Mapping the future: Professor Hyam Rubinstein discusses the MineOptima project

16 January 2018 (Last Updated January 17th, 2018 10:42)

The MineOptima project was developed by Professors Thomas and Rubinstein of the University of Melbourne to apply their work on network optimisation and microchip design to the planning and design of the most efficient mines. Julian Turner talks to Professor Rubinstein about their work.

Mapping the future: Professor Hyam Rubinstein discusses the MineOptima project
Level development design done for Newmont using PUNO showing footwall drive and ore drives. Credit: Courtesy of MineOptima

Professor Hyam Rubinstein − along with colleagues Doreen Thomas, Nick Wormald, Marcus Brazil, David Lee and Peter Grossman − has spent 20 years developing MineOptima, optimisation software tools that maps networks of mining tunnels deep underground so that ore deposits can be accessed more efficiently.

Professor Rubinstein is a fellow of the Australian Academy of Sciences and was president of the Australian Mathematical Society.

Julian Turner (JT): Where did the idea come from to use software and algorithms to map mines?

Hyam Rubinstein (HR): A young mining engineer at Olympic Dam at Roxby Downs in South Australia asked a question – how deep should a shaft which was being developed to access this huge deposit be? Should the shaft be shallower with the ramps sloping down to the deposit or deeper so the ramps were more horizontal?

We were able to formulate this as a new mathematical model, where the edges of a graph represent the access network of the ramps and shaft. There is a key gradient bound on the ramps for navigability and also the costs of the edges depend on development, but also haulage over the lifetime of the mine. Our initial software tool was called Underground Network Optimiser (UNO).

JT: MineOptima incorporates two software tools: DOT and PUNO. How do these work?

HR: The Decline Optimisation Tool (DOT) looks at a network of declines and ramps accessing a number of ore zones at specified levels. Our algorithms offer methods of finding networks that satisfied gradient and turning circle constraints, and minimise network development and haulage costs. Barrier avoidance is another important consideration as the access tunnels need to avoid or minimise contact with faults, old workings and the ore zone itself.

From around 2005-2010 we developed the Planar Underground Network Optimiser (PUNO), which finds an optimal design for the access infrastructure close to the ore zone, whether that be footwall or hanging wall drives, ore drives or connections to vent raises. DOT and PUNO are designed to work in tandem, so that the output of PUNO becomes the input of DOT, with draw points and tonnages for the design produced by PUNO being the starting point for the DOT design.

JT: What advantages does MineOptima software offer compared with rival mapping solutions?

HR: MineOptima software builds on the theory of shortest networks with geometric constraints, which we developed over a 20-year period. The components of the network are represented as edges of a graph, where an edge could represent a helical ramp.

The optimal network is constructed starting at the bottom level by a process called dynamic programming. There are various steps which are computationally difficult, so powerful heuristics like simulated annealing aid in rapid evaluation.

There are two important advantages to using our software over traditional computer-aided design (CAD) tools. Firstly, our software produces designs that minimise cost, taking both development and haulage into account. This is very challenging to achieve, even for experienced mine planners, since the best network can be located so that the distance to larger deposits decreases, but the distance to smaller deposits increases. We estimate for complex designs a saving of around 10% in costs over traditional methods using our software.

The second advantage is that by using our software mine planners can run multiple scenarios very rapidly. For example, they can look at different mining methods, as well as different cut-off grades and level spacings, which can lead to quite different access designs. For feasibility and strategic designs, a planner can run up to 100 designs in a week and then schedule these to compute net present value.

This type of sensitivity analysis means that in the future, mine planners can have more confidence that the choices they make are the best under a range of commodity prices.

JT: How did you to apply your work on network optimisation using microchips to the much larger 3D mining environment?

HR: Our main challenge was to understand and correctly model the geometric constraints involved in mining as compared to classical shortest networks, such as those that occur in microchip design.

For microchips, wiring traditionally runs in two orthogonal directions − although recently more complex designs are used but with similar principles. Finding the optimal layout of a microchip has been shown to be very computationally difficult, as the size of the microchip grows.

For mining, the constraints lead to much harder computational issues. So we had to do a lot of research and analysis to ensure that our algorithms were as efficient as possible, so that large mining access networks could be handled.

JT: How does MineOptima use geological information from exploratory drilling and algorithms to explore and locate access tunnels?

HR: MineOptima software is designed to start with the output of stope design software, but can be used with a geological block model. PUNO works by connecting up the stopes at a level by an access network optimising development and haulage costs.

DOT starts with a choice of draw points and tonnages on each level. DOT has the ability to choose one from a group of points, so that the best possible draw point can be found on each level. This can lead to substantial savings and efficiency over predetermining the draw points. DOT can work from a geological block model by consolidating the tonnages without knowledge of the detailed stope design. The block model is constructed from exploratory drilling information using geostatistical techniques such as kriging.

JT: How will MineOptima help mining companies access and extract ever-deeper ore deposits?

HR: As mining goes deeper there will be a stronger push for automation. Ventilation is critical for conventional mining and the temperature gradient means cooling is also vital for deep mines. For all these reasons, having the most efficient and cost-effective access network, involving ventilation design early, is crucial to an economic outcome. Current CAD tools cannot achieve such designs.

JT: What next for MineOptima and how close are you to achieving your goal of creating ‘200 access network designs in two weeks’?

HR: We created 70 designs in four days on a large underground deposit with complex access design features. DOT, PUNO and UNO were all used for this due to the interesting distribution of ore zones.

In 2010, we set-up MineOptima as a spinoff from the University of Melbourne, to commercialise DOT and PUNO and entered into licensing agreements with several major software suppliers. We also were very fortunate to get sponsorship, ideas and projects from Rand and Tribune resources.

During this period, we developed UMOID, a software tool which integrates the position of vent raises with an optimal access decline. In 2017 RPM Global bought MineOptima and the rights to our software tools, which are available through a number of suppliers.