Related Experiment Video
Updated: Jan 10, 2026
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
5.2K
ACI-GNN: Lightweight All-Channel Interaction Graph Neural Network for Multi-Sensor Coal-Rock Cutting Recognition
Zhixin Jin1,2,3, Jie Cheng1,2, Wenyan Cao1,2,3,4
1College of Safety and Emergency Management Engineering, Taiyuan University of Technology, Taiyuan 030024, China.
Sensors (Basel, Switzerland)
|November 27, 2025
Summary
This study introduces a novel graph neural network (GNN) model for multi-sensor coal and rock cutting state recognition, significantly improving accuracy and robustness. The GNN model enhances feature capture and reduces redundancy, outperforming traditional neural networks in noisy conditions.
Area of Science:
- Mining Engineering
- Artificial Intelligence
- Signal Processing
Background:
- Traditional neural networks struggle with low single-sensor accuracy and channel redundancy in coal/rock cutting state recognition.
- Existing models exhibit poor performance and limited robustness in complex industrial environments.
Purpose of the Study:
- To develop a robust multi-sensor coal and rock cutting state recognition model using graph neural networks (GNNs).
- To enhance feature capture, reduce channel redundancy, and improve recognition accuracy and noise resistance.
Main Methods:
- A novel GNN-based model comprising a feature encoder, information exchange module, and feature decoder was proposed.
- The model enhances inter-filter communication within layers to improve feature extraction.
- Comparative, ablation, and noise-resistance experiments were conducted on multi-sensor datasets.
Main Results:
- The proposed GNN model achieved significant accuracy improvements over baseline models (CNN3, ResNet, DenseNet).
- The ResNet model with the ACI block demonstrated superior noise resistance, maintaining 93.27% accuracy at 6 dB noise.
- Embedded deployment confirmed real-time performance (<216.1 ms/window on NVIDIA Jetson Nano).
Conclusions:
- The GNN-based model effectively addresses challenges in coal and rock cutting state recognition.
- The model exhibits excellent robustness, accuracy, and real-time processing capabilities.
- It shows promising applications in resource-constrained underground mining environments.
