Information and communication and chemical technologies

Vol. 1 No. 30 (2026) / 2026-03-30 / Number of views: 137

AUTOMATED ANALYSIS OF TECHNOGENIC WASTE GRANULOMETRY: ТЕХНОГЕНДІК ҚАЛДЫҚТАРДЫҢ ГРАНУЛОМЕТРИЯСЫН АВТОМАТТАНДЫРЫЛҒАН ТАЛДА

Authors

K.Kulazhanov Kazakh University of Technology and Business
https://orcid.org/0000-0001-5202-3501
Nazarbayev Intellectual School
https://orcid.org/0009-0008-1214-0867
K.Kulazhanov Kazakh University of Technology and Business
https://orcid.org/0000-0002-0744-0389

Keywords

machine learning, models, granulometric composition, man-made waste, machine vision, algorithm

Link to DOI:

https://doi.org/10.58805/kazutb.v.1.30-1396

How to quote

Akishev K.М., et al. “AUTOMATED ANALYSIS OF TECHNOGENIC WASTE GRANULOMETRY: ТЕХНОГЕНДІК ҚАЛДЫҚТАРДЫҢ ГРАНУЛОМЕТРИЯСЫН АВТОМАТТАНДЫРЫЛҒАН ТАЛДА”. Vestnik KazUTB, vol. 1, no. 30, Mar. 2026, doi:10.58805/kazutb.v.1.30-1396.

Abstract

The relevance of the study is due to the need for prompt and reliable control of the granulometric composition of man-made waste from metallurgical and energy production when used in the construction industry. Traditional methods, in particular sieve analysis, are characterized by high labor intensity and limited applicability in conditions of mass and automated control, which makes it advisable to introduce digital analysis methods.

The paper proposes a comprehensive method for determining the granulometric composition of slag and ash granules based on the analysis of SEM images using computer vision and machine learning technologies. The method includes the stages of 1240 image preprocessing, particle segmentation using the U-Net model, YOLOv5-based object detection, geometric feature extraction, and fraction classification using an ensemble of Random Forest, XGBoost, and SVM algorithms. Verification of the results was performed by comparing the obtained distributions with the reference sieve analysis according to GOST 5459.2-20 and the ASTM C136 standard.

The scientific novelty of the work lies in the application of the hybrid architecture "computer vision — machine learning" for the analysis of man-made waste with pronounced morphological heterogeneity. The quantitative results showed the achievement of classification accuracy of up to 95% with an average absolute error of 6-7%, and the qualitative analysis confirmed the stability of the method in complex shapes and dense packing of particles. The practical significance of the research lies in the possibility of automating the quality control of secondary raw materials and reducing the analysis time. The prospects for further research are related to the transition to online photography, integration with production lines, and expansion of the dataset of man-made materials.