“M-INFERENCE”: A SOFTWARE PACKAGE FOR RAPID REFERENCE CLASS FORECASTING FROM COMPLEX MINE WATER GEOCHEMISTRY DATASETS

The United States is heavily reliant on imports of several mineral commodities that are vital to the Nation’s security and economic prosperity. A policy directive included in Executive Order 13817, signed December 20, 2017, “A Federal Strategy to Ensure Secure and Reliable Supplies of Critical Minerals” requires that the United States streamline leasing and permitting processes for critical mineral development in a safe and environmentally responsible manner. A major existing hurdle for successful permitting, and therefore, establishment of new critical mineral mines is the ability to design mining operations, control technologies, and reclamation plans that are demonstrably able to meet environmental criteria. The best existing technology for making this demonstration uses forward-looking hydrogeochemical models of the mine, with parameters defined from first principals, and, largely, bench scale testing of site-specific materials. Model construction and parameterization takes years to complete, and cost millions of dollars. To the address this issue, MineraLogic LLC proposes to develop a software package (working name of “M-Inference™”) that transforms the complex geochemical mine water data that is routinely collected by mining companies into actionable knowledge that can used by mine developers, operators, and the government to facilitate efficient mine design and permitting. This software package will include a built- in water chemistry database, a front-end user dashboard, and algorithms for geochemical speciation and statistical routines. Consistent with DOE specifications, M-Inference will allow users to compare modelled outcomes and operational data with an internal dataset, identify likely future performance of engineered systems, integrate operational data with geochemical models, visualize chemical trends and relationships, and elucidate underlying geochemical processes. Phase I tasks to be conducted in the development of M-Inference will result in a working prototype: formulation of a trial database; integration with geochemical speciation code; creation of code for executing algorithms for descriptive statistics and statistical process control charting; and design of a front end user dashboard. The goal for the Phase I project is to produce a working program with a database for one type of commodity deposit as a minimum viable product for this technology. The theory underlying this proposed approach to prediction is adopted from “reference class forecasting”, which has shown to be highly effective in forecasting complex scenarios in finance and engineering applications (among others), but, to our knowledge, has never been applied to prediction of hydrogeochemical systems.