Ultimate failure generally occurs after saturation of matrix cracking. The fibers break when the applied load is close to the maximum. Matrix damage and ultimate failure thus appear to be successive phenomena.

The ultimate failure of a tow of parallel fibers involves two steps:

• a first step of stable failure and

• a second step of unstable failure.

During the first step, the fibers fail individually as the load increases. In the absence of fiber interactions, the load is carried by the surviving fibers only (equal load sharing). Fiber interactions cause tow weaken­ing. The ultimate failure of a tow (second step) occurs when the surviving fibers cannot tolerate the load increment resulting from a fiber failure. At this stage, a critical number of fibers have been broken.

The ultimate failure of a longitudinal tow coated with matrix also involves a two-step mechanism and global load sharing when a fiber fails. In the presence of multiple cracks across the matrix and associated interface cracks, the load-carrying capacity of the matrix is tremendously reduced or eliminated. The matrix-coated tows behave like dry tows subject to the typical stress field generated by the presence of matrix cracks. The ultimate failure of a matrix-coated tow occurs when a critical number of fibers have failed. This mechanism operates in the tows within textile CVI SiC/SiC composites. The ultimate failure of the composite is caused by the failure of a critical number of broken tows (>1) depending on the stress state: ~1 under an axial tension, >1 in bending.

It is worth pointing out that the failure mechanism ofCVI SiC/SiC composites differs from that observed in polymer matrix impregnated tows, where local load sharing prevails when a fiber fails. In these composites, the fibers fail first. Therefore, the uncracked matrix is able to transfer the loads.

2.12.6.2 Reliability

The ultimate failure of CVI SiC/SiC composites is highly influenced by stochastic features. As fibers are brittle ceramics, they are sensitive to the presence of flaws (stress concentrators) that are distributed randomly. As a consequence, the strength data exhibit significant scatter, as illustrated by Figure 7.39’40 The figure shows that the magnitude of the strength and scatter decrease from single fibers to tows, then to infiltrated tows, and finally to woven composites.

0. 012 0.01 0.008

0­0.006

M

0.004

Q

0.002 0

0 500 1000 1500 2000 2500

Stress (MPa)

Figure 7 Strength density functions for SiC fibers (NLM 202), SiC fiber tows, SiC/SiC (1D) minicomposites, and 2D SiC/SiC composites.

As a result of the previously mentioned two-step failure mechanism, the ultimate failure of an entity is dictated by the lowest extreme of the strength distri­bution pertinent to its constituent: that is, tows versus filaments, infiltrated tows versus fibers, and 2D com­posites versus infiltrated tows. The lowest strength extremes correspond respectively to the critical num­ber of individual fiber breaks («17% for the SiC Nicalon™ fibers and for the SiC Hi-Nicalon™ fibers) and to the critical number of tow failures (>1). The gap between tows and SiC infiltrated tows results from the method ofstrength determination: the critical number of individual fiber breaks was taken into account for tow strength determination, whereas the strength of infiltrated tows and composites was under­estimated because the total cross sectional area of the specimens was used.

The flaw populations are truncated during the successive damage steps, which leads to a homoge­neous ultimate population of flaws.40 This process of progressive elimination of flaws governs the trends in the ultimate failure. The tensile stress-strain curves obtained on a batch of several CVI SiC/SiC test specimens coincide quite well (Figure 5), whereas the strength data exhibit a certain scatter (Figure 5). This scatter is limited (Figure 8). Dependence of composite strength on the stressed volume is not significant (Figure 8). Furthermore, dependence on the loading conditions is not so large (Figure 9): for instance, the flexural strength is 1.15 times as large as the tensile strength40,41 when measured on specimens having comparable sizes (Figure 9).

The Weibull model is not appropriate to describe the volume dependence of strength data,40 as the weakest link concept is violated. However, the

1

Ulti

permission from Springer.

Weibull modulus (m) can be extracted from the sta­tistical distribution of the strength data: m is in the range of 20-29. This value provides an evaluation of the scatter in strength data. It reflects a small scatter.