Failure property distributions for conventional and highly crosslinked ultrahigh molecular weight polyethylenes

To make stochastic (probabilistic) failure predictions of a conventional or highly crosslinked ultrahigh molecular weight polyethylene (UHMWPE) material, not only must a failure criterion be defined, but it is also necessary to specify a probability distribution of the failure strength. This study sought to evaluate both parametric and nonparametric statistical approaches to describing the failure properties of UHMWPE, based on the Normal and Weibull model distributions, respectively. Because fatigue and fracture properties of materials have historically been well described with the use of Weibull statistics, it was expected that a nonparametric approach would provide a better fit of the failure distributions than the parametric approach. The ultimate true stress, true strain, and ultimate chain stretch data at failure were analyzed from 60 tensile tests conducted previously. The ultimate load and ultimate displacement from 121 small punch tests conducted previously were also analyzed. It was found that both Normal and Weibull models provide a reasonable description of the central tendency of the failure distribution. The principal difference between the Normal and Weibull models can be appreciated in the predicted lower-bound response at the tail end of the distribution. The data support the use of both parametric and nonparametric methods to bracket the lower-bound failure prediction in order to simulate the failure threshold for UHMWPE.