Bimodal impact strength distribution in HDPE composites: Formulational effects and Bayesian inference

A bimodal impact strength distribution was found in notched impact specimens of HDPE composites with calcium carbonate, carbon black, SEBS and stabilisers. The bimodal distribution was only found at moderate-to-high calcium carbonate loadings, with the likelihood of low impact strength increasing with increasing stabiliser loading and decreasing with increasing SEBS/CB masterbatch loading. Bayesian methods were used first to confirm bimodality and then to investigate the effects of formulation on the performance of the system, based on a hierarchical model with quadratic and interactive terms and switching based on the sampling of a Bernoulli distribution with a logistic regression informing the probability of high or low impact strength. The results are contextualised through micrograph fractography and, briefly, differential scanning calorimetry. Results are also reported for unnotched impact tests, with negative correlations for impact strength with calcium carbonate and stabilisers, a positive correlation with SEBS/CB and interactive effects.