The clear implication for the smaller operator (2,000 tonnes per day or less) is that automation holds little potential benefit. Many gold producers find themselves in this production category and are presumably influenced by this widely employed line of thinking.
To demonstrate the fallacy of this notion, we offer some simple illustrative calculations. These calculations were based on ore reserves, production and cost data for five gold producers with daily throughputs in the range of 385 to 730 tonnes per day. (Source: Anon 1, 1988). Assuming essentially no plant instrumentation, it was estimated that an installed computer-based regulatory process control system would cost approximately $500,000, with an annual associated support/maintenance cost of approximately $150,000. Demanding a 30% discounted before-tax rate of return, the following average improvement figures were calculated: Reduction in operating costs ………… 4.7% (Constant tonnage and recovery), or Increase in recovery ……………….. 1.9% (Constant operating costs and tonnage), or Increase in tonnage
………………. 4.4% (Constant operating costs and recovery), or some combination thereof.
The important point is that, with the possible exception of the recovery increase, these performance improvements are not large and are very likely to be realized with lower level process control. This is especially true when one considers that, in practice, improvements due to process control typically manifest themselves in all these measures.
Clearly, the North American gold industry should move to more fully exploit the benefits of process control to help meet the generic economic optimization criterion. With this in mind, we have written this paper more for the benefit of the neophyte than the more experienced user. (The latter, we assume, is a convert to this technology.) INTRODUCTION
Figure 1 illustrates the three major elements of a control system. (While not strictly part of the system, one can make a strong case for a fourth element — maintenance and/or support — which has a strong influence on all three other elements.) We believe that, of the three, control hardware is of the least concern given the quality/support of present major vendor product lines. The control strategy can pose some problems, but a good starting point is to adopt the control tactics employed by the best process operators. In terms of instrumentation, gold operators can borrow considerable technology for comminution circuits from successful base metal mine control installations. This is not meant to imply that the transfer cannot occur within the gold mining community but, given the relatively short history of control technology in this industry, it is more likely that the transfer will occur from outside (at least in the short term). In the area of leaching and other downstream processes, opportunities for borrowing technology are limited. However, there is a fairly intensive effort under way to develop/evaluate on-line sensors for gold hydrometallurgy applications.
In spite of the sensor/strategy limitations, it is still relatively straightforward to implement a successful process control system based on proven technology. In fact, if there is a major obstacle to process control implementation, it lies at the centre of the triangle, i.e., the mineral processing control engineer. Under today’s market conditions, it is exceedingly difficult to secure a person who will function in this role. Management must be sensitive to such things as training, career-pathing and remuneration. What follows is not a comprehensive review of process control in gold milling. Readers interested in such a review are directed to the work of MacLeod and Bartlett (1987). The remainder of this section deals with control generalities and some specific process examples in selected gold mill circuits. Several examples are drawn from Noranda’s Hemlo mill (Larsen et al., 1988), since the authors believe this plant to be fairly representative of the state of process control in North American gold milling. CRUSHING
Since autogenous or semi-autogenous milling circuits require only primary crushing for feed ore size preparation, there is little need for any form of continuous control. Such is not the case for the conventional crushing plants which prepare feed for rod or ball mills. These plants include secondary crushing and, very frequently, a tertiary crushing stage. Screens are included in the circuit to scalp crusher feed and/or ensure a positive control over crushing plant final product size. The remarks which follow are restricted to these more conventional plants. The typical operating objective of the crushing plant is to maximize throughput at minimum product size. The constraints are to provide adequate time for preventive maintenance and to avoid lost production time due to unit overloads. This latter point is particularly important if the crushing plant is to operate near maximum throughput.
It appears that few operators attempt anything beyond discrete control and some rudimentary continuous control. More comprehensive control strategies exist and have been documented in a number of base metal applications (e.g., Board, 1982; Flintoff et al., 1987; Austin and Flintoff, 1987; Norby and Hales, 1986). Since crushing circuit topology is effectively independent of ore mineralogy, many of the concepts employed in these strategies can be readily applied to gold mill crushing circuits.
Figure 2 (Austin and Flintoff, 1987) depicts the major elements of a crushing plant (continuous) control strategy. The importance of constraints is recognized explicitly through the watchdog control block. In essence, this is a form of intelligent process- monitoring and alarm-handling. The supervisory control block includes the standard elements of regulatory set point supervision but also addresses optimization issues. We believe this structure forms a solid base developing process control strategy. The example chosen for discussion here was taken from Hemlo Gold Mines (Larsen et al., 1988). Although only in its infancy, the control strategy is evolving according to the structure shown in Figure 2.
The Hemlo crushing plant flowsheet is shown in Figure 3. Some of the instrumentation that is germane to the description which follows is also shown on this figure.
