Inference of Geological Material Groups Using Structural Monitoring Sensors on Excavators

Abstract

In mining, correctly characterising geological grades is very important as it directly relates to ore quality assessment and downstream processing. Significant effort has been placed into creating the geological block models for the mine site, through site exploration and in-lab assay analysis. Yet the blasting, digging and various processing can cause non-negligible movement of material and invalid prediction of material content due to the now obsolete model. On the other hand, it is well known to excavator operators that digging effort is closely related to the hardness or lumpiness of the material underneath, and therefore this may be exploited to indicate the material type post blasting.

This paper proposes a method that can automatically infer the geological material types of mined material during excavation at the digging location by applying machine learning methods to the force, energy and kinematic information collected from sensors mounted on the diggers. Therefore, the digging equipment is being used as a sensor for this purpose. Conversely, we also show how knowledge of material type can lead to accurate prediction of digging effort category. Further, per bucket material information can be utilised throughout the material movement pipeline. A case study was conducted in a test region at Pilbara iron ore deposit situated in the Brockman Iron Formation of the Hamersley Province, Western Australia. Initial results show strong level of inter-dependency between sensor measurements and excavated material type, demonstrating the potential of material inference at the bucket level.

L. Liu and M. Balamurali—These authors contributed equally to this work.