In mineral processing plant , although there may be elements of discrete I/O in control strategies (e.g., opening or closing cyclones to maintain header pressure), they are normally designed to continuously modulate manipulable variables (e.g., feeder speed, valve position, pump or mill speed) to ensure that the controlled variables (e.g., ore flow, water flow, tank level, particle size) are at or near the set-point value. 

For that reason the emphasis in this section is on continuous control, and Figure 1 conveniently summarizes the levels of continuous control, while offering insight into the structure of control strategies.

Fig 1. The levels of continuous control

Fig 1 illustrates that there is some performance benefit associated with each level of control strategy. It also carries the important implicit message that control strategies are hierarchical. That is, one cannot build an effective supervisory strategy if the regulatory strategies underpinning it are ineffective.

This is not merely a point of academic interest, because numerous studies in all process industries have highlighted unexpectedly poor performance of the low-level controls. A similar argument can be advanced for optimizing controls, which must rely on the supervisory level.

The definitions of what fits where in this hierarchy are debatable, but some general attributes are associated with each level.

Regulatory Control Strategies

_ Strategies are mostly implemented with feedback loops aimed at stabilizing process inputs, such as ore and water flows, or bin and tank levels.

_ These loops almost always involve the PID controller, generally using only the Proportional Integral-Differential (PID) functions.

_ Typical control intervals range from one to a few seconds.

_ They are always implemented on the DCS or PLC operating software.

_ Occasionally dead-time compensation and gain scheduling are required (e.g., long conveyor belts and multiple feeders).

_ For highly nonlinear systems, it may be necessary to resort to other control options, such as fuzzy logic or self-tuning controllers.

Supervisory Control Strategies

_ These strategies calculate set points for the regulatory strategies in the pursuit of some operational

objective, such as maximum throughput subject to a maximum particle constraint.

_ The simplest form may be the cascade PID loop that delivers a set point to the associated regulatory

loop.

_ Typical control intervals range from a few seconds to a couple of minutes.

_ They are almost always implemented on a dedicated PC on the DCS/PLC network.

_ Strategies are frequently multivariable in nature; for example, attempting to control grinding circuit product particle size and recirculating load (see example below).

_ They are prone to interaction problems; i.e., the multivariable nature of the problem leads to

competition among the supervisory loops.

_ They often require sophisticated approaches, such as heuristic, model-based, or blended

approaches.

Optimizing Control Strategies

_ These strategies calculate operating objectives for the supervisory strategies, based on some economic objective function.

_ Typical control intervals range from a few minutes to an hour.

_ They are almost always implemented on a dedicated PC on the DCS/PLC network.

_ They tend to employ optimization techniques based on plant experimentation (SSDEVOP) or analytical techniques (e.g., multivariable search) that employ adapted process models.

_ They tend to look at the coordination of several circuits to ensure that local optimization of each does not lead to the suboptimization of the whole plant.

Using a primary ball mill circuit, we can create a simple illustration of the structure outlined in Figure 2. Figure 2A shows the process flow, instrumentation layout, and typical regulatory controls for such a circuit. It should be noted that control strategies are usually documented through a combination of loop narratives and Process/Piping and Instrumentation Diagrams. In both cases, there are standards for preparation one should use, but in this example a quasi-PID representation is employed. There are four regulatory loops to stabilize inputs and internal variables. R1 is a PID loop that measures tonnage, W, and regulates feeder speed, VS, to maintain the set point, entered by the operator. R2 and R3 are PID water flow stabilization loops that ensure flow set points are maintained in the presence of variation in supply pressure. R4 is a sump-level control PID loop that would ensure the tank does not run dry or overflow.

Fig2A An illustration of regulatory control in a primary ball mill circuit

There are a number of possible operating objectives for such a circuit, but here we assume that the goal is to maintain the product particle size at some set point, and to manage the circulating load so as to ensure maximum ore feed rate while avoiding a ball mill overload. The approach adopted for this example is to employ cascade PID loops, one delivering a set point to the sump water addition to maintain particle size, and the other sending a set point to the tonnage loop to maintain circulating load.

This is shown in Figure 2B. Here S1 is the standard ratio controller aimed at maintaining a constant slurry density in the fresh feed to the mill, which could just as easily have been shown at the regulatory level. S2 is the supervisory cascade loop to regulate circulating load, and S3 is the particle size cascade loop. For completeness, the optimizing strategy is depicted as O1, and it endeavors to ensure that maximum revenue is generated across the plant by avoiding capacity imbalances that would lead to downtime, and by continually reevaluating the optimum grinding circuit throughput and particle size for maximum plant net revenue.

Fig2B. An iIlustration of supervisory and optimizing control in a primary ball mill circuit

The development of a successful control strategy requires a very good understanding of process dynamics and operating characteristics, a broad knowledge of control tools, and the ability to clearly articulate the control problem; that is, what needs to be done and where the priorities lie. At a minimum, it will comprise a blend of regulatory and supervisory techniques that must work well, both independently and together. Although it has not been discussed previously, good operator training and strategy documentation are also required to achieve the benefits attached to the investment. In the past, the mineral processing control community has been rather poor at the last two steps, and as a consequence, has frequently been condemned to repeat history when new process control people are brought on-board.