The recovery and concentrate grade of valuable minerals are the most widely accepted measures of assessing metallurgical performance. In mineral sand industry challenges are being found to monitor the heavy mineral product quality during processing in wet plant stream. In this situation to achieve and maintain these requirements a technique for on-site rapid measurement of percent HM values with the required degree of accuracy is desirable in both tails and concentrates. This paper summarizes application of new image analysis technique developed for rapid measurement of percent HM.
To evaluate a technique for the research task the focus was to recognize the limitations of various physical properties that can be used, such as specific gravity, bulk density, radioactivity, electrical properties, magnetic susceptibility and the image analysis which is based on colour intensity of minerals. However there are some measurement difficulties encountered in different situations during the measurements of specific gravity, bulk density, radioactivity and electrical measurements (relative permittivity) of mineral sands. The magnetic susceptibility measurements (MSM) provide an effective 'coarse' method for both dry and slurry samples over the 0 to 100 percent HM range and an acceptable indication for range 0 to 12 percent HM with low error values. For extreme light end (range 0 to 3 percent HM) provide unacceptable error values while for the direct application of MSM to the heavy end (over 90% HM) are not successful. However for the heavy end by separating the magnetic components using a Perm Roll could make a significant contribution, assuming constant conditions of Perm roll and constant ratio between non magnetic to magnetic minerals of heavy minerals.
A simple image analysis method is developed which enables reliable, quantitative assessment of heavy minerals sands in the macroscopic scale. The basic principle of this technique is classification of grey level differences between heavy mineral (ilmenite, rutile, leucoxene, zircon.) and light mineral (quartz) components. Samples are prepared on the sticky transparency sheet and then images acquired scanning through an ordinary flat bed scanner (range of resolution of 600 to 1200) as grey scale images. The scanner output consists of grey scale images with 256 grey levels to be analysed by the Scion image analysis software. Since mineral compositions vary from one site to another, separate sets of calibration samples prepared by mixing known weight of heavy minerals and light minerals. The statistical evaluation of the results showed that the calibration lines are extremely ore specific and the results and accuracy depends on degree of quartz staining (colour of quartz).
The technique can be further improved in the presence of stained quartz by developing software to discriminate individual characteristics of stained quartz with heavy minerals. The requirement of monitoring percent HM values also can be achieved by comparing the area values of the test sample (equal to calibration sample weight) with designated calibration values from time to time followed by adjustments of spirals. However this is a promising method to measure extremely low percent HM (low end) with less than 1% stained quartz in tails and therefore to minimize use of heavy liquids which have number of health and safety issues.
For low percent quartz (high end) in concentrate, the high intensity of predominantly darker grains masks the low grey values eliminating accurate results