Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Graph-Driven Models for Gas Mixture Identification and Concentration Estimation on Heterogeneous Sensor Array Signals

Ding Wang, Lei Wang, Huilin Yin, Guoqing Gu, Zhiping Lin, Wenwen Zhang

TL;DR

This work tackles the generalization gap in gas mixture identification and concentration estimation across heterogeneous sensor datasets. It introduces two graph-enhanced deep models, GraphCapsNet for classification and GraphANet for regression, that operate on temporal graph representations of sensor data and employ dynamic routing and self-attention, respectively. Across UCI and a self-developed dataset, the methods achieve exceptional performance with accuracy above 0.98 and R^2 above 0.96, demonstrating strong cross-dataset robustness and minimal need for per-dataset tuning. The results suggest these graph-driven approaches offer scalable, real-time gas analysis suitable for industrial deployment, with future work extending to additional sensor types and more varied environments.

Abstract

Accurately identifying gas mixtures and estimating their concentrations are crucial across various industrial applications using gas sensor arrays. However, existing models face challenges in generalizing across heterogeneous datasets, which limits their scalability and practical applicability. To address this problem, this study develops two novel deep-learning models that integrate temporal graph structures for enhanced performance: a Graph-Enhanced Capsule Network (GraphCapsNet) employing dynamic routing for gas mixture classification and a Graph-Enhanced Attention Network (GraphANet) leveraging self-attention for concentration estimation. Both models were validated on datasets from the University of California, Irvine (UCI) Machine Learning Repository and a custom dataset, demonstrating superior performance in gas mixture identification and concentration estimation compared to recent models. In classification tasks, GraphCapsNet achieved over 98.00% accuracy across multiple datasets, while in concentration estimation, GraphANet attained an R2 score exceeding 0.96 across various gas components. Both GraphCapsNet and GraphANet exhibited significantly higher accuracy and stability, positioning them as promising solutions for scalable gas analysis in industrial settings.

Graph-Driven Models for Gas Mixture Identification and Concentration Estimation on Heterogeneous Sensor Array Signals

TL;DR

This work tackles the generalization gap in gas mixture identification and concentration estimation across heterogeneous sensor datasets. It introduces two graph-enhanced deep models, GraphCapsNet for classification and GraphANet for regression, that operate on temporal graph representations of sensor data and employ dynamic routing and self-attention, respectively. Across UCI and a self-developed dataset, the methods achieve exceptional performance with accuracy above 0.98 and R^2 above 0.96, demonstrating strong cross-dataset robustness and minimal need for per-dataset tuning. The results suggest these graph-driven approaches offer scalable, real-time gas analysis suitable for industrial deployment, with future work extending to additional sensor types and more varied environments.

Abstract

Accurately identifying gas mixtures and estimating their concentrations are crucial across various industrial applications using gas sensor arrays. However, existing models face challenges in generalizing across heterogeneous datasets, which limits their scalability and practical applicability. To address this problem, this study develops two novel deep-learning models that integrate temporal graph structures for enhanced performance: a Graph-Enhanced Capsule Network (GraphCapsNet) employing dynamic routing for gas mixture classification and a Graph-Enhanced Attention Network (GraphANet) leveraging self-attention for concentration estimation. Both models were validated on datasets from the University of California, Irvine (UCI) Machine Learning Repository and a custom dataset, demonstrating superior performance in gas mixture identification and concentration estimation compared to recent models. In classification tasks, GraphCapsNet achieved over 98.00% accuracy across multiple datasets, while in concentration estimation, GraphANet attained an R2 score exceeding 0.96 across various gas components. Both GraphCapsNet and GraphANet exhibited significantly higher accuracy and stability, positioning them as promising solutions for scalable gas analysis in industrial settings.

Paper Structure

This paper contains 26 sections, 7 equations, 10 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Gas Sensor Array Signal Acquisition System and Data Collection
  3. Gas sensor array and signal acquisition
  4. The overview of sensor array from UCI dataset and our lab
  5. Data acquisition protocol
  6. UCI and self-developed sensor array datasets
  7. UCI sensor array dataset
  8. self-developed sensor array dataset
  9. Methodology
  10. The architecture of GraphCapsNet
  11. Graph encoding module
  12. Dynamic routing module
  13. The loss
  14. The achitecture of GraphANet
  15. Experiment
  16. ...and 11 more sections

Figures (10)

  • Figure 1: (a) Standard gas configuration system. (b) Custom-developed gas sensor array.
  • Figure 2: (a) Data segment of UCI dataset. (b) Custom data sample.
  • Figure 3: The heat map of data distribution. (a) CO-C2H4 from UCI dataset (b) CH4-C2H4 from UCI dataset (c) H2-C2H4 from custom dataset.
  • Figure 4: The data processing of UCI data and custom data.
  • Figure 5: Architecture of the GraphCapsNet model for gas recognition.
  • ...and 5 more figures