s11) from matrix failure (governed by matrix stresses s22 and t12). Figure 22 compares CML data with the theoretical predictions for the [90/A/0/A/90/A/0]s laminate. The fully interactive criteria HillTsai and TsaiWu underestimate the strength of the laminate. The Hashin criterion is closer to the

22

Strength Analysis of Unidirectional Composites and Laminates

Figure 20 Stressstrain curves characterizing fiber and matrix failures.

Figure 21 Comparison of ultimate strengths for a [0/A/+45/A/745/A/90]s laminate under unidirectional loading using different lamina failure criteria.

data than the fully interactive criteria. Again, the criteria (Maximum Stress, Maximum Strain, and HashinRotem) which separate fiber failure from matrix failure best match the data. Data from the [0/A/+45/A/745]s laminate are displayed in Figure 23. Overall, Maximum Stress, Maximum Strain, and HashinRotem are clearly the best fit for the data. In the off-

axis region from 08 to 208, the TsaiWu criterion closely fits the data. However, the overall trend of TsaiWu's prediction is quite different from the experimental data. Data from the [90/A/+30/A/730]s laminate are displayed in Figure 24. Once again, Maximum Stress, Maximum Strain, and Hashin Rotem match both the magnitude and the trend of the data.

Laminate Strength Analysis

23

Figure 22

Comparison of ultimate strengths for a [90/A/0/A/90/A/0]s laminate under unidirectional loading using different lamina failure criteria.

Figure 23

Comparison of ultimate strengths for a [0/A/+45/A/745]s laminate under unidirectional loading using different lamina failure criteria.

1.20.3.5

Discussion

From the four different sets of laminate strength data presented in the previous section, it appears that the interactive criteria (HillTsai and TsaiWu) and Hashin criterion (which couples s11 and t12) not only underestimate ultimate laminate failure, but cannot correctly

predict the trend of the data. On the other hand, Maximum Stress, Maximum Strain, and HashinRotem criteria all perform quite well. Simply put, those criteria which separate fiber failure completely from matrix failure are relatively insensitive to inaccurate lamina strengths Y and S. There is some matrix strength sensitivity in these criteria from the

24

Strength Analysis of Unidirectional Composites and Laminates

Figure 24 Comparison of ultimate strengths for a [90/A/+30/A/730]s laminate under unidirectional loading using different lamina failure criteria.

Figure 25 Comparison of ultimate strengths for [0/A/+45/A/745/A/90]s with different lamina shear strengths using HillTsai Criterion.

effects of intermediate ply failures (failure preceding ultimate fiber failure). However, the fully interactive criteria and the Hashin criterion are considerably more sensitive to inaccurate matrix strengths. This can be illustrated by varying the lamina matrix strengths used in the theoretical analysis of the laminates.

The sensitivity of the interactive strength criteria to matrix strength (Y or S) is demonstrated by the ultimate strength curves for the [0/A/+45/A/745/A/90]s laminate in Figure 25. The HillTsai strength criterion is employed for the laminate strength analysis using three different lamina shear strength values. The curve

Conclusions

25

Figure 26

Evidence of 90 8 ply matrix cracking in [0/90/0] and [0/90/0]s laminates of AS4/3501-6 graphite epoxy composite.

HillTsai with S*1.0 is the same as the Hill Tsai curve from Figure 21. The other two curves in Figure 25 are for the shear strengths 2 6 S and 2.5 6 S, respectively, with all other material constants kept the same. It is clear that by simply increasing the shear strength of the lamina, the HillTsai criterion takes on a completely different trend, predicting an increase in laminate strength instead of a decrease as the off-axis loading angle increases from 08 to 22.58. With these high shear strengths, the HillTsai criterion now correctly predicts the trend and magnitude of the experimental data. Another issue often raised is whether matrix cracking actually occurs in composite laminates with well-dispersed laminae. To answer this question, we tested [0/90/0] and [0/90/0]s laminates of AS4/3501-6 graphiteepoxy composite. Coupon specimens were tested under tension. Figure 26 clearly shows the presence of matrix cracks in the 908 plies at the load about 95% of the laminate strength. Thus, we conclude that matrix cracking does occur and ply-by-ply discount in laminate failure analysis is justified.

