The Effect of Particle System on Quasistatic Compression Behavior and Damage Accumulation of High Solids Loading Polymer-Particle Composites

This paper studies the effect of particle system on the quasistatic compression behavior and damage accumulation of high solids loading polymer-particle composites. To isolate the effect of particle system, two binary polymer-particle composites are fabricated with either crystalline silica sand or sodium chloride at 80% nominal particle weight fractions in a polydimethylsiloxane matrix. The particle size distributions are controlled with a median particle size of 312 +/- 12 micrometers. Near fully-dense right circular composite cylinders with nominal diameter of 25.4 mm and nominal height of 18.8 mm are compressed under monotonic and cyclic loadings in a uniaxial, unconfined configuration. Under monotonic loadings, the compression modulus of the sodium chloride composites are nearly 7 times larger than the compression modulus of the silica sand composites. Novel experiments performing cyclic loadings at small strains assess material damage through changes in peak stresses and apparent stiffness with strain and cycle number. Both composites reached a decreasing peak stress with cycle number at the same strain suggesting damage similarly initiates at low strains and accumulates with repeat loadings. The bulk composite strength and damage observations are discussed in the context of single-particle fracture strength estimated from a pressure-volume analysis of their uniaxial confined compression behavior. The increased bulk compressive strength of the sodium chloride composite despite the lower estimated particle strength suggests that particle shape has a larger effect on low-strain compression behavior than particle strength in these high-solids-loading composites. Quantitative segmentation of X-ray micro-computed tomographic images using an unsupervised k-means clustering algorithm into polymer and particle spatial distributions observes that damage initiates at the sample boundaries. It is presented that the unsupervised segmentation of the low-contrast X-ray images quantitatively agrees with measurements of ingredient mass fractions taken during fabrication and is largely insensitive to typical image pre-processing steps of image sharpening and filtering.