Rapid updating of resource knowledge with sensor information

Abstract

Resource models are generally constructed from directly observed data (e.g., grades of drill cores) that have relatively high accuracy. However, the accuracy of resource models is therefore limited by the scale on which the data are collected. As mining progresses, more information becomes available on different scales from various types and sources of data (e.g., blast hole samples, sensors on drill rigs, conveyor belts and draw points). This continuous stream of production data can be used to update resource knowledge in near real-time. The ensemble Kalman filter has been successfully applied to update resource and grade control models. However, due to the Gaussianity assumption, the ensemble Kalman filter must be combined with some kind of Gaussian transformation, such as a normal score transform. Multi-Gaussian transformations can yield better results in terms of reproducing relationships between multiple grade variables. This poster presents a case study demonstrating the application of the ensemble Kalman filter and the projection pursuit multivariate transform for sequential updating of multivariate geostatistical models.