CIM Bulletin, Vol. 97, No. 1084, 2004
P. Radziszewski and S. Caron
Industrial practice in mineral processing shows that an important problem in understanding and monitoring in-mill parameters is due to the lack of fundamental knowledge and appropriate sensors. However, over the last decade, DEM modelling has shown an interesting potential in describing the fundamentals of grinding mill behaviour, in improving the understanding of internal grinding mill dynamics, and in developing solutions to industrial practice. With the objective of bringing this emerging technology into the concentrator control room to assist operators, this paper aims to describe the main elements needed to develop on-line mill performance monitoring. These main elements are the on-line interface, the performance criteria, and the underlying models of the comminution process.
Since the end of the 1990s, COREM has been developing a real-time platform through its Process Optimization Project System (POP) where one of the goals is to provide a range of soft sensor information for on-line process monitoring, diagnosis, and control.
Further, recent trends have described internal dynamics of mills using DEM as a function of mill dimensions, lifters profiles, charge volume, and rotation speed. Keeping in mind the objective of bringing this software technology into the control room, two main avenues present themselves: (1) define a ‘look-up’ image library of each charge motion scenario and a search engine to find the appropriate image describing the current operating conditions, or (2) simplify the DEM model to allow real-time simulation of the charge motion as a function of mill operation.
The current project aims to address the second avenue and will profit from previous work on charge motion simulation (see Figure) that shows the possibility of simulating charge motion in close to real-time. The trade-off to increase calculation speed (minutes as compared to hours) is to decrease precision. However, simulation results show gross power prediction confidence limits of 12%, which appear to have a good potential for on-line applications. With the availability of the POP real-time platform, the current project aims to study the possibilities of using a simplified charge motion simulator for on-line applications.
For a mill, the notion of performance is defined as the cost (energy and wear) of grinding a ton of ore to a desired size distribution which can define a cost function. The eventual use of this cost function in an on-line application could lead to a monitoring structure where the mill cost function performance index is a function of mill-manipulated control variables. Diagnosis could be an operator-driven process where the operator would use the on-line simulator to explore improvements to mill performance behaviour. This could be achieved using a design of experiments and response surface analyses to determine the best mill control action. In such a scenario, changes in mill-predicted performance are determined as a function of changing physical parameters of the mill due to time (liner and lifter wear, media wear, ore characteristics, and charge volume) and control choices determined by the mill operator. If the control choice decreases the cost function performance index below its actual value, the mill control variable can be changed.
In the short and medium term, the development of this on-line real-time simulator could address some intermediate measures of mill performance, such as mill charge shoulder and toe positions, liner and media wear, and mill charge volume. In this particular case, the main focus would be to validate charge motion and tie it to a given DCS through the POP system.
