Correct sampling systems and statistical tools for metallurgical processes

"The optimization of metallurgical processes requires an in-depth understanding of the heterogeneity carried by any constituent of interest. It emphasizes that knowledge of the Theory of Sampling (TOS) is the necessary step to ensure that valuable and valid research programs, pilot plants, feasibility studies, plant design and plant operations have been set on solid foundations. Too many publications today are solely based on the properties of the sample, with no mention of how this sample relates to anything in the environment where it was collected. Unfortunately sampling is an error-generating process that leaves no apparent tracks of its shortcomings. Indeed, it is easier to track down analytical errors than sampling errors. Consequently, bad sampling practices plague the scientific community and the industry and are responsible for severe money losses, unsatisfactory science and processing, and sometimes litigation. A point-by-point strategy is presented addressing sampling problems in early stages of a project. Sampling system requirements for metallurgical accounting and process control are listed. The use of Chronostatistics is suggested to determine if sampling systems are performing satisfactorily and to identify the various types of process variability that may not be easily detected by a conventional inspection of available chronological data. This strategy will help engineering firms, manufacturers, managers and supervisors to provide, install, use and maintain key sampling systems located at very important points in a process. Equipped with the appropriate sampling and statistical tools it becomes easier to minimize reconciliation problems with the mine, with suppliers and with clients. Contrary to the beliefs of many people, good sampling systems are not expensive because in the long run they help to reduce hidden cost and better regulate plant productivity; they most certainly minimize the number of unnecessary meetings looking for solutions where there are none. The suggested strategy will help specialists to extract maximum information from the available data, as large investments and crucial decisions depend on it."