Fuzzy Logic Control In Coal Preparation Plants

On-line analyzers have made possible the use of fuzzy logic control (FLC) for local control applications in coal preparation plants. This includes FLC of heavy media vessels, cyclones and froth flotation. These circuits can be controlled in real time in a feedback control loop with an on-line analyzer. Control strategies can be developed using statistical process control (SPC) techniques that are first applied offline in manual mode through the use of control charts. This process is then brought on line using FLC in conjunction with an on-line analyzer to automate the circuit. The advent of programmable logic controllers into preparation plants has enabled the implementation of regulatory control. As shown in Fig. 1, regulatory control is the first level of process control as applied in a processing facility. This involves: ? The logic sequences for start-up and shutdown procedures and all the necessary interlocks for safety purposes. ? The systems to control individual parameters and individual units. Various sensors provide information that is useful to plant operators, while some can be used as signals to simple control loops (Couch, 1996). This is the foundation of process control. It must be performed well if there is to be any success in supervisory control (level 2). Supervisory control involves overseeing and directing the operation of unit operations, or of the whole plant. This includes more complex loops with cascade control. And it may involve feed forward and feedback systems. There have been some applications of this kind of control to thickeners and flotation circuits in coal preparation plants. In addition, there are supervisory controls on some dense medium circuits and, in a few cases, on jigs. These are implemented with proportional integral- derivative (PID) controllers that are available in the PLC or distributed control system (DCS) operating soft ware. In a coal preparation plant, there might exist 200 to 400 PID loops. In a coal-pro¬cessing plant, this number is increased to 2,000 loops or more. It is important to use the correct algorithm or function to control a process variable to the point where automatic control shows a markedly decreased variability over manual mode. This is where fuzzy logic and neural networks have an advantage over traditional model-based expert systems. Because dynamic models of coal preparation unit operations do not exist, model-based expert systems that are prevalent in minerals processing cannot work. Not enough is known about the dynamics of the gravity separation processes employed in coal preparation. Therefore, an artificial intelligence (AI) system consisting of a subset of fuzzy logic, genetic algorithms and neural networks must be applied in coal preparation plants if a supervisory control system is to be successful across a range of feed characteristics and variable set points.