Dam break analysis for BRDA 5 at Worsley Alumina Refinery

Abstract

South32 Worsley Alumina Pty Ltd operates several bauxite residue storage facilities at their refinery in Western Australia. The bauxite residue slurry is thickened and pumped onto the storage facilities which are constructed on an upstream basis using compacted earthfill embankments. BRDA 5 is the newest of the residue storage facilities, established in 1999, and has a current maximum height of embankment slope of 42 m, of which a compacted earthfill starter embankment makes up the first 22 m. Since 2010, the residue has been treated after deposition using amphirols, which assist in dewatering and compacting the residue. As part of their residue management commitments, Worsley undertook a dam break analysis entailing a review of all available geotechnical testing data, a fault-event analysis to evaluate the probability of removal of support to the residue through slope instability, overtopping, geotechnical piping or failure of a buried structure leading to liquefaction of the residue and a dam break. To model the flow of the liquefied residue, pilot-scale dam break testing was carried out from which estimates of the rheology of the residue were made. The rheology has been incorporated into a flow slide model that applies stream power entropy theory to define the energy changes due to viscous effects and frictional resistance. Benchmarking of the results against modelling using computational fluid dynamics has been made. This paper presents an overview of the dam break assessment process and results that covers a liquefaction potential assessment, a semi-quantified risk assessment based on a fault-event analysis approach, rheological testing of liquefied consolidated residue in a specially developed box rheometer and probabilistic modelling of the dam break, the outflow hydrograph, and the flow inundation areas. It is concluded that incorporation of rheology representative of liquefied consolidated residue in dam break modelling results in significantly reduced prediction of inundation area.