Segregation measurements in granular mixture with AI

Abstract

Segregation is a significant factor that can affect the uniformity of a mixture. In order to address and mitigate segregation, accurate data is necessary to characterize particles throughout the process. However, obtaining data in a manner that does not impact the mixture can be challenging, particularly in non-laboratory settings where conditions may not be as controlled.
In laboratory settings, stable lighting and coloured particles can be used to aid in differentiation. However, this approach may not be feasible for all materials. Therefore, the development of a tool that can identify particles without the need for colouring or consistent lighting is highly valuable.
This study focuses on a common mixture found in blast furnaces, consisting of coke, pellet, and sinter. The similarities in colour schemes and overlapping sizes of coke and sinter present challenges for particle recognition. Unstable lighting further complicates the differentiation process.
The proposed method is applied to a mixture containing all three components to demonstrate its capabilities. This method effectively distinguishes between the various particles, providing information on both the material and area of each particle. The measurement data is utilized to assess material and size segregation within the mixture. Material segregation is evaluated using a square grid and ring configuration, while size segregation is determined by categorizing particles into three groups based on diameter, including Feret diameter and area-equivalent diameter.