On the Performance Assessment of ANN and Spotted Hyena Optimized ANN to Predict the Spontaneous Combustion Liability of Coal

Abstract

ABSTRACT Spontaneous combustion is the greatest hurdle faced in coal mining, requiring a more precise solution that is effective to predict the likelihood of its occurrence. The phenomenon has long been a concern affecting flora and fauna, mining machinery, coal reserves, and consuming essential and nonrenewable resources. To predict this occurrence, the application and performance evaluation of artificial neural network (ANN) and spotted hyena optimized ANN (SHO-ANN) were used to develop reliable models in predicting coal spontaneous combustion liability. The coefficient of determination (R2), mean error (ME), and root-mean-squared error (RMSE) were used to assess the performances of the models based on sixty-eight (68) datasets. The datasets were used for training, testing, and validation. The models provide excellent predicting ability with an R2 value close to 1, errors close to zero, while most of the data points fall within ± 3% error bands. Sensitivity analysis was conducted using the cosine amplitude method and the result shows volatile matter (VM) and Oxygen (O) as having the highest influence on the Wits-Ehac and FCC (Feng, Chakravorty, Cochrane) liability indices, while O and nitrogen (N) have the highest influence on crossing point temperature (XPT). In addition, this study provides a graphic user