1.20.4

CONCLUSIONS

Six commonly used composite strength criteria were investigated and compared for applications in unidirectional fiber composites and their laminates. These strength criteria were evaluated using lamina and laminate strength data which do not involve any outof-plane deformation failure modes such as delamination. Moreover, all the experimental data used in the evaluation were obtained from specimens under a state of uniform stress; i.e., no stress gradient effects were included. Based on the foregoing environment, the following conclusions have been obtained. (i) At the lamina level, those criteria (such as the HashinRotem criterion) which separate the fiber failure mode from the matrix failure mode are the most reasonable and accurate. (ii) For fiber-dominated laminates, Maximum Stress, Maximum Strain, and Hashin Rotem failure criteria outperform other criteria. These criteria are insensitive to variations in matrix strengths (Y and S) which are very difficult to obtain in situ.

26

Strength Analysis of Unidirectional Composites and Laminates

Z. Hashin and A. Rotem, J. Composite Materials, 1973, 7, 448464. P. Labossiere and K. W. Neale, Solid Mechanics Archives, 1987, 12(2), 6595. M. N. Nahas, J. Composites Technology & Research, 1986, 8(4), 138153. R. Narayanaswami and H. M. Adelman, J. Composite Materials, 1977, 11, 366377. R. B. Pipes and B. W. Cole, J. Composite Materials, 1973, 7, 246256. C. T. Sun and I. Chung, Composites, 1993, 24(8), 619 623. C. T. Sun and K. C. Jen, J. Reinforced Plastics and Composites, 1987, 6, 208222. C. T. Sun, B. J. Quinn, J. Tao and D. W. Oplinger, US Department of Transportation Federal Aviation Administration Report No. DOT/FAA/AR-90/1-9, 1996. C. T. Sun and S. G. Zhou, J. Reinforced Plastics and Composites, 1988, 7, 515557. S. R. Swanson, M. J. Messick and Z. Tian, J. Composite Materials, 1987, 21, 619630. S. R. Swanson and B. C. Trask, Composites, 1988, 19, 400406. S. R. Swanson and B. C. Trask, Composites Science and Technology, 1989, 34, 1934. S. R. Swanson and Y. Qian, Composites Science and Technology, 1992, 43, 197203. J. X. Tao and C. T. Sun, Mechanics of Composite Materials and Structures, 1996, 3, 225239. S. W. Tsai, `Strength Characteristics of Composite Materials', NASA CR-224, National Aeronautics and Space Administration, 1965 April. S. W. Tsai and E. M. Wu, J. Composite Materials, 1971, 5, 5880. A. Voloshin and M. Arcan, Experimental Mechanics, 1980, 20, 280284. J. Z. Wang and D. F. Socie, in `Composite Materials: Testing and Design', ASTM STP 1206, vol. 1, American Society for Testing and Materials, Philadelphia, PA, 1992, pp. 136149. J. Z. Wang and D. F. Socie, J. Composites Technology & Research, 1994, 16(4), 336342. E. M. Wu and J. K. Scheublein, in `Composite Materials: Testing and Design 3rd Conference', ASTM STP 546, ASTM, Philadelphia, PA, 1974, pp. 188206.

(iii) The interactive failure criteria (Hill Tsai, TsaiWu, and Hashin) are sensitive to variations of the matrix-dominated lamina strengths (i.e., Y and S). Accurate in situ composite strengths are critical to the use of these criteria. Because of the interaction among all stress components, sudden switching of failure modes makes the failure envelope (in stress or strain) very jumpy. (iv) To predict lamina matrix failure in a laminate, the in situ transverse strength (Y) and shear strength (S) should be used. In addition, thermal residual stresses must be included in the laminate analysis. (v) Experimental results indicate that matrix cracking does take place even in laminates with well-dispersed laminae. Thus, the ply-by-ply discount of stiffnesses in failed laminae is justified. (vi) The Parallel Spring model for stiffness reduction is adequate for analysis of laminate strength. The drastic ply stiffness reduction (the concerned stiffness is set equal to zero after ply failure) does not cause appreciable errors in the predicted laminate strength for fiber-dominated laminates.

1.20.5

REFERENCES

V. D. Azzi and S. W. Tsai, Experimental Mechanics, 1965, 5, 283288. G. E. Colvin and S. R. Swanson, J. Eng. Mat. Technol., 1990, 112, 6167. W. C. Cui, M. R. Wisnom and M. Jones, Composites, 1992, 23, 158166. I. M. Daniel and O. Ishai, `Engineering Mechanics of Composite Materials', Oxford University Press, New York, 1994. Z. Hashin, J. Appl. Mechanics, 1980, 47, 329334.

Copyright # 2000 Elsevier Science Ltd. All rights reserved. No part of this publication may be reproduced, stored in any retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise, without permission in writing from the publishers.

Comprehensive Composite Materials ISBN (set): 0-08 0429939 Volume 1; (ISBN: 0-080437192); pp. 641666