Identification of a Coal Separation Process using a Soft Monitoring System with a Genetic Algorithm

1. Abstract A simulation analysis of a new method to identify a raw coal separation process to produce concentrate, middle product and refuse is presented in the paper. The washability characteristic of a raw coal, the partition curve of a gravitational separator and separation densities have been identified. The washability characteristic determines the yield and ash content (or other quality attribute) of elementary density fractions in a given size fraction of a raw coal. The partition curve of a separation machine (heavy media vessel or jig) determines the probability of elementary fraction flow to concentrate, middle product or refuse. This probability is the function of density and size of elementary fraction and the separation density of the machine (heavy media density or a set point in the refuse removal controller in the case of jigs). On-line identification is performed with the use of a soft monitoring system consisting of belt scales for tonnage measurements of products, on-line radiometric ash monitors and software interpreting these measurements. The signals from monitoring system create a vector determining properties of products. These signals are compared to the respective signals generated in the on-line simulator working in parallel with the identified system. Sets of washability characteristics, partition curves and separation densities are automatically generated in the on-line simulator. They create a vector which determines the simulated properties of the three products. The best set of generated values is chosen on the basis of the minimum “distance” between the two vectors according to the accepted criterion. The search for the minimum distance of those two vectors is performed with the use of a genetic algorithm. The method has been tested using a simulation model for the three-product coal separation process in a jig.