USE OF STATISTICAL APPROACH IN TAILINGS FILTRATION: LEARNING FROM CLOTH FAILURES

Residue filtration is an increasingly important process in the alumina industry and its associated red mud generation. Tailings filtration is not a process that is exclusive to the alumina business, and lessons can be learned from other mining sectors.

This paper describes a methodology and platform where cloth failure in filter presses is leveraged as a mean of driving improvement. Briefly laying out the basic concepts behind tailings filtration using filter presses, the author introduces the problems often encountered in practice and the limitations of the fundamental theory (in helping to solve the former) due to complexity. A methodology based on statistical analysis and systematic empirical testing is then introduced, with the support of a case that spans two years of improvements in a Canadian mine site and the gains obtained.

The core of the paper explains the steps in implementing such a process. From establishing a statistically significant status quo, to diagnosing the failures using the data and finally launching an iterative, failure driven improvement process. In this process, multiple models of cloth and/or plates can be tested simultaneously, each with significant sample sizes. Each new generation of tests starts from the best performing technology of the previous generation, in what could be described as an evolutionary approach. The author explains how the data can be used to model a probabilistic reliability function, and to become predictive in a strategy to minimize unplanned maintenance, and cost. Lastly, a platform facilitating the implementation of the described approach in filtration plants is presented.