# Classification & Concentration of Heavy Minerals

The recovery of heavy minerals by gravity concentration during grinding is practiced widely in the minerals industry. Studies have shown that the grade and recovery of the concentrate depends not only on the performance of the concentrator but also on the degree of liberation achieved in the mills and the scavenging action of the hydrocyclones closing the grinding circuit. In the current study, consideration is given to the development of mathematical models to quantify and unify the operations of comminution, liberation, classification and concentration.

## Performance of a Production Circuit

An important first step in understanding gravity concentration is to quantify the mass flow of gold and total solids as a function of particle size for each stream of a complete circuit.

Composite samples were collected from the circuit over a period of about one hour at the following locations:

– fresh feed
– pebble mill discharge
– hydrocyclone overflow
– hydrocyclone underflow
– plane table concentrate
– plane table tailings.

It is evident from the differences in assays between the hydrocyclone overflow and underflow products that much of the gold in the circuit is either liberated or associated with minerals of high specific gravity. Mineralogical examinations of metallurgical plants have indicated that at particle sizes below about 150 µm much of the gold is free and occurs as isolated grains. Because of the ductile nature of gold, one would expect these grains to become distorted in shape rather than in size during grinding.

## Theoretical Framework

In most ore-dressing operations it is common to characterize the properties of particulate material in terms of mass-size and grade distributions. Size distributions are readily measured by standard sieving techniques. In principle it is possible to separate particles into different grade classes using heavy liquids. In practice it is usually difficult to achieve such separations, except for the simplest of ores. Gold ores are particularly difficult to treat in this way; grades are extremely low (typically parts per million) and the mineralogy is very complex.

The size distribution of the ore is expressed in terms of the mass fraction retained on each sieve of a nest of standard testing sieves. The size class corresponding to the largest particles of interest is denoted by the index i=1 and the smallest particles are denoted by i=n. Size class n corresponds to all particles smaller than dn. When barren particles or particles containing locked gold break out of a given size class, say class j, the specific rates of breakage can be characterized by the selection function, Sj . The primary fragments of breakage are distributed so that a fraction, bij, of the particles originating in class j appear in class i. The function bij is referred to as the breakage function. The selection and breakage functions of the free gold are given by Sj and bij respectively.

Although more general forms for the breakage, selection and liberation equations were chosen originally with many additional parameters, it quickly became apparent that the parameters could not be estimated uniquely from the available data. Because of this problem, the equations had to be simplified to their final form in order to reduce the number of experimental parameters. As a result it can be argued that the models only have empirical value, even though the size distribution and assay data are correlated accurately.

Plane tables have found favour with many South African mines because they have no moving parts and can be operated continuously with minimal supervision and maintenance. Additionally, plane tables have high throughput capacities that make them ideal for primary or rougher concentration duties.

The deck of the table is covered with rubber matting that has riffles running parallel to the direction of pulp flow. The flowing pulp forms a layer or film over the matting. Particles are subjected not only to hindered settlement forces, but also to forces arising from the velocity gradient across the layer of pulp. Velocities increase from zero at the surface of the table to a maximum at the exposed surface of the pulp.

If consideration is given only to the effects of hindered settlement, one would expect large light particles to be recovered along with the small heavy ones. However, when large particles settle to the surface of the table, they remain exposed to forces arising from the velocity gradient in the flowing layer of pulp. Thus large particles tend to saltate at high velocities on the matting and so the probability that they will by-pass the slots will be high.

It was concluded that a circuit based on gravity concentration and flotation would not lead to tailing grades that would be acceptable in terras of conventional metallurgical practice. However, the economics of underground processing are obviously very different and reject grades of about 1 g/t could well be acceptable. Unfortunately, a detailed discussion of the economics of underground processing is beyond the scope of this paper, although the problem is currently being looked into.

A systems approach has been adopted to describe the behaviour of gold in grinding circuits. Such circuits are of course but a sub-system of a complete ore dressing operation. It is apt therefore that the ideas developed in this paper should be extended to other unit operations commonly used in the treatment of gold ores, such as washing, screening and sorting.