Table of Contents

Figure 4 is a schematic diagram of a ball mill/cyclone control system. This diagram shows the instrumentation and the calculated variables used in the control strategies.

The prior detailed process analysis study produced two major conclusions:

- Reduce mill feed size
- Reduce ball charge size

The reduction of the mill feed size could be reduced by modification of the tertiary crusher system by increasing the power draught and proper screen sizing. Detailed scale up experiments indicated that the ball charge size should be reduced due to the small contribution to the grinding efficiency.

It was identified that the disturbances to the ball circuit are:

- Mill feed size distribution due to bin segregation and crusher circuit operation,
- Ore hardness and ore mineralogical structure and composition due natural mining characteristics,
- Pumping/Classification limitations and equipment wear.

The initial detailed process analysis study indicated that the calculation of inferred calculated variables can provide with adequate information for the development of the control strategies.

The translation of raw measurements into adequate indices (process indicators) permits the operator to concentrate into a few valuable variables after prechecking by the computer system. This concept contributed to the definition and development of the ball mill load index, the grind index and the transport index calculation obtained from raw measurements filtering and mass balance calculations. The circulating load flowmeter and density gauge were not considered vital measurements for implementation of the control strategies since these values can be obtained from calculations from other measurements and where not used by the controls.

The pairing of the Controlled Variables and Manipulated variables was set using the process table presented in Table 1.

The primary objectives for the grinding control system can be stated as to provide with a flexible and adaptable easy to use system to either:

- maintain an optimal throughput depending on the ore conditions. Which translates to: provide to flotation a constant size distribution, at a size selected to meet metallurgy requirements, so as to improve overall recovery.
- or to maintain a stable operation while assisting the operator in maximizing the throughput while avoiding frequent upsets or spills and maintaining an adequate grind. Which translate to: eliminate overloads and choking of cyclone apexes which reduce throughput because of the need to cut feed rate while “grinding out” the mill and try to reduce overgrinding, thus increasing throughput.

The pairing of the variables was made with the required flexibility concept in mind. So both criteria can be met by the same system by adequate setting of the controller modes.

A simplified overall function block schematic for the new grinding controls is shown in Figure 5. At the left side the ball mill load constraint and transport indices are shown. At the right side of the figure the infered grind index and the sump level controllers manipulating the mill discharge water are depicted. Notice that there are seven controllers with ten indices and only three manipulated variables.

The ball mill load control system: The main control scheme is the ball mill load controller. Any additional capacity of the ball mill depending on the grind settings is sensed by the ball mill load controller and the feed rate is increased (if desired to bring the ball mill its capacity). It was implemented as a cascade controller to set the new feed rate based on a constraint control with a long term horizon control provided by a 20 minutes mill load calculated variable and limits on pumps amperage and sump level rate of change. The operator enters the constraint and alarm settings from a special panel. The rate of change is provided on a seven minute basis to identify large disturbances or a change in mill grinding characteristics. The pump amperage is also closely monitored as the second constraint and the sump level rate of change as the third constraint.

The success of the implementation of this control strategy resides in the adaptive nature of it. The system is at all times monitoring the state of the ball mill transport and loading thus advising to the operator where to set an alarm to start reducing feed. The gain in mill throughput by working closer to active constraint is depicted in Figure 6. This strategy works very well with difficult to grind ores and very well with softer ores. Throughputs of up 400 TPH can be achieved as depicted in Figure 7 when running regular ores from a regular tonnage of 250 TPH, while maintaining the metallurgical performance indexes in range.

**Ball Mill Load Detection Scheme**

This scheme uses the reasoning followed by the operators in interpreting the power draught chart with the earlier electronic controllers. A technique was developed to infer the ball mill loading and to set a constraint procedure using the new system. The developed algorithm uses the computer memory to store the history of several variables and to calculate process variables rate of change and moving averages. Adequate signal processing techniques were used to interpret the raw power draught signal from the ball mill motor. A first order filtered raw measurement was processed by a least square filter to obtain the rate of change and a reasonable base value for a typical process time constant period.

**New feed rate controller**

The new feed rate controller is the secondary controller. It takes the cascaded signal from the ball mill load controller or the operator feed rate set point when in automatic. This controller maintains the feed rate selected by the ball mill controller by raising or lowering the speed of a variable speed feeder.

**Grind index control strategy**

Various techniques are available to measure on stream particle size. Usually these expensive instruments require considerable maintenance and periodic tedious calibrations. For the type of circuit in study it was not economically justified the installation of a particle size analyzer for all the grinding lines. The inferential size prediction was chosen due to the simplicity in implementation with current state of the art control systems. The calculation used resulted from the evolution of a rigorous inferred particle size-estimator developed for the dynamic simulation of grinding circuits.

The grind index is calculated performing a mass balance around the ball mill circuit to estimate the cyclone overflow density (viscosity effect) and a correction factor due to specific power consumption (size effect).

The grind cut index is maintained by controlling the addition of water to the cyclone feed. This water control method sets the cyclone feed viscosity at the proper value to give a split at the desired cyclone overflow particle size distribution. Because viscosity is principally a function of density and surface area, controlling to a fixed size by adjusting of hydrocyclone water continuously compensates for viscosity changes in the cyclone feed that occur because of particle size changes in the mill discharge slurry.

The grind index controller is implemented as a regular PI controller cascading a water setpoint to sump water addition controller when the pump condition is normal. A switch was implemented for the ball mills without a variable speed pump to maintain the sump level within reasonable limits to control the sump level by manipulating the sump water. In this case the grind index controller is left in manual. The implementation of this loop uses the water flow instead of the valve controller direct to prevent the adverse effect on the grind cut index and to be able to maintain a reasonable amount of water under critical conditions.

**Ball mill transport index control strategy**

A simple ratio controller was implemented to regulate the solids content in the mill by manipulating the feed water. A calculated percent solid index is used to monitor the transport in the ball mill. A separate study was initiated to develop a measurement of this property (Bascur, 1985). The large volume (capacitance) of circulating load versus feed water involved load precludes a good control. The historical inventory tracking of the ball mill power draught decouples this water addition by adding in proportion to the feed rate set by the ball mill load controller.

**Sump level controller**

When a variable speed pump is available in the grinding line the sump level is controlled by setting the pump speed. If the variable speed pump is not available a dead band controller was implemented with an automatic switch coordinated with the grind cut index controller. When the pump is new the sump level will be lower so water is added to keep the pump from cavitating and the grind index is kept in manual. At other times, the water is adjusted according to the grind index except when a high level sump is indicated due to an old pump. The logic involved in implementing such switch strategy is relatively simple using a current state of the art distributed control system.

**Water flow controller**

This controller accepts the output from either the sump level controller or the inferred grind index as a water flow set point (cascaded). The controller then maintains the automatic valve in the required position. An hysteresis compensation lookup table was implemented to avoid over controlling due to valve wear. All actuators were implemented with output clamping and hysteresis compensation.