Developing Real Grade- Recovery Curves for Sorting

"INTRODUCTION Sorting technology has advanced rapidly recently. Of the two possibilities particle sorting has advanced especially in introducing new sensors and the use of multiple sensors. The analyzing and computing lead times have also shortened. This has made it a viable technology option for many applications but not for the most important sorting task at a mine; the run-of mine ore. Bulk sorting, however, has advanced in smaller steps, even if the technology has the capability to tackle the run-of-mine ore and offers a high potential to bring mines substantial economic benefits. The sensors, which are useful are restricted by the difficult operating conditions or by the ore characteristics. One of the important features of run-of-mine sorting technology is the major material handling infrastructure required. The capital expenditure can be substantial requiring good understanding of the sortability curve (grade-recovery curve) behavior of the ore. This paper discusses one industrial case, where substantial amount of data measured from the ore feed belt has been obtained. This data was then analyzed for the grade recovery curves obtainable from that ore with sorting. EXPERIMENTAL The XRF measurement used was the FCA device from IMA Engineering (Fig.1). It is a specially built version of XRF designed to operate with coarse material. The measurement was situated at the main belt conveyor taking ore from the primary crusher to the stockpile. The crushed ore was about 80% -300 mm at a capacity of 1100-1200 tph (Fig.2). Measurement times were varied from 30 seconds to a minute. Measurements were obtained over several months.As the XRF is a surface distribution measuring technique, the measurement was first tested upon accuracy and reproducibility against plant daily wet chemical assays and against on-line Courier feed analysis. As there is a varying delay in the ore stockpile, the delay time was used as a variant in the validation. Different measurement times were tested for accuracy. A one minute measurement time gave a total measurement relative standard error, which was both for Cu and Ni below 3%. The results showed also very robust correlations; for Cu, the correlation coefficient was R > 0,94 for one minute measurements over three months and R > 0,86 for Ni respectively. Fig 3. Shows a measurement period of two hours showing substantial feed grade variations (given as relative value; 1.0 represents the mean value during the measurement month)."