An artificial neural network and principle component analysis based model for methane level prediction in underground coal mines

High methane levels in underground coal mines interfere with mining activities and increase the risk of fires and explosions. Therefore, early warning and predicting systems are imperative in ongoing underground coal mining exploitation areas. In this paper, a hierarchical approach made of the principal component analysis (PCA) and the artificial neural network (ANN) model is proposed to improve the prediction accuracy of methane levels. The PCA was used to evaluate those factors most influencing methane levels. The variables extracted by the PCA were used as inputs parameters to the artificial neural network ANN model. An ideal number of neurons was developed for both conventional inputs and PCA-extracted variables. To train the model four algorithms were employed. The algorithm which proved to have the highest accuracy was Levenberg-Marquardt, with a supervised method of learning adopted. The study demonstrates that the hierarchical model achieved better performance and slightly improved prediction accuracy than the ANN model with original input parameters. It is also proven that a higher prediction is dependent on the variables derived from the PCA and the training algorithm adopted.