Python: Optimize Flotation Plant Operation Using Machine Learning - Business Edition

Python: Optimize Flotation Plant Operation Using Machine Learning - Business Edition

Introduction

Metals R' Us is a mining company, and now we have the data of their flotation plat. This flotation plant is used to purify Iron; the engineers will feed the original materials into the plant, mix the materials with some chemicals, shoot air bubbles toward them, and get a material with a higher Iron concentration. However, the quality of the purifying process is inconsistent, so we are now analyzing the dataset from the company to identify the possible reasons.

No alt text provided for this image
source: https://www.pexels.com/zh-tw/photo/2101135/


Key Findings

After looking into the data set, we found that we should operate the flotation plant differently with different percentages of Iron and Silica in the original material. I will demonstrate the instruction for each type of ore at the end of the article.

Data

The dataset provides information on the following categories:

  • Timestamp
  • The original material (% Iron Feed, % Silica Feed)
  • What they mixed in the original material (Starch Flow, Amina Flow)
  • The mixture's information (Ore Pulp Flow, pH, Density)
  • Processes in the flotation (Column 1-7 Air Flow and Level)
  • Final material (% Iron Concentrate, % Silica Concentrate)

The goal is to have a higher percentage of Iron concentrate.

Data Analytics

First, I try to see if any factors directly affect the final Iron concentrate percentage. Unfortunately, I did not see a trend in any of the factors after visualizing the relationship between each column and the final Iron concentrate percentage. Therefore, I decide to use machine learning techniques to analyze the data.

No alt text provided for this image

As we are interested in getting a high Iron concentration at the end of the process, I will use only the best performance result with the machine learning technique. Our machine-learning technique will group the data with similar characteristics together. After observing each group, I discovered some interesting results.

Result

I found that we should tailor the flotation process for different types of ore.

This is the process instruction for some of the material types:

Ore: Low Iron + High Silica

Starch: Medium-High

Amina: Medium-Low

Flotation 1- 5 Air Flow: Medium

Flotation 6-7 Air Flow: High

Flotation 1-3 Air Level: Medium

Flotation 4-7 Air Level: Low


Ore: High Iron + Medium-low Silica

Starch: Low

Amina: Medium-high

Flotation Air Flow: Low

Flotation 1-3 Air Level: Medium

Flotation 4-7 Air Level: Low


Conclusion

Before Metals R' Us process the flotation plant, they should examine the ore and refer to the following chart to identify the type of ore and the suggested operation metrics.

No alt text provided for this image

Thank you for reading my articles. Just a quick reminder that I am looking for a full-time position as a data analyst now! If you are hiring or know anyone hiring, please message me on Linkedin or email me at zzonach@gmail.com to further discuss it! Any information or support will be well appreciated!

As a metallurgist involved in mineral processing, I’m curious to know which valuable iron ore this plant is currently processing. Understanding this would help me better comprehend the clustering analysis you performed based on mineralogy. Thanks!

Like
Reply
Aryanto M.

Remote only | Data Scientist Lead | AI Strategy & Engineering | Recommender Systems, NLP & CV | Delivered almost IDR 2.0 Trillion in Business Value via ML/DL

2y

Nice reading, thank you Zona Chiang!

To view or add a comment, sign in

More articles by Pei-Yu Chiang

Others also viewed

Explore content categories