A method is presented by which fatigue life predictions of polymer matrix composites may be improved. First a "critical element", whose failure defines global failure of the material. is identified. The global stiffness changes of the specimen during a fatigue test are monitored and taken to be inversely proportional to the increase in applied stress on the critical element. Using a cumulative damage model, the complicated stress history of the critical element is reduced to a "critical element SN curve", which defines the fatigue response of the critical element. The residual strength of the critical element may then be continuously evaluated to predict failure. The statistical nature of material strength is accounted for by forcing the critical element SN Curve to yield a specimen initial strength distribution in fatigue which is equivalent to the quasi-static tensile distribution. In contrast to most methods, the predictions are based on the stiffness history of the specimen in question, rather than on generalized phenomenological models. The critical element SN curve is then applied to (90°/0°) crossply materials to evaluate their fatigue response. Simulations and the variation of experimental parameters are examined for their effect on the predictions.

The unidirectional fatigue predictions were vastly improved over the traditional SN Curve. While the crossply predictions were not as good, they still demonstrated the applicability of the critical element SN curve to a material with a different geometry. Additionally. such a method may have application in real-time durability evaluation of composite components.