Wireless Sensor Network Based Underground Coal Mine Environmental Monitoring Using Machine Learning Approach

Abstract

Underground coal mines are associated with environmental problems and serious hazards resulting in the loss of coal resources and valuable lives. Wireless sensor network (WSN), nowadays, is widely adopted for continuous monitoring of environmental parameters and other aspects of underground coal mines. Sensor nodes deployed in underground coal mines are meant for sensing the environmental parameters and transferring the data to ground monitoring station for processing. Since the monitoring data are imprecise and vague in nature, it is imperative to analyze the data for taking precautionary measures. In this paper, we applied data mining or Machine Learning (ML) technique to predict the occurrence of mine fire hazard in underground coal mines. Machine learning concept is inherited from Artificial Intelligence (AI) which has an ability to learn from the past experiences. We applied supervised learning method on the collected data set which partitioned into training and testing data sets. The purpose of training data set is to train the model, whereas the testing data set is meant for evaluation of the trained model. Learning method is implemented at the base stations or sinks instead of monitoring station for taking immediate real-time decision on sensed data in case of any hazard. Learning method for monitoring of different hazard conditions are simulated using WEKA (data mining and machine learning) Tool. This system is more reliable and responsive to any kind of hazards as compared to offline monitoring system used in underground coal mines.