Grinding to liberate, flotation to recover: a copper rougher

97 t/h of a 2 wt% chalcopyrite ore in quartz, crushed to a 2 mm median, is ground in a grate-discharge mill and floated in an 8-cell rougher bank. The feed declares its composition by size (psd_minerals) and its texture (150 micron grains), so every unit acts on mineral by size, not on one distribution. The mill is Whiten's perfectly mixed mill with a 0.6 mm grate: nothing coarser can leave, so the coarse fraction is held and ground until it passes. That is why the mean residence time comes out at 643 s against the 300 s the fines see, and the mill holds 17.9 t. The product P80 is 240 microns for 615 kW. Chalcopyrite breaks 1.5 times faster than quartz here, so it ends up enriched in the fines. Liberation is computed, not assumed: at 150 micron grains only 17% of the chalcopyrite sits in fully liberated particles at this grind. The rougher floats every particle type at a rate set by its grade, so a composite floats slower than a liberated grain and drags quartz with it. It recovers 87.0% of the chalcopyrite into 15.4% of the mass, a concentrate grade of 11.3% against 2% in the feed, and leaves 0.31% in the tailings. The sensitivity is the grind-recovery trade. Grind longer and recovery and grade rise together (80% at 100 s, 93% at 900 s) while the mill's power rises with them, which is the decision a concentrator's grind size is set on. Bounded, and stated. Every ore parameter here is illustrative and fitted to no deposit: the selection rate, the grain size and the flotation rate constants are measured per ore (the breakage-fit tool measures the first from a batch test). A composite's flotation rate is taken as proportional to its grade (linear exposure); there is no froth recovery, no size dependence of the rate beyond liberation, and one valuable mineral.