# Research on odor prediction methods for coal spontaneous combustion based on E-nose technology

## TL;DR

This study uses an electronic nose to detect and predict coal spontaneous combustion by analyzing odors, with promising accuracy for early warning in high-risk coal areas.

## Contribution

A new programmable temperature E-nose device and PCA-SVM model were developed to accurately predict stages of coal spontaneous combustion using odor data.

## Key findings

- Acetaldehyde was the dominant volatile compound detected during early spontaneous combustion stages with high feature importance.
- PCA effectively differentiated CSC stages, explaining 92.45% of the variance in odor characteristics.
- The PCA-SVM model achieved over 95% accuracy in classifying CSC stages based on odor data.

## Abstract

This study uses an electronic nose to predict Coal Spontaneous Combustion (CSC) based on odor information. In this study, we built a Programmable Temperature Coal Electronic Nose Testing Device (PTC E-nose). Its purpose was to detect odors from gases emitted by lignite during spontaneous combustion (30–200 °C). Odor characteristics at different stages of spontaneous combustion were analyzed, and principal component analysis (PCA) was used to illustrate their distribution across the various stages. Multiple machine learning models were employed to predict CSC stage. Results showed that acetaldehyde was the dominant volatile compound detected by the E-nose during the early stage of CSC, with a feature importance score of 0.38 in the random forest analysis. The response intensity of the acetaldehyde sensor showed a significant correlation with coal temperature (R²=0.97), suggesting its utility as an effective indicator for predicting spontaneous combustion, with benzene serving as an auxiliary predictor. The odor characteristics during different stages of CSC were distinctly different. PCA effectively differentiated among CSC stages (explaining 92.45% of the variance). The Principal Component Analysis-Support Vector Machine (PCA-SVM) model demonstrated the ability to identify characteristics of CSC odor and accurately classify stages, achieving recognition accuracy exceeding 95%. This study concludes that using E-nose technology for odor detection can be effectively used for monitoring and early warning of CSC. It is especially suitable for high-risk areas such as goaf areas, fractured coal seams, fault zones, and abandoned roadways.