A stochastic method for representation, modelling and fusion of excavated material in mining

摘要

The ability to safely and economically extract raw materials such as iron ore from a greater number of remote, isolated and possibly dangerous locations will become more pressing over the coming decades as easily accessible deposits become depleted. An autonomous mining system has the potential to make the mining process more efficient, predictable and safe under these changing conditions.ududOne of the key parts of the mining process is the estimation and tracking of bulk material through the mining production chain. Current state-of-the-art tracking and estimation systems use a deterministic representation for bulk material. This is problematic for wide-scale automation of mine processes as there is no measurement of the uncertainty in the estimates provided. A probabilistic representation is critical for autonomous systems to correctly interpret and fuse the available data in order to make the most informed decision given the available information without human intervention.ududThis thesis investigates whether bulk material properties can be represented probabilistically through a mining production chain to provide statistically consistent estimates of the material at each stage of the production chain. Experiments and methods within this thesis focus on the load-haul-dump cycle.ududThe development of a representation of bulk material using lumped masses is presented. A method for tracking and estimation of these lumped masses within the mining production chain using an 'Augmented State Kalman Filter' (ASKF) is developed. The method ensures that the fusion of new information at different stages will provide statistically consistent estimates of the lumped mass.ududThere is a particular focus on the feasibility and practicality of implementing a solution on a production mine site given the current sensing technology available and how it can be adapted for use within the developed estimation system (with particular focus on remote sensing and volume estimation).