SPDE for boosting simulation performances
SPDE is a two-year R&D consortium aiming at developing a new engine for estimation and simulation adapted to the specific context of Mineral Resource Estimation that will offer a quantum leap in performance. The key ingredient to allow that quantum leap is based on solving Stochastic Partial Derivative Equations, hence the name given.
Anyone involved with today’s mining industry understands that the sector is facing very tough challenges. And it is under enormous pressure from due cost control and budgetary management constraints that innovation and ingenuity must still find their way to propose new ways of tackling traditional issues. Mineral Resource Estimation (MRE) is no stranger to that conundrum and their practitioners all know the multi facets of the game: integrate more and more data, boost productivity, and come up with an answer fast and at the same time enriched with an assessment of the uncertainty that can be attached to that answer. All of that in a fraction of the time they used to have at their disposal and less resources to double check…
On the issue of quantitative risk assessment for MRE, the geostatistical framework (via the use of conditional simulation) has been providing a technically suitable answer to the mining industry for some years now.
Despite the inherent ability of simulation to tackle risk analysis a series of factors have prevented their incorporation to the mainstream chain of information processing prevalent in MRE. The factors that have impeded simulation’s progress vary in nature from a series of cultural or regulatory blocks to more technical elements. The main technical block has undoubtedly been that of performance: it still takes an awful length of time to produce simulation realisations in sufficient numbers to allow meaningful risk analysis to be performed. A problem made worse when the size of the dataset increases as tends to be the case nowadays with the emphasis put on optimising the use of all available data.
It is in that context that Geovariances in partnership with the Center for Geostatistics from Mines Paris Tech launched a new consortium in December 2016 meant to develop a new engine for estimation and simulation adapted to the specific context of MRE that will offer a quantum leap in performance (drastic reduction of time required to produce large numbers of realisations, using large numbers of data). The key ingredient to allow that quantum leap is based on solving Stochastic Partial Derivative Equations that are linked to the random functions at the heart of the geostatistical paradigm, hence the name given to the consortium SPDE.
The consortium boasts 4 sponsors (Eramet, Areva, Newcrest Mining and Kinross), with a couple more due to join in July (Anglo American and BHP) and other giants (like Rio Tinto) having expressed very strong interest to joining in.
The consortium will amongst other things:
1. Tackle the “Big NDATA, NGRIDSIZE, NSIMUS” problem;
2. Facilitate and fully integrate the use of local constraints (in particular orientation and anisotropy ratios);
3. Work in multivariate space.
The end result will be a software prototype that performs estimation and most of all simulation in large numbers, with exceptional computing time performances.