Abstract

If variability of input data for rheological measurements is not adequately included, their associated uncertainty and subsequent modelling can be underrated. Mineral pulp rheology determination is commonly done through triplicate tests, with such variability reported as multiples of a standard deviation, with the potential for underestimation. In the present work, a novel statistically-based methodology for the estimation of uncertainty in the rheological characterization of mineral suspensions -and other parametric models- is proposed. From the variability of the experimental measurements and the analytical propagation of errors, a set of rheological profiles are generated using Monte Carlo simulations within a variability frame. The corresponding inverse problem for curve-fitting is solved individually, resulting in distributions of fitted parameters, which were statistically-analyzed to obtain representative values for both the parameter and its true variability. The methodology proposed herein has been used to explore the applicability and limitations of the Herschel-Bulkley and Bingham models under specific experimental and data analysis protocols, where the relevance of including low-shear-rate measurement points or yield stress measurements using alternative methods is exposed. Additionally, we present a case study on the effect of the concentration of NaCl on the rheological response of synthetic tailings consisting of quartz suspensions doped with kaolinite, bentonite and kaolinite-bentonite blends, using the proposed methodology with a concentric cylinder rheometer. Results show predominantly decreasing trends in yield stress as salt concentration increases, with non-monotonical behavior and strongest variability associated to the quartz-bentonite blend.