New study shows AI can predict mining disasters

Artificial intelligence can predict gas incidents in coal mines within half an hour, according to a new study exploring how the technology can reduce the risk of disasters.

The study of coal mines in China compared 10 machine learning algorithms to see which AI method could predict changes in methane gas levels 30 minutes in advance and alert users of anomalies. “A Comparative Study of Ten Machine Learning Algorithms for Short-Term Forecasting in Gas Warning Systems” was published in the journal Scientific reports.

Gas explosions or fires in underground mines pose significant risks: nearly 60% of coal mine accidents in China are caused by methane gas.

In 2020, China accounted for 46% of global coal production, and the country had more than 3,200 high-gas coal mines with explosion hazard levels.

Author and adjunct associate professor at Charles Darwin University’s (CDU) School of Science and Technology, Niusha Shafiabady, said the results showed that out of the 10, four machine learning algorithms produced the best results.

“Linear regression is one of the most efficient algorithms with better performance for short-term forecasting than others,” said Associate Professor Shafiabady.

“Random Forest often shows statistically lower error performance and achieves the highest prediction accuracy. Support Vector Machine performs well and has shorter computation time on small datasets, but will require too much training time as the dataset size increases.

“The results of this study will help the coal mining industry reduce the risk of accidents such as gas explosions, protect workers and improve the ability to prevent and mitigate disasters that will cause financial losses in addition to potential loss of life.”

The study was conducted with Charles Darwin University, University of Technology Sydney, Australian Catholic University, Shanxi Normal University and Central Queensland University.

Associate Professor Niusha Shafiabady, who is also a research fellow at the Peter Faber Business School at Australian Catholic University, said the findings had multiple applications.

“This method works for all coal mines, and the same principles can be applied to other industries such as aerospace, oil and gas, agriculture and many others,” she said.

“This is an example of an application where AI can be used to save lives and mitigate health and safety risks.”

A

A previous study led by Associate Professor Shafiabady found that increased monitoring of wind, gas density and temperatures in coal mines can also help reduce the risk of disasters.