The “best” control strategy for a given plant is the strategy that most closely fulfills the overall operating objectives. The best strategy will likely not be “optimum” for any given specific goal but rather will involve a series of trade-offs made in consideration of the many factors that influence the operation in any significant way.
Documentation was very important with so many strategies being tested. Two drawings were made for each strategy: (1) a grinding circuit process instrumentation drawing similar to Figures 4 and 5 but with much more detail and (2) a terminal strip connections chart complete with required controller setting (see Figure 12). The drawings enabled easy reconnection of any given strategy at a later date.
This strategy was rejected after a short test because the size loop 1 created an ambiguity in the required sense of control action for the density loop 3. Consider two possible control system conditions in explanation:
A) All three loops on automatic and the rod mill feed getting hard. As the ore gets hard, the rod mill discharge and cyclone feed becomes coarser causing the overflow (and underflow) to move in the coarse direction. The size loop adds water to the sump reducing the density in the cyclone feed. The total response, even though an increased underflow increased the circulating load, was toward a lower density. This required that the control sense of loop 3 be connected to reduce feed for reduced density.
B) The size loop on manual and the other loops on automatic as the feed becomes hard. As for the first example, the cyclone feed, overflow and underflow become coarse, but, since the size loop in on manual, no water is added. The total response for this case is an increase in density as the circulating load increases. This requires that the control sense of loop 3 be connected to reduce feed for increased density, just the opposite of the first example.
In addition to these specific examples, the density response was generally unpredictable as an indicator of circulating load.
Under normal reclaimer operation, this strategy worked quite well. When the front loader was pressed into service however, the system couldn’t cope with the rapid, large ore changes. The top strip chart in Figure 6 shows how this strategy worked when the reclaimer was feeding the plant while the lower chart indicates front loader operation. Because this strategy held internal circuit parameters fixed, there was no “cushion” to absorb input effects. Changes at the input (ore) were fully reflected as changes in the output (particle size). This strategy did cope with minor ore changes (top chart—Figure 6) during reclaimer operation but circuit lags made effective control impossible for large changes. Size control was very poor under these conditions and attempts to tighten it resulted in undesirably large variations in feed rate.
An attempt was made to improve the operation of strategy B by changing loops 2 and 3 as indicated. Since control was not any better (and perhaps worse), it was quickly abandoned and remained undesignated as a separate control strategy.
Since sump level is a good indicator of circulating load, this strategy can significantly reduce the cost of a control system by eliminating the mass flow measurement. Here it was tested since it could provide reliable back-up for other circulating load measurements. Unfortunately, it never functioned as well as desired.
Control action of the cyclone pressure/pump speed loop to hold cyclone pressure constant greatly accentuated sump level variations forcing the sump to run over or run dry before the feed rate could respond. Switching the pump speed to manual greatly improved operation but the pump curve flatness and sump depth were not sufficient to permit the self regulation of sump level necessary for this approach. Figure 7 illustrates the problems encountered.
This strategy worked well except for worst ore conditions. At worst ore conditions, this strategy resulted in larger and faster feed rate changes than desired even when the gain of the flow controller was set to minimum.
With sump level controlling pump speed, the speed of the pump becomes a good indicator of circulating load. This strategy was tested as another alternative to the direct flow measurement (like strategy C above). This strategy was identical to D except that pump speed controlled feed rate rather than cyclone feed flow. Since these two quantities were essentially the same in the short term, the performance was nearly the same as strategy D except that it was easier to tune the controller within operating range for bad ore conditions. The basic problem with this strategy was pump wear. Two pumps, and controller tuning and set point had to be changed depending on which pump was in service. Operator acceptance was also rather low for this strategy because they did not understand the logic of using pump speed to control feed rate. Figure 8 shows strip charts of this strategy. Note the good response to set point changes and front loader operation.
This strategy yielded the best overall performance and came the closest to accommodating all of the operating objectives. Controller tuning was easy, response was good, and operator acceptance was high. Figure 9 shows strip charts of this strategy for very consistent ore conditions except for two periods of front loader operation. In the left hand set, the size loop 1 and the mass loop 3 were switched to manual just after the front loader was put into service. Compare this operation with the fully automatic operation in the right hand set. Figure 10 shows strip charts for less consistent but more “normal” ore conditions. In a 15 hour period, the control system varied the feed rate from a low of 312 dry T/H to a high of 408 T/H. At that point, the front loader had been placed into service and could not keep up with this tonnage demand. The mass flow to tonnage control loop set point had to be lowered. Note the small variations in mass flow that were permitted by controller tuning in order that large, short term feed rate variations be held to a minimum.
