Variability in Particle Size Distribution Due to Sampling

Abstract

Abstract Sampling particulate matter for particle size distribution (PSD) analysis is a task routinely performed in geotechnical and geoenvironmental engineering. Pitard (2019) mentions that “for a sample to be representative of anything, the first rule to fulfill is to ensure that the sample is representative of all the particle size fractions.” Several sampling techniques exist for obtaining samples of particulate matter from a lot, but their representativeness has rarely been documented, either experimentally or theoretically. To this end, this article studied the representativeness of four sampling techniques applied to moist and dry particulate matter, namely riffle splitting, fractional shoveling, 2-dimensional incremental sampling (2-DIS), and grab sampling. Bias being small because of experimental design, relative variance was used to assess sampling performance. Except for the largest size fraction (>9.5 mm), for which all sampling techniques gave poor results because of insufficient sample mass, riffle splitting was the most reproducible technique (CV = 6.47 %) and showed the smallest increase in variability compared to the fundamental relative sampling variance (i.e., a CV increase of 0.66 %), followed by fractional shoveling (7.68 %, 2.59 %), grab sampling (11.7 %, 6.51 %), and 2-DIS (16.3 %, 11.1 %). For fractional shoveling, sampling dry matter (CV = 19.2 %) significantly increased sampling variability compared to moist matter by 11.5 %. Furthermore, theoretical estimation of minimum sample mass requirements showed that mass requirements in ASTM D6913/D6913M-17, Standard Test Methods for Particle-Size Distribution (Gradation) of Soils Using Sieve Analysis, can lead to larger sampling variance than expected. Guidelines for specimen procurement ASTM D6913/D6913M-17 were also analyzed and judged insufficient with respect to fundamental sampling principles.