A Comprehensive Review of Modern Technologies in Industrial Vibrating Screens: From Discrete Element Modeling (DEM) and Machine Learning to Mechanical Design, Crack Dynamics, and Intelligent Fault Detection
A Comprehensive Review of Modern Technologies in Industrial Vibrating Screens: From Discrete Element Modeling (DEM) and Machine Learning to Mechanical Design, Crack Dynamics, and Intelligent Fault Detection
sina soleimani1 mahmod eskandari nasab2
1) M.Sc. Student in Mineral Processing Engineering, Shahid Bahonar University of Kerman
2) Ph.D. in Mineral Processing Engineering, Shahid Bahonar University of Kerman
IOI Code
:
ACCD-ACZDCD
Publication :
10th International Conference on Applied Researches in Science & Engineering (carseconf.com)
Abstract :
Vibrating screens are among the key equipment in particle separation and classification processes in the mining, recycling, mineral processing, and renewable energy industries (such as solar panel recycling). In recent years, numerous studies have focused on four main axes: (1) simulating the behavior of irregular particles using the Superquadric Discrete Element Method (Superquadric DEM) and developing hybrid deep learning models (CNN-LSTM) for rapid prediction of separation behavior; (2) designing reconfigurable mechanical structures to adapt to fluctuations in production demand; (3) dynamic analysis of cracked cross-beams as one of the most common causes of failure in linear mineral vibrating screens; and (4) developing intelligent fault detection systems for screen mesh using deep learning-based object detection algorithms. In this paper, these four axes are reviewed and compared in an integrated manner, and the key quantitative findings of each study, including separation efficiency, prediction model accuracy, structural stresses, and fault detection accuracy, are presented in the form of tables and figures. The results show that circular displacement and vibration angle are the most effective operational parameters on separation efficiency, CNN-LSTM models with R² accuracy higher than 0.94 and speeds up to 72 times faster than DEM simulation are suitable alternatives for rapid process prediction, and improved YOLOv5 algorithms incorporating ConvNeXt V2, EfficientRep, and ODConv modules can achieve a mean average precision (mAP) for fault detection of more than 97 percent.
Keywords : vibrating screen discrete element method (dem) non spherical particles cnn lstm deep learning reconfigurable vibrating screen cracked beam dynamics machine vision based fault detection yolov5