A Machine Learning Approach for Fire-Fighting Detection in the Power Industry

Firas Basim Ismail*, Ammar Al-Bazi, Rami Hikmat Al-Hadeethi, Mathew Victor

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


Coal kept in the coal storage yard spontaneously catches on fire, which results in wastage and can even cause a massive fire to break out. This phenomenon is known as the spontaneous combustion of coal. It is a complex process that has non-linear relationships between its causing variables. Preventive measures to prevent the fire from spreading to other coal piles in the vicinity have already been implemented. However, the predictive aspect before the fire occurs is of great necessity for the power generation sector. This research investigates various prediction models for spontaneous coal combustion, explicitly selecting input and output parameters to identify a proper clinker formation prediction model. Feed-Forward Neural Network (FFNN) is proposed as a proper prediction model. Two Hidden Layers (2HL) network is found to be the best with 5 minutes prediction capability. A sensitivity analysis study is also conducted to determine the influence of random input variables on their respective response variables.

Original languageEnglish
Pages (from-to)475-482
Number of pages8
JournalJordan Journal of Mechanical and Industrial Engineering
Issue number5
Publication statusPublished - Dec 2021

Bibliographical note

Funding Information:
This research was financially supported by Universiti Tenaga Nasional, Malaysia through BOLD refresh publication fund 2021(J510050002-BOLD refresh2025-Centre of Excellence).

Publisher Copyright:
© 2021 Jordan Journal of Mechanical and Industrial Engineering. All rights reserved


  • Artificial neural network
  • Clinker formation prediction models
  • Coal-fired power plant
  • Spontaneous combustion of coal


Dive into the research topics of 'A Machine Learning Approach for Fire-Fighting Detection in the Power Industry'. Together they form a unique fingerprint.

Cite this