HIGH-TECH OPERATIONS — THE MANLESS MILL

Distributed control systems (DCS) were introduced in the mid-1970s, according to Isabelle Clauzier and Burton Roberts of B.R.I.C. Systems Integrators, who delivered a paper on the subject at an operators conference in Sudbury this past February. The first to incorporate DCS was an iron ore mining company in Minnesota. Today, they can be found in autoclaves, concentrators and pelletizing plants. This system offers the advantage of “distributing clusters of controllers throughout the plant. Each controller has a direct communication path to a central control room. Distributed intelligence, distributed risk, graceful control degradation, battery back-up and redundant communication paths were some of the features implemented into the DCS architecture,” they told conference participants. They went on to say today’s DCS systems are highly sophisticated and reliable. And they cannot crash.

Both Inco and Falconbridge have introduced DCS at their Sudbury operations. In the remainder of this article, authors Robert Spring and Mark Franklin of Noranda Technology Centre offer their experiences with DCS installed at the Brunswick concentrator.

In October, 1989, Brunswick Mining & Smelting added the G2 real-time expert system to the existing Fisher Provox DCS. One feature of the G2, supplied by Gensym Corp., is its high-speed, 2-way communications link with the DCS. This permits real-time access to read and/or change the values of the more than 500 field instrument signals connected to the DCS. The other major features of G2 are its schematic flowsheet approach to programming and its natural language text processing for entering rules.

It was decided that the first test of G2 would be an on-line dynamic material balance. The test would demonstrate how a plant flowsheet could be made part of a computer program for process control. The first step in constructing a knowledge base was to define the different types of process equipment that exist in the concentrator. The types of equipment that were defined include rod mills, ball mills, cyclones, conditioners, flotation cells, as well as other common types of mill equipment. A small picture of the piece of equipment was drawn using a computer mouse and then the input and output connections to the piece of equipment were located on the drawing. Once the equipment types were defined, a picture of the flowsheet was constructed using the computer “mouse.” Figure 1 is a small portion of the flowsheet (the complete flowsheet includes more than 100 streams).

The next step in the design of the knowledge base was to write a short set of generic rules that determined how flow moved to and through the various pieces of equipment. These rules were written using G2’s English-like grammar. For example, the following is a valid G2 rule describing how flow moves from one piece of equipment to another:

The input-flow of any process-equipment P = the output-flow of the process-equipment connected to the inlet of P

Other rules describe how the equipment modifies the input flow. The various types of equipment were modelled using different combinations of plug flow, mixer, splitter and adder components. For example, the model for the discharge ore flow rate from a grinding mill is as follows:

The discharge-ore-flow of a grinding mill = the average value of the feed-ore-flow between four minutes ago and two minutes ago

Although the mill model appears trivial, the temporal character of the model is very significant, particularly in the context of dynamic material balance calculations. Like the real mill discharge, the modelled discharge does not immediately react to changes in the feed ore flow rate: two minutes elapse before any change is seen in the discharge ore flow rate. Also, like the real discharge, the modelled discharge ore flow is an average value and never changes abruptly; the enormous ore volume contained in the grinding mill buffers any sudden changes. This model captures the essential dynamic behavior of mass flow through grinding mills.

Only a few such generic rules are needed to interpret the most complex flowsheet. This greatly reduces the time and effort needed to develop a system for process control. G2 uses fewer than 50 rules to balance the material flow in the three grinding circuits and in all of the flotation circuits.

The final step in the construction of the knowledge base was the addition of sensors to tell G2 which data to send to the DCS system. The reagent flow set-points in DCS system are ratioed directly from the simulated mass flow rates calculated with G2. Figure 2 shows a typical response of the rougher feed mass flow to a change in the rod mill feed. Since the reagent additions to the rougher circuit are tied to the rougher feed, they will automatically turn themselves on and off or adjust themselves to varying rod mill feed tonnages. This minimizes the production of below-specification concentrates, while ensuring maximum recovery. A detailed internal audit has shown that the G2 expert system hardware and software working as part of this automatic reagent control scheme has already paid for itself in the first four months of operation.

The marriage of the G2 real-time expert system and the Provox distributed control system takes advantage of the strengths of both systems. G2 provides a programming environment matched to the skills and needs of process engineers. The DCS provides an “interface” to sensors and actuators, efficient single-loop control and a robust interface for the flotation operators. Moreover, by integrating G2 with the existing DCS, the expert system was put “on-line” without the expense of new wiring and without the need to train the flotation operators on new consoles.