top of page
  • Writer's pictureLeonardo Lara

Development of a deep neural network and a PSO algorithm to predict ore hardness using X-ray diffraction and atomic emission spectroscopy

Our latest paper on the application of Deep Neural Networks (DNN) and Particle Swarm Optimization (PSO) to geomet models and predict Drop Weight Index (DWi) and Bond Work Index (BWi) from mineral samples using ICP-AES and XRD has just been published by Thiago Augusto de Almeida and Marcos de Paiva Bueno in Minerals Engineering journal.


The results for the DNN and PSO are very promising for both DWi and BWi:


DNN Model:

  𝐑² = 𝟗𝟗.𝟓% 𝐟𝐨𝐫 𝐃𝐖𝐢 

𝐑² = 𝟗𝟗.𝟖% 𝐟𝐨𝐫 𝐁𝐖𝐢


PSO Model:

𝐑² = 𝟗𝟗.𝟒% 𝐟𝐨𝐫 𝐃𝐖𝐢 

𝐑² = 𝟗𝟗.𝟕% 𝐟𝐨𝐫 𝐁𝐖𝐢


These high accuracy rates demonstrate the potential of DNN and PSO in enhancing predictive geometallurgical modeling in comminution projects instead of conventional regression methods. Our models were tested against a real project dataset, yielding results that closely align with the reference measured data.


The full article is available for download here:

Development of a deep neural network and a PSO algorithm to predict ore
.pdf
Download PDF • 8.25MB

4 views0 comments

Comments


bottom of page