Open Cut » Overburden Removal
The ability to automatically classify and report swing cycles according to their limiting motion (swing, hoist or drag) is an essential capability in order to flag potential dragline performance losses and drive continuous improvement efforts.
This project uses data generated by dragline performance monitoring systems to develop a cycle classification algorithm capable of functioning without the need for continuous manual input or manual calibration. Two algorithms have been developed to classify dragline cycle dependencies and potentially dig modes. The first of these uses a simple graphical approach based on coincident limits defined by the locus of points where the swing and hoist motors work at full capacity for equivalent amounts of time. The algorithm is capable of distinguishing three cycle dependencies: swing, hoist and drag-limited cycles.
Over 200,000 cycles of dragline performance monitoring data from Dragline 302 at Curragh mine were retrospectively analysed using the coincident limit algorithm. On the basis of the cycle dependency breakdown and the range of hoist distances, the algorithm was able to discriminate between four of the eight dig modes employed at Curragh mine. These are descriptors that identify different types of dragline digging operations, for example: key cut, blocks and chop cut.
The coincident limit algorithm was coded and deployed on the PegasysTM software system of DL302 at Curragh mine. Some 75,000 cycles were logged. 33% of these cycles were classified as being hoist-limited, which demonstrates the need to take into account hoist limited cycles when benchmarking dragline performance. 2,000 of these cycles were independently validated against the algorithm developed using MATLAB.
The coincident limit algorithm cannot discriminate hoist-then-swing and swing-then-hoist cycles. To overcome this problem, a second algorithm was developed in MATLAB using a feed-forward artificial neural network. Data from 40,000 cycles were used to train and validate this algorithm. A classification error rate of 2.7% was achieved. This algorithm was not implemented in the field due to time constraints.
Traditional measures of dragline productivity consider the number of BCMs per operating hour. This measurement takes no account of the percentage of hoist limited cycles. Just as haul truck work is calculated in tonnes × equivalent horizontal kilometers, development of coincident limit graphs opens the possibility of measuring dragline work as BCM × equivalent swing angle. This has the potential to be applied to the benchmarking of dragline performance across different operations where different pit and block geometries influence the proportion of hoist-limited cycles.
The potential exists to develop a predictive algorithm to provide dragline operators with advice on how to improve dragline dig sequencing and when to walk the machine.