Designing of Intelligent Expert Control System Using Petri Net For Grinding Mill Operation

The paper utilizes Petri_like_net structure in developing an intelligent expert control system to achieve optimum grinding by regulating the parameters of the grinding mill. The work establishes an appropriate theoretical background that helps to predict dynamic breakage characteristics of particle size distribution of materials, adequately supported by experimental data. Feed forward neural network has been employed in the work to adapt the dynamic breakage characteristics of the mill by imparting training using generalized back propagation learning method. The mill has been tuned with the trained parameters and no further training is required so long the desired output is unchanged, irrespective of the input particle size range. In the next phase of the work, using acoustic sensors sound of different mills are recorded and analyzed to identify the operating states of the machines. The rule base adopted from the mill has been designed with the help of the recorded information and implemented using Petri_like_net structure. In real time, the reasoning process generates inference based on the deviation of acoustic signal from the recorded value, which is subsequently converted to control action to achieve optimum grinding. The time of reasoning is compatible with the particle residence time.