For the purposes of illustration, attention will be restricted to continuous control; however, watchdog control does play an important role in this area of the plant. Examples of watchdog control include shutdowns based on low crusher r.p.m., high crusher power spiking, high fine ore bin levels, etc. With regard to regulatory control, there are two pi (proporti
onal and integral action) loops — one controlling the tonnage on the standard crusher feed belt by manipulating the coarse ore bin vibrating reclaim feeders; the other controlling shorthead crusher power draft by manipulating the pantleg feeder. Supervisory control is limited to a single p cascade loop.
Variations in shorthead crusher feed rate will affect the level in the pantleg feeder, and it is necessary to adjust the feed tonnage to avoid spills or the possibility of insufficient crusher feed. Since the pantleg feeder functions like a sump pump, only proportional control action is used in the cascade loop. This supervisory loop handles the imbalance between the two regulatory loops, as detected through level changes. The control law is given in Equation 1. The idea of using t (out) as the controller bias is simply based on the observation that t (out) is responsive to disturbances such as variations in ore hardness or size distribution.
A recurring theme in recent design practice is to omit any surge capacity between the crushing stages. While this doesn’t alter steady state plant performance and has obvious capital advantages, it has a deleterious impact on control. Surge capacity is important in a dynamic context since it accommodates, and indeed facilitates, the measurement of circuit imbalances. To remove it makes for a rigid and complicated system with extensive transport delays which, in turn, necessitates conservative or safe (sub-optimal) operation. GRINDING
Grinding is the last stage of particle comminution before the ore is delivered to the leaching circuit. The importance of this area of the plant cannot be overstated; yet, in many plants, the leaching circuit receives a disproportionate amount of attention. (This argument holds for the crus hing plant as well.) The reasons for this situation are not clear; however, Hulbert and Barker (1984) offer some insight. They note that grinding circuits can operate as open-loop stable with respect to major upsets (sump overflow, mill overload, etc.) for extended periods of time. Despite this stability, disturbances occur which cause inefficient operation and variation in distribution of product size.
The importance of controlling the grinding circuit product size distribution to minimize recovery losses will be case-specific. The notion can be illustrated by means of a simple calculation. Figure 4 shows a typical gold recovery-versus-grinding-circuit product particle size relationship. The point (s*, r*) is the target operating point and is usually identified through some sort of economic optimization calculation. For mathematical simplicity, it will be assumed that, for small variations about the operating point, the curve can be approximated by a second order truncated Taylor Series Expansion. Replacing the partial derivatives with coefficients, yields the model in Equation 2:
Suppose one were to have conducted the following hypothetical experiment at a fixed fresh ore feed rate. The grinding circuit product particle size was sampled continuously, yielding the trajectory shown on the left hand side of Figure 5. Converting this information to a probability density function would produce a curve something like that shown on the right hand side of Figure 5. With Equation 2 and the probability density function, p(s), it is possible to compute the mean gold recovery over the sampling period by applying the statistical expectation operator as in the following Equation 3:
Equation 5 provides some simple yet quantitative insight into the control problem. (It can be used in control system justification studies.) For example, a particle size control system would operate to make s = s* and d = 0. Moreover, it would minimize excursions from set point and thus reduce the particle size signal variance, os2. From equation 5, this clearly would have the desired response of making r = r*. By inspection, it is easily seen that the value of the control system increases with increases in the magnitude of the co-efficients a and b. Further interpretation of Equation 5 is left to the reader. Suffice to say that particle size control can be an important aspect of grinding circuit operation.
As mentioned in the crushing control discussion, opportunities exist for gold operators to borrow grinding control technology from successful base metal applications. In some instances, the technology itself is not directly applicable, but the concept can be used. Such is the case with the fairly extensive work that has been done on pebble mill control in South Africa (Gossman et al., 1982; Flook and Plasket, 1982; Pauw et al., 1985).
The typical operating objective of a grinding circuit can generally take one of several forms:
* Maximize tonnage subject to the physical constraints of the circuit.
* Maximize tonnage at some particular product size specification, subject to the physical constraints of the circuit.
* Minimize particle size at constant tonnage, subject to the physical constraints of the circuit.
Criterion 1 appears to be relatively rare in gold milling, but there are examples for criterion 2 (Larsen et al., 1988) and criterion 3 (Parashyniak et al., 1988). It is clear from the statement of these control objectives that unit overload protection can be an essential part of any robust industrial control strategy (ref. Fig. 2: watchdog control). Once these techniques are established, they can be integrated into a higher-level control algorithm which will address the operating objective.
]]>
Be the first to comment on "COMPUTER PROCESS CONTROL Part 1"