McArthur River Mining
At McArthur River Mine, we were asked to improve their flotation process using our KSX expert system.
We were asked to specifically address the following issues by the client:
- Downstream stream issues impacting on the flotation performance rather than the reagent additions, air rates or pulp levels. The issues could be coarse grind size, blocked cyclones, changes in water chemistry, dirty process water.
- Reagent additions g/t is different for each orebody and also depending on where the ore comes from in the pit (North, South) ores can treat extremely differently.
- They had tried to split the circuit up to concentrate on individual bank recoveries, however the effect of residual reagents caused poor performance and it was difficult to sync the overall performance
- Setting limits, trigger points proved difficult as each orebodies can react very different.
Control Overview and Solution
Manipulated (Output) Variables were:
Rougher – Xanthate, CuSO4, Air
Cleaner – Xanthate, CuSO4
Input Variables were:
Rougher – Mill Feed Rate, Zn Grade (Feed, Tails, Concentrate, Final Concentrate) Lead Grade (Feed), Rougher Recovery, Overall Recovery
Cleaner – Mill Feed Rate, Zn Grade (Tails, Concentrate, Final Concentrate)
Achieve target final concentrate grade, while maximizing recovery, and minimizing reagent usage.
The system used a set of fuzzy logic rules to adjust primary collectors, secondary collectors (Xanthate and CuSO4), and air. It used upper and lower limits for reagent dosages based on grams of reagent per ton of zinc (and also lead for the secondary collector) so that as mill feed rate and head grade changed, the dosages would automatically change. This approach tied the dosing of reagents back to the ore entering the mill. Within the g/ton ranges, the dosage would be adjusted up and down based upon metallurgical performance as indicated by the Courier online x-ray analysis. On the rougher, when reagent additions were nearing the far edges of the control area, air would be adjusted.
Validation of the X-ray analysis were performed before each change, assuring that the values were fresh, and within believable ranges. If large changes in readings occurred, the system would wait until subsequent readings confirmed the change. This allowed for stable and smooth control. In the histograms for reagent addition, you can see that the dosage is much more variable under the expert control, but on average lower, and the grade and recovery less variable, and higher. This is exactly what we want. The operator typically cannot make as many changes, and makes bigger ones. Thus operator control sticks with certain setpoints, and the performance varies, while the expert system moves the setpoints much more, and the performance is more stable and improved.
More variation (more fine adjustments being made) in dosing, but a much lower total usage (9% savings).
Improved recovery on the cleaner by 8%
Reduced usage by 14%
Less variation (measured by std deviation), and a 14% increase
With expert on, you can see that there was less variation and a higher average value (5% Increase)
Usage reduced by 2%
Rougher recovery was increased by 9%
|Antofagasta – Los Pelamres”>|