Representative sampling and use of HHXRF to characterize lot and sample quality of quartzite at a pyrometallurgical ferrosilicon plant

Material sampling is a critical component in mining and mineral processing industries. Nonetheless, sampling is often considered to be a simple matter and, as such, non-rigorous sampling protocols are often applied. The use of inappropriate methods produces inferior, non-representative estimates of sampling target composition. To address weaknesses in sampling protocols and evaluate the representativeness of collected samples, we performed a feasibility study of the ability of handheld X-ray fluorescence (HHXRF) to achieve a satisfactory characterization of a raw material lot at a pyrometallurgical ferrosilicon plant. Using composite and grab samples, we determined the various sampling error manifestations stemming from the fundamental sampling error, grouping and segregation error, as well as increment delimitation, increment extraction, and increment preparation errors), and performed a first foray determination of optimal sample mass, and estimated the heterogeneity within the sampling target. HHXRF results were compared with the results obtained using laboratory XRF. A first estimate of optimized sample mass for HHXRF was 10 kg, given the large size of crushed quartz blocks used in ferrosilicon plants—roughly cubic, 10 cm per side; accuracy improved with increased sample mass (18% error with a 10 kg sample versus 35% error when using a 1 kg sample). A 10 kg sample is also the mass a technician can realistically transport from the sampling site to the preparation facilities. The main contribution to the global estimation error is from primary sampling. Variographic analysis illustrated a sill equal to the nugget effect, indicating that two adjacent samples are no more similar than two samples separated by larger distance; this suggests equal spatial heterogeneity at all scales larger than the increment mass in the sampling target. Analytically, the HHXRF and desktop XRF results compared very well. Overall, the error associated with our first attempt at field composite sampling was half of that obtained via grab sampling for both the HHXRF and desktop XRF protocols. Relative to conventional analysis based on grab sampling and analysis via desktop XRF, the use of handheld XRF coupled with composite sampling would appear to be a feasible approach for an improved sampling protocol for obtaining fit-for-purpose characterizations of industrial quartzite.