Open Cut » Overburden Removal
The objective of this project was to develop prototype software that would allow a mine to determine the optimum dragline excavation sequence, ie. where the dragline stands, where it digs from and dumps to, and when and where it moves. Currently, many of these decisions are made by the dragline operator or supervisor "on the fly". Operations can try different sequences and monitor productivity or use dragline planning software to compare a number of "what if" scenarios but this is very costly in terms of time, lost opportunity and the potential to make things worse. As with all research software, the product is not of commercial quality although it is sufficiently robust and documented for testing by knowledgeable engineers.
This program is based on the genetic algorithm where the behaviour of the unit to be optimised is described by a sequence of control values. A population of possible dig sequences is maintained that are progressively improved as new sequences are created. Sequences that are not as good as those in the rest of the population are progressively removed. Genetic algorithms are based on an assumption that some systematic changes to the control sequence will make improvements to the performance. They have the advantage that they can supply intermediate solutions as the algorithm is run and they can be restarted to further improve the solutions found.
However the genetic algorithm cannot be applied directly to dragline excavation as nearly all random changes and sequence swaps applied to the control sequence result in infeasible behaviour. Also it is not possible to use the typical genetic algorithm technique of starting with random control sequences for the initial population, as these are invariably not feasible. To overcome this, the optimisation problem is divided into "tuneable" stages that can bias the random choices so that good decisions are made more often. New sequences are checked to ensure they are feasible and the completion time is calculated. If the time is sufficiently short, the new sequence becomes part of the population of good sequences. This produces progressively improving dig sequences that approach the optimum and the best sequences found can be examined when the optimisation is halted.
The optimisation algorithm is generating complete excavation sequences for a production scale dragline. The results from four variations of side-cast operations are documented and while the resulting excavation sequences are feasible, further work is required to determine how close to optimum they are and then to fine tune the algorithm's control parameters to improve the solution. Although only a side-cast operation has been tested, the nature of the optimisation means that more complex operations should be just as feasible. Ultimately this optimisation algorithm could be integrated with other technologies such as dragline planning software, digital terrain mapping and dragline monitoring systems with GPS to create a very powerful dragline optimisation system.