Characterisation of dewaterability from equilibrium and transient centrifugation test data

Abstract

A knowledge of the properties that quantify the rate and extent of particulate suspension dewatering through sedimentation and consolidation is critical to the prediction of dewatering performance for industrial processes such as gravity settling, centrifugation and filtration. The strength of a network of particles in compression is quantified as a compressive yield stress or modulus, Py(ϕ). The rate of dewatering is inversely related to the hindered settling function, R(ϕ), which is a measure of hydrodynamic resistance. Dewaterability characterisation over a broad range of solids concentrations is a challenging task requiring tests including batch settling and pressure filtration. Unfortunately, for many systems, these tests are prohibitively slow and an alternative such as centrifugation is useful in accelerating the characterisation. This paper validates algorithms for the characterisation of dewaterability from centrifugal sedimentation data obtained in a laboratory centrifuge with interface height detection and provides guidelines for obtaining accurate results over a broad range of solids concentrations.Typical dewaterability data for a flocculated industrial mineral slurry and a biological sludge are used as inputs for the numerical prediction of centrifugal sedimentation behaviour. Equilibrium centrifugation at a range of rotation speeds has been simulated and Py(ϕ) data determined using a method described by Buscall and White (1987). The result is shown to be accurate to within 10% of the expected values. Transient centrifugation data has been simulated and R(ϕ) determination using a new algorithm is shown to be accurate over a broad range of solids concentrations extending beyond the gel point, provided a new data rejection criterion is utilised. In addition, errors at low concentration in R(ϕ) data, caused by dilution, are described. It is shown that currently available analysis techniques are unlikely to quantify these errors.