Deep Learning-based Gangue Sorting for Coal Plants

In a paper published in the journal Scientific Reports, researchers introduced a deep learning (DL) --based, non-contact gangue recognition and pneumatic sorting system for coal-washing plants. This system, featuring a dynamic database and precise impact energy matching, significantly enhances gangue identification accuracy and sorting rate. Demonstration experiments show a separation time of less than 3 seconds, considerably improving raw coal purity compared to traditional methods.

Study: Deep Learning-based Gangue Sorting for Coal Plants. Image Credit: Petr Szymonik/Shutterstock.com
Study: Deep Learning-based Gangue Sorting for Coal Plants. Image Credit: Petr Szymonik/Shutterstock.com

Related Work

Past work on coal gangue sorting has transitioned from labor-intensive methods to advanced technologies, including mechanized and intelligent systems. Traditional methods face issues with efficiency and accuracy, prompting the adoption of DL and image-based sorting, which have shown promising results. Challenges such as noise, motion blur, and precise control in pneumatic systems still need progress. Recent innovations include multi-axis robotic arms and image processing, enhancing sorting speed and accuracy.

System Design Overview

The coal gangue sorting system is designed with a belt conveyor that is 4.4 meters long, 1 meter wide, and 2.3 meters high. It includes a 0.7-meter feeding port, a 1.3-meter paving queuing module, a 1.2-meter coal gangue identification module, and a 1.3-meter sorting space. The workflow involves capturing and identifying coal and gangue, with the system using a cubic darkroom fitted with an industrial-grade digital camera array and uniform strip lighting for image capture. The DL-based recognition module utilizes the residual network with 50 50-layer (ResNet-50) network, optimized through transfer learning, to classify coal and gangue based on image analysis.

The system framework integrates an application for coal gangue identification using machine vision easy robotics laboratory interface configuration (MERLIC) software. It features speed input, image adjustment, object search, and result processing modules. Identification data, including coordinates and processing time, is transmitted to the sorting system controller over Ethernet. The MERLIC Designer component creates an intuitive human-computer interface, enabling easy interaction and adjustment of system controls.

Control System Overview

The coal gangue sorting control system features a comprehensive design framework. Data packets are transmitted via Ethernet to a communication module. This module converts the Ethernet packets into serial port data and relays them to the sorting module controller. The controller processes the data to regulate cylinder actions and manage the sorting workflow. It also modifies the stepper motor's rotation based on the conveyor belt speed to ensure precise impact by the cylinder.

The hardware design includes key components such as guide rails, solenoid valves, cylinders, magnetic switches, and stepper motors, with the chip selected for the core control circuit. The communication module uses the winchiphead technology (CH9121) network serial communication chip, while the cylinder control circuit employs a dual electronic control solenoid valve. The sorting mechanism is adjusted using a stepper motor controlled by the SD-20504 digital driver, which ensures accurate positioning.

The software design incorporates a main program to process identification data and control sorting actions. Cylinder and stepper motor programs handle timing and movement based on belt speed and image recognition data. The system aims to improve raw coal selection efficiency, reduce labor intensity, and minimize energy consumption, leading to lower pollutant emissions and a better ecological impact.

Experimental System Evaluation

This study experimentally assesses the prototype of the coal gangue sorting system by feeding coal and gangue at various belt speeds. The prototype evaluates recognition and sorting performance for coal with a 30% gangue content, using gangue particles ranging from 50 to 120 mm in size. Preliminary results indicate that smaller particles might lead to the cylinder breaking the gangue, but larger particles may not be separated effectively due to insufficient force.

Experimental results show that the prototype achieves over 97% recognition accuracy, more than 92% recall, and over 91% sorting rate within a belt speed range of 0.5–1 m/s. The system performs optimally at 0.5 m/s, with accuracy, recall, and sorting rates reaching 99.2%, 98.3%, and 98.3%, respectively. In comparison, previous methods using image processing and multi-layer perceptrons achieved average recognition accuracy of 96.45% and sorting rates of 90.76%.

The study also compares energy efficiency between pneumatic cylinder sorting and other methods like spray guns and robotic arm sorting. The cylinder-type method is more energy-efficient, simpler, and more effective for sorting coal gangue. This method's advantages include lower energy consumption and improved application benefits for handling coal gangue of similar sizes.

Conclusion

To sum up, this study introduced a pneumatic intelligent coal gangue sorting system based on DL, addressing issues with traditional coal selection technologies. The system accurately identified and sorted coal gangue particles ranging from 50 to 120 mm, utilizing the high-performing ResNet-50 model to achieve over 97% accuracy. Compared to existing methods, it offered greater efficiency and lower energy consumption, with a sorting rate exceeding 91% and a separation time under 3 seconds. Experimental results confirmed that this system simplified operations, enhanced sorting performance, and reduced production costs.

Journal reference:
Silpaja Chandrasekar

Written by

Silpaja Chandrasekar

Dr. Silpaja Chandrasekar has a Ph.D. in Computer Science from Anna University, Chennai. Her research expertise lies in analyzing traffic parameters under challenging environmental conditions. Additionally, she has gained valuable exposure to diverse research areas, such as detection, tracking, classification, medical image analysis, cancer cell detection, chemistry, and Hamiltonian walks.

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Chandrasekar, Silpaja. (2024, July 25). Deep Learning-based Gangue Sorting for Coal Plants. AZoAi. Retrieved on October 08, 2024 from https://www.azoai.com/news/20240725/Deep-Learning-based-Gangue-Sorting-for-Coal-Plants.aspx.

  • MLA

    Chandrasekar, Silpaja. "Deep Learning-based Gangue Sorting for Coal Plants". AZoAi. 08 October 2024. <https://www.azoai.com/news/20240725/Deep-Learning-based-Gangue-Sorting-for-Coal-Plants.aspx>.

  • Chicago

    Chandrasekar, Silpaja. "Deep Learning-based Gangue Sorting for Coal Plants". AZoAi. https://www.azoai.com/news/20240725/Deep-Learning-based-Gangue-Sorting-for-Coal-Plants.aspx. (accessed October 08, 2024).

  • Harvard

    Chandrasekar, Silpaja. 2024. Deep Learning-based Gangue Sorting for Coal Plants. AZoAi, viewed 08 October 2024, https://www.azoai.com/news/20240725/Deep-Learning-based-Gangue-Sorting-for-Coal-Plants.aspx.

Comments

The opinions expressed here are the views of the writer and do not necessarily reflect the views and opinions of AZoAi.
Post a new comment
Post

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.

You might also like...
Revolutionizing Gemstone Analysis with Deep Learning