Intelligent based decision-making strategy to predict fire intensity in subsurface engineering environments

Abstract

The prospect of subsurface structures taking uncontrollable fire is a significant cause of stress held by people from all over the world. Mine fires and explosions have caused countless casualties and material losses for decades. A fire in a mine has the ability to rapidly contaminate the air of the entire mine, which might result in the loss of life and, in certain cases, the suspension of mining industry processes. Mine fires deplete coal supplies, release greenhouse gases and toxic chemicals, and cause ground subsidence from coal volume loss. This study aims to explore some interesting aspects of mine's fire data using Catboost and light gradient boosting machine (LightGBM) methods in order to minimize the human fatalities and material losses during the construction of deep underground engineering projects. Firstly, 120 samples demonstrating several distinct characteristics were gathered from an underground coal mine in Turkey. The prediction framework was developed employing the training set and the appropriate configuration of hyperparameters. The findings indicate that LightGBM algorithms achieved a more comprehensive performance in comparison with Catboost, with an accuracy of 92 % and 89 %, respectively. Hence, the proposed intelligent decision-making model can be employed in preventing and warning system of fire risks in subsurface engineering projects. The suggested strategy would serve as a reliable guide for predicting fire intensity in deep underground projects to ensure safe subsurface environments. This research promotes safe, hygienic, environmentally friendly, climate-smart and sustainable development by reducing greenhouse gas emissions.