Understanding sampling variation – a vital aspect of industrial research experiments

Abstract

Industrial research experiments are conducted in various scales in the mining industry. Regardless of the experimental purpose, sampling and analysis is normally always a part of the experimental process to collect necessary data. However, in order to ensure that the experiment will enable valid conclusions, the understanding and minimisation of sampling variation is crucial. Two effective methods for evaluation of sampling variability in any process sampling situation are the duplicate and replication experiments. The application of sampling experiments in the early phases of a demo- or pilot-scale experiment is an effective way to both understand the total measurement system variability, as well as the possibility to improve sampling methods if the sampling variability is deemed too high to enable representative results to use for experimental evaluation. LKAB is an iron ore mining company in the north of Sweden where experiments are conducted in all parts of the process value chain with regularity. In the current state of the world, encountering more and more threats to our global climate and environment, a focus for LKAB has been to reduce the use of fossil fuel as well as to minimize waste and tailings. One of LKABs current environmental initiatives is to investigate the feasibility of recovering and processing apatite from tailings of the LKAB beneficiation process. Further processing of recovered apatite will generate critical raw materials, phosphorous, rare-earth elements, and fluorine. To increase the understanding of the process variability of various analytical parameters in a pilot-scale experiment within this project, both duplicate and replication sampling experiments were conducted during one of the pilot-scale campaigns. The sampling experiments were applied to three separate sampling locations where two different sampling methods were used. Results show that both sampling method and sampling experimental method can affect the results obtained. The case study showed that the sampling variability was higher for sampling locations where grab sampling was applied, in comparison to composite sampling that generated lower sampling variability at one of the sampling locations in the pilot plant. This indicate that the composite sampling method can produce more representative results and should be favourable in future process experiments. The results also indicate that the duplicate sampling experiment is more robust to outliers in comparison to the replication experiment. The duplicate experiment is also able to quantify the process variability and evaluate the relative sampling to process variability which can be an advantage in some cases.