The Value of Good Sampling and the Effects on Net Smelter Returns

"The data that many mineral processing plants make use of, is provided by samples taken from sampling systems which are inadequately designed or improperly maintained, and therefore cause detrimental effects to plant operations, production planning and metallurgical accounting! Poorly designed or maintained sampling systems will likely produce low probability samples which give the operator a false sense of security of the product and the process. If data is the basis for action (Flintoff, 1992, p.2), then using the resulting suspect or bad assays by operators, control systems, or metallurgist and administration personnel will act and make decisions which may adversely affect the profitability of a processing plant. Here we will discuss some of the fundamental sampling problems, propagation of errors and a method to estimate what impact this has on an operations Net Smelter Return.INTRODUCTIONIn order to control a mineral processing plant at its best potential, it is necessary to obtain the best possible measurement results of the metal or mineral in question. The sample which is delivered to a measuring instrument, or lab, needs to be of the highest probability to be able to achieve this. This paper will discuss the effects of sampling errors on grade / recovery and grade / Net Smelter Return (NSR) and provide a method to estimate the monetary advantages of using probabilistic sampling in a processing plant.REVIEW OF SAMPLINGSampling, by definition, is the removal of a small representative portion from a total consignment, or flow, for the purpose of accounting or process control.• A sample can only be considered representative, if each and every increment collected, in each of the sampling stages, is representative.• Each particle of the sampling lot must have the same probability of being included in the final sample.• If both of the above conditions are met, then the final sample will be representative of the complete sampling lot.It is commonly known that in order to take truly representative and probabilistic sample, best practices indicate that:• The total stream should be sampled.• The sample cutter should intersect the sample at right angles to the flow.• The sample cutter should travel through the stream at a linear and constant speed and the speed deviations should be less than 5%."