Tuning of real-time algorithm for crushing plants using a dynamic crushing plant simulator

Paper in proceeding, 2012

Real-time algorithms have earlier been successfully implemented in crushing plants for the selection of set-points. The algorithms build on principle models of how the on-line adjustable parameters affect the process. The dynamics in a crushing plant are usually consequences of altered states of the plant due to factors such as natural variation of the processed material, degrading equipment performance and more or less stochastic discrete events. Although the real-time algorithms are carefully designed in order to optimize the process, they are not fine-tuned themselves. In this paper a method for tuning real-time algorithms using a dynamic crushing plant simulator is presented. With the plant realistically modeled in a dynamic simulator the constants of the real-time algorithm can be selected before the algorithm is implemented in a real plant. Thus the algorithm can optimize the plant better and quicker. The method is demonstrated with a simulation of a real crushing plant.