A hybrid geometallurgical study using coupled Historical Data (HD) and Deep Learning (DL) techniques on a copper ore mine

This research work introduces a novel hybrid geometallurgical approach to develop a deep and comprehensive relationship between geological and mining characteristics with metallurgical parameters in a mineral processing plant. This technique involves statistically screening mineralogical and operational parameters using the Historical Data (HD) method. Further, it creates an intelligent bridge between effective parameters and metallurgical responses by the Deep Learning (DL) simulation method. In the HD method, the time and cost of common approaches in geometallurgical studies were minimized through the use of available archived data. Then, the generated DL-based predictive model was enabled to accurately forecast the process behavior in the mineral processing units. The efficiency of the proposed method for a copper ore sample was practically evaluated. For this purpose, six representative samples from different active mining zone were collected and used for flotation tests organized using a randomizing code. The experimental results were then statistically analyzed using HD method to assess the significance of mineralogical and operational parameters, including the proportions of effective minerals, particle size, collector and frother concentration, solid content and pH. Based on the HD analysis, the metallurgical responses including the copper grade and recovery, copper kinetics constant and iron grade in concentrate were modeled with an accuracy of about 90%. Next, the geometallurgical model of the process was developed using the long short-term memory neural network (LSTM) algorithm. The results showed that the studied metallurgical responses could be predicted with more than 95% accuracy. The results of this study showed that the hybrid geometallurgy approach can be used as a promising tool to achieve a reliable relationship between the mining and mineral processing sectors, and sustainable and predictable production.