Mine planning is one of those vitally important but highly frustrating disciplines. The need to take into account geological data, available on-site capabilities, production costs and the ebbs and flows of the global commodity market mean that data has to be constantly updated in order to guarantee the safest, most productive setup over the short, medium and long-term.
Ever-shifting data results in scenarios having to be constantly remodelled, often with information that’s incomplete or based on estimates from past performance, giving you models that don’t always accurately reflect reality at the given point. If you’re not careful this can lead to slippages in the production schedule and cost overruns, which no miner can afford at the moment.
Technological mix poses a challenge to mine scheduling
Thankfully, technology has enabled considerable improvements over the past several years. Today’s 3D models are much easier to manipulate and interrogate. They also allow you to change easily between an overall view of a mine’s operations and the geologic and chemical data of a specific segment, helping the planner maintain a complete picture.
But miners still encounter information bottlenecks caused by the need to convert one file type to another, or to transfer data from one standalone computer system to another. It is also common to find large analogue chunks amid the digital process chain. Entering data by hand into Excel spreadsheets is still a common occurrence and one that inevitably leads to errors.
Maptek, a mining-focused 3D modelling and design technology company, has put a lot of effort into digitising these process and integrating them into a single software package. A new software release is the company’s latest offering. Evolution 4.5 consists of new and improved versions of Evolution Strategy and Evolution Origin, which are both focused on short and medium-term cut-off optimisation, essentially, identifying and dealing with waste. The idea is that these programmes work in tandem with Vulcan, a 3D modelling software, to give you all the planning tools you might need in one place, with no need for advanced software knowledge.
“Spreadsheet-based products with optimisation engines using techniques like linear programming, dynamic or mixed integer linear programming, were traditionally used for reserving,” says Steve Craig, manager of mine scheduling at Maptek. “Setting up and generating results was incredibly time consuming…Maptek Evolution challenges the tradition.”
The financial upside of waste disposal
Cut-off grade optimisation is an often undervalued aspect of mine scheduling. This is a little surprising given the high strippage ratios that many types of mine produce, and the fact that haulage can account for around 45% of a mine’s operating costs.
Evolution Strategy is designed to generate cut-off grade optimisation policies. Essentially it uses production data and information on the quality of each parcel of material to determine the most appropriate path for that material. For example, it can use commodity market and cashflow data to calculate a cut-off grade, then automatically categorise materials based on where they fall on that scale. These materials can then be organised and transported to the right location.
“If a block has a grade which is below the heap leach cut-off grade then it becomes waste,” explains Craig. “If above the heat leach grade but below the tank leach grade, then it becomes heat leachable. If greater than the tank leach grade, it becomes tank leachable. And all these parameters can be changed on a periodic basis.”
The data on how much material is going through which process stream is accumulated on the system and can be visualised in a number of different ways. You can even apply estimated variable costs to each process and let the computer calculate a best and worst-case cashflow scenario. Craig believes that cut-off grade optimisation planning can lead to an uptick in project value of as much as 25%.
The right vehicle for the job
The Origin module generates detailed scheduling scenarios for haulage, waste landform and blending and is the tool with which the high-level policies developed on Strategy are carried out. For example, you can use it to allocate certain vehicle types to particular materials or areas of the mine, taking into account things like their speed and fuel consumption and the gradient, rolling resistance, and maximum and minimum speed of the route they have to take.
In the latest update, you can use Origin to automatically apply a haulage schedule to a network you’ve designed and imported from Vulcan, its 3D modelling counterpart, as well as to automatically assign waste deposits to the most economic location within a dump based on metrics, such as fuel cost.
“Mines are complex environments and scheduling production is just as complex,” says Craig. “Planners must consider cut-off grade, route and equipment allocation, cycle times, fuel burn and waste dump locations. There’s universal benefit in being able to present an integrated, holistic 3D view of a mine site by simultaneously displaying multiple models, waste dumps, haul networks and topography.”
Evolution is so-called because of it is based on an evolutionary algorithm, which is inspired by natural processes such as reproduction and cell mutation. Essentially, as the algorithm runs it learns from huge amounts of operational data saved up from other mining projects, improving its effectiveness through a process of selection, just like we see in evolution. Evolution should in theory get better and better as it tweaks itself to match the multiple objectives of the user, an exciting thought for the role that machine learning could play in mining's future.
“Classical techniques rely on manipulating a single solution,” Craig said in a 2015 blog post. “To speed up the calculations, instances of the same problem can be resolved using more computers or parallel processing. However, all we get are more solutions of the problem in less time, with no improvement in quality. This is the beauty of using evolutionary algorithms. We can run with a population of solutions in parallel.”