EXPERIMENTAL DESIGN FOR CONTINUOUS IMPROVEMENT AND OPTIMISATION *T. G. Vizcarra1, T. J. Napier-Munn2 1 Senior Process Specialist, JKTech Pty Ltd, Indooroopilly QLD Australia, 4068 (*Presenting author: t.vizcarra@jktech.com.au) 2 Emeritus Professor, Julius Kruttschnitt Mineral Research Centre, University of Queensland, Indooroopilly QLD Australia, 4068 ABSTRACT Process improvement initiatives will frequently culminate in trials of new reagents, processing strategies, equipment, and other innovations, in full-scale production circuits. The challenge with these types of plant trials is that day-to-day variability frequently overwhelms the effect of the trial, largely because of unavoidable changes in ore properties and feed mineralogy. Consequently, even if a trial leads to a performance improvement, this can often be camouflaged by noise and variability in the production data. This is a cause of considerable risk in subsequent decision-making, particularly with regards to changes that are capital intensive, and whether these should be implemented into daily production strategies. However, methods to design experiments that effectively manage these complications are available, and frequently used in other industries and scientific disciplines. These types of experiments apply statistical principles to quarantine the effect of the trial and isolate it from interfering factors that cannot otherwise be controlled. Even if the metallurgical effect is small relative to the variability in the data, these statistical methods can establish whether the effect is real and will therefore provide economic benefit to the operation. They will also indicate whether any perceived improvements are merely coincidental, and simply an artefact of the ‘noise’ in the data. JKTech has had key involvement in a number of these plant trials over the years. In this paper, three case studies are presented to illustrate the power of these experimental designs in optimising process response and improving business performance. KEYWORDS Continuous improvement, process optimisation, experimental design, mineral processing INTRODUCTION The profitability of a mineral processing plant is largely driven by day-to-day decisions made at the operational level. Whether or not these decisions are ‘good’ or ‘bad’ is ultimately determined by the response of the processing circuit, in turn reflected in the production data generated during each shift. Ideally, this data would be easy to interpret, putting operators in an immediate position to identify conditions that maximise metal production. In practice, however, there are several impediments to this. By far the most severe difficulty is the fact that conditions are never constant, given that the circuit feed continually changes in grade, hardness and mineralogy. These on-going changes mean that the response of the target variable (e.g. throughout,
RkJQdWJsaXNoZXIy MTM0Mzk2