Dragline Automation System: Optimal Excavation Sequencing - ACARP Project Report C23033

Report Abstract Dragline excavation sequencing describes the complex problem of determining the sequence of tub positions and the associated material movement task for each tub position for dragline strip mining. This project has examined computational methods for computing excavation sequences. It is thought that these methods can be the base for an operator decision support that instructs the operator where to position, what material to take, when to move, and where to move to. The project has been conducted by the Smart Machines Group at The University of Queensland and the SME Mineware. This report describes the progress made and identifies a pathway toward commercialization. The methodology of the study has been to: · Understand the requirements of an operator assist for excavation sequencing from the perspective of dragline operators, supervisors, and engineers; · Characterize the decision making processes for excavation to understand the functional elements of the sequencing problem and how they interrelate, and to identify sub-problems that can be solved; · Develop a software framework for providing the functional capabilities of excavation sequencing; · Apply identified strategies to address excavation sequencing for a steady state excavation scenario and make comparisons with actual operation. Dragline excavation sequencing is a complex problem. There are many factors that influence the decision to excavate material from a particular position: the volume of material that can be reached and how the volume is distributed; the efficiency of digging the reachable material; the average cycle time for moving the material to spoil; the available spoil room and how the available room is distributed; and the impact of future excavation requirements. These factors result in a large decision space with competing objectives and constraints. The approach taken for this project was to impose structure to the excavation task through the prescription of material assignment to a position (i.e. layered excavation strategy describing target profiles for the block walk-up sequence), prescribing an order of material removal, and applying a front-to-back spoiling strategy to minimize swing angle. These strategies were considered appropriate for the test strip and could be augmented to include strategies appropriate for other strip geometries. The imposed structure for the block excavation represents a template for excavation based on actual execution (e.g. percentage of block volume removed at each walk up position, block length, block shape) and established strategies for removing material to spoil. This structure reduces the dimensionality of the decision making process to one that seeks to identify the walk-up positions that will maximize the lineal advance of the strip. The excavation sequencing algorithm developed has been applied to data collected from Dragline 17 at BMA's Peak Downs site. A strip was chosen to provide a steady state operation for approximately 200m of lineal advance. In this section there were no ramps, corners or discontinuities in the target profile. This allowed several complexities of general operation to be marginalized. The dragline was operating on an in-pit bench, initially constructed from an extended key process. The overall excavation task was to excavate the bench and residual material left in the key. The algorithm was compared in its performance to operators and out-performed them. Not surprisingly operators will make different decisions based on experience / prejudices and the quality of their perception of the required current and future material movements. To arrive at a decision requires trading off various objectives and constraints. For the test strip the operator cycle data shows a mean swing angle of 85 degrees with a mean cycle time of 73 seconds. The swing angles recorded ranged from 10 degrees up to 200 degrees with a standard deviation of 32.5 degrees. The source of this variation is unknown but it is suggested that the variation is typical of what might be expected from multiple operators across different shifts. When the algorithm was applied to the same data the mean cycle time was reduced to 51 seconds. An attribute of effective sequencing is better utilization of spoil. The computed solutions both produce spoil that sits against the design spoil shell representing minimum movement of material. If a decision support tool were able to sufficiently guide operators to deliver better sequencing it is estimated that an overall 10% productivity improvement could be achieved. Sensitivity analysis showed that a further 10% productivity improvement could be achieved by moving from a 15m to 25m blocks. A planned Phase 2 of this work will verify the value proposition and augment the existing dragline sequencing capabilities by extending the ideas and deploying to a production dragline for evaluation.