Comdale Technologies of Toronto has developed an expert system to control the flotation circuit at Cominco’s Polaris mine, in the Northwest Territories. Genex, an expert system tool kit, was used to develop the “smart” computer program which is now being tested at the lead-zinc mine. Genex offers a means of incorporating machine “intelligence” into the decision-making process. It takes information from sensors and from mill operators and makes decisions which dictate the operating and control actions that should be taken in the flotation circuit. The system was to be in full operation by mid- December, 1987.
The program springs from work done in the Mining Engineering Department at Queen’s University in Kingston, Ont., where Prof John Meech has developed an automatic control system based on the theory of “fuzzy logic.” This fuzzy logic controller was designed and tested with a computer simulation model of a secondary crushing plant. His research found that plant throughput using the fuzzy controller is greater than that of manual control for an equivalent level of power use (but more on that later). One of Meech’s students, C. A. Harris, is now president of Comdale, where genex was developed.
In addition to a number of industrial and commercial firms applying genex, Queen’s is using the tool to develop an expert system to help teach basic mineralogy.
Expert systems — a subfield of artificial intelligence (see Artificial Operators in our July, 1987 issue) — are “smart” computer programs that capture human expertise and attempt to imitate the human thought process. Expert problem-solving involves the analysis of large amounts of specialized knowledge. Much of this knowledge is based on rules-of-thumb or aids to learning developed and refined by human experts over many years of practical experience. The knowledge gained from such an expert (say an experienced crusher-man or mill operator) is programmed into an expert system by a knowledge engineer. This knowledge can then permit a novice mill operator to solve problems with a higher degree of competence.
Expert system software uses specific knowledge to carry out specific tasks. All expert systems consist of an inference engine and a knowledge base. An inference engine is the “thinking” part of the system — an algorithm that organizes the order in which the knowledge is to be considered and decides the sequence of activities. A knowledge base, on the other hand, is a file that contains the rules concerning important facts as they pertain to the problem to be solved. The key difference between this technology and conventional modelling is that inexperienced personnel with minimal mathematical skills can participate in the design and use phases of the creation process. Knowledge can be transmitted using ordinary, everyday language.
The control system, designed and developed by Comdale, is based on the technology of expert systems. Genex is an effective tool for the development of expert systems. This tool is available for construction of expert systems to be used as consultants, or as real-time controllers (as is being done at the Polaris mine).
Genex is a microcomputer-based expert system software package which contains a sophisticated inference engine to operate on rule-based knowledge. It contains tools to assist in the creation of the knowledge base. The reasoning, which genex uses to make its decisions, is contained in a knowledge base that is developed with the help of an experienced process operator. All knowledge is implemented in the form of “if-then” rules. Once the expert’s knowledge has been put into the computer, genex allows anyone to use this knowledge for decision-making, to request explanations regarding conclusions the expert system makes, to seek justification for the system’s actions, and to allow the general transfer of knowledge from the system to the user.
Genex can communicate with a number of processes, retrieve data, examine these data in a manner similar to an experienced operator, make decisions, and communicate its control actions to the final control elements or human operators in the plant.
The genex software package consists of two programs: the genex Rule Compiler and the genex Knowledge Application Program. The genex Rule Compiler is used to compile these rules and generate reports and output files which are used by the genex Application Program for problem-solving. Genex is available for execution under the dos and xenix operating systems on an ibm personal computer and compatible micro- computers.
The fuzzy logic controller is, naturally enough, based on the theory of fuzzy logic, which is used in expert systems that deal with the simulation of the human thought process. It does this by introducing the concepts of vagueness and imprecise measurement into the interpretation of information. Fuzzy logic deals with imprecise boundaries. Consider the concept of “old.” To categorize a person as “old” is subjective and depends on one’s point of view. Fuzzy theory provides a means to translate human concepts described by linguistic variables such as “old,” “high,” “low,” and “large” into forms which mathematically represent the vagueness of these concepts. The strategy employed by the fuzzy logic controller is based on the actual operating practices of experienced crushing plant personnel.
The fuzzy logic controller was tested on a dynamic simulation model of a secondary crushing plant developed at Queen’s. It attempts to maximize plant throughput at a specific product size. Constraints include maintaining high-power utilization and avoiding equipment overload. Although effective control of a crushing plant can be achieved manually, disturbances caused by changes in feed rate, feed size distribution, ore hardness or specific gravity usually result in crushing plant operators maintaining equipment at reduced levels of efficiency with some minor feed rate adjustments. This is done in order to avoid trip-outs or equipment failures. The fuzzy logic controller was designed and tested on a secondary crushing plant. The configuration is shown on page b * The plant consists of three 7-ft cone crushers, together with two 8×20-ft, 0.76-inch screen decks. All crushers have 350-horsepower motors enabling the plant to supply a production rate to a milling operation of about 11,000 to 16,500 tons per day. Two internal surge bins are available which allow for independent control of feed rate to each of the 5-unit operations.
Figure 2 is a typical record of productive activity for the secondary and tertiary crushers and screen under manual control. (For simplicity, only one tertiary crusher and screen output chart have been reproduced.) The charts represent a period of 160 to 320 minutes from the start of the shift.
Figure 3 illustrates response curves for the secondary and tertiary crushers and screens, while operating under fuzzy control. The period of operation shown is from shift start to 160 minutes.
For secondary crusher operation, from the start of the shift the controller increased the secondary crusher feedrate in order to attain an “ok” amperage level. However, ore at the start of the shift was difficult to crush, and so two high-chamber-level alarms were triggered at the 15-minute and 35-minute marks. Controller response to these alarms was an immediate reduction in feedrate. There is nothing “fuzzy” about a high-chamber- level alarm. At about 40 minutes, the ore became less difficult to work and the fuzzy controller increased throughput to maintain power draws at “ok” levels. Between 70 and 100 minutes, the controller operated at high throughput rates (about 825 tons per hour), but thereafter the feed tonnage was reduced as ore conditions deteriorated.
Fuzzy control of the tertiary crusher shows increasing feedrate from about the 20-minute mark. After attaining “ok” power draws, the controller continuously adjusted the tertiary crusher feedrate in response to changing ore parameters and fluctuations in amperage. At about 130 minutes into the shift, the controller reduced the crusher feedrate as the tertiary bin level became “very low.” This low bin level was caused by low screen tonnages. As the bin level improved, the controller authorized increased tertiary crusher feedrate.
At shift start, both screens were operational. The reduction in secondary crusher feedrate in response to a high chamber situation at the 15-minute mark, therefore, led to a drawdown in screen bin level. This drawdown produced a “very low” screen bin state. Decision-making by the fuzzy controller dictated a continuous reduction in screen feedrate until the situation was corrected. Screen feed control assumed stable operation after 40 minutes and continued until the 130-minute mark, when low screen bin levels again caused screen feed rate reductions.
The coincidence of low tertiary and screen bin levels at the 130-minute mark occurred because “difficult” ore was fed to the secondary crusher. This ore caused the controller to reduce the circuit feedrate. This affected the over-all circuit performance, since the operating strategy promotes low bin levels in order to maximize tonnage throughput. Although this allows for surge capacity to be available whenever needed, the plant operation then becomes susceptible to secondary throughput changes. Additional rules might be useful to modify the operation of the tertiary crushers and screens when the secondary crusher is operating at a lower than normal tonnage rate.
The fuzzy logic controller shows potential for improving crushing plant throughput above that achieved by manual control for an equivalent level of power utilization. Figure b provides a comparison of fuzzy control and manual control as affected by the secondary crusher, closed-side setting. Lower shift production tonnages indicate the processing of hard, variable ore while higher tonnages are achieved with softer, more consistent ore. The range of shift production for the fuzzy controller lies above that of manual control. REFERENCES Harris, C. A., and Meech, J. A., Fuzzy Logic: a Potential Control Technique for Mineral Processing. cim Bulletin, September, 1987, p. 51-59. Meech, J. A., Using GENEX to Develop Expert Systems. For presentation at the First Canadian Conference on Computer Applications in the Mineral Industry, Laval University, Quebec, March 7-9, 1988, p. 8. Joyce Musial is a Toronto-based geologist and freelance writer.
Be the first to comment on "Research FUZZY LOGIC"