Initial applyication of a dynamic, mechanistic mill model

This research represents a shift away from empirical models that lock the industry into 50-year old inefficiencies with little response to operating variability. The mechanistic and intrinsically dynamic mill model under development steps up to a new paradigm using the latest computing capabilities and physical understanding of breakage and transport to transform predictive capabilities for design, optimisation and on-line control with variable feed. The milling process is broken down into discrete sub-processes, each of which can be modelled as an object in the whole model and in the model code. These objects can be upgraded or exchanged without having to rewrite or restructure the remainder of the code. Sub-processes include: feed into mill; transport along mill; presentation to impacting events; range of impacting forces; probability of being impacted in different zones in the mill as a function of particle size and density; suspension in slurry; and discharge out of the mill. The unit and processes surrounding the mill (feed and recycle conveyors, slurry pump and pipe, cyclone and recycle crusher) have also been modelled, in order to incorporate the effect of recirculating load coming from a cyclone or a recycle pebble crusher, respectively for ball mills or SAG mills. Each unit is modelled dynamically and PID controllers have been added to control pumping flowrate from sump and water addition for cyclone feed dilution. The paper demonstrates the predictive capability of a dynamic, mechanistic model and is used to test the sensitivity to the sub-processes, comparing different formulations and highlighting the aspects that require further development. Validation on pilot tests and application to industrial data is presented. This approach is currently undergoing industrial trial on sites, for use in on-line model-informed process control. Considerable further development is required to incorporate the full mechanistic structure planned for this model.