How can we estimate particle types from Mineral Processing data?
Dr Stephen Rayward

How can we estimate particle types from Mineral Processing data?

Summary

Particle liberation data is particularly useful as the basis for simulating mineral processing data.?Because of the conventionally held belief that obtaining such data is difficult, such an approach is seldom used.

This article provides an introductory discussion on whether it is possible to estimate particle liberation information directly from plant data. A YouTube video has been created that discussed this approach in more detail.

Introduction

Mineral processing is the process of taking composite particles, breaking them to ‘liberate’ the mineral from the particles, and then to separate the valuable mineral from the non-valuable mineral using processes such as flotation.

No alt text provided for this image
Figure 1 When particles are broken mineral is ‘liberated’.


But ‘liberation’ is not a binary process in that particles are composite and then at a particular size become liberated.

Instead liberation occurs increasingly as size decreases.

Near-full liberation is achieved by breaking the particles to the finest possible size.?But breakage requires energy and is an expensive process.?So a compromise is required whereby energy consumption is minimised whilst achieving a desired separation.

So from a technical viewpoint, the enthusiastic mineral processor engineer will wish to know precisely the level of liberation achieved and identify the optimal compromise between separation and breakage.?They then use a simulation system to provide guidance on how to optimise plant operation.

Obtaining liberation information is not at all straightforward.?Mineralogical systems are available but getting high quality mineralogical information is expensive, time-consuming and often difficult.?In general mineralogical systems are underutilised.

The author of this article developed numerous mathematical models to utilise mineralogical information (previous name Stephen Gay).?Because of the practical impediments to using mineralogical data Stephen leant toward the idea of estimating liberation data from the plant data itself.

This raised the question:

can liberation data be estimated directly from plant data?

Estimation of particle types

The answer to the above question is necessarily complex, and there is not a simple yes, no answer.

However there are mathematical strategies that can assist.

In this article we refer to a mathematical technique called Principal Component Analysis (PCA).

To my knowledge there is no specific published paper that discusses identification of particle types using PCA.?There are however 1000s of papers on PCA and a sprinkling of papers in Mineral Processing.?PCA is certainly more widely used in Geology, and Geostatistics.

PCA is related to the covariance; and is used in clustering or classification algorithms.

An understanding of classification algorithms leads directly to the idea of particle types.?Therefore, the idea of using particle types is both novel and obvious.

Action Plan

  • An overview YouTube video is available, and if you are interested in knowing more about this approach in more depth a Udemy course will be made available (if there is sufficient interest).
  • Of course the best way to learn the overall simulation and optimisation approach is via a ?3 day direct course, ‘Modelling and Simulation of Mineral Processing Plants’.
  • Contact [email protected] if you are interested in direct or online courses.

References


Yusuf Enes PURAL

Mineral Processing Engineer | Data Science and Artificial Intelligence Enthusiast

1 年

PCA is mainly an unsupervised learning method used for dimensionality reduction. I'm curious to know how it can be applied to predict particle liberation. Providing more details would help in understanding the specific approach. Also, I'm unsure about the reference to "plant data." Could you please clarify which data is being considered? In any case, it seems necessary to establish a connection between this data and actual liberation data to develop an effective model. I'd appreciate further insights into these aspects.

回复

要查看或添加评论,请登录

Mineral Processing Analysis的更多文章

社区洞察

其他会员也浏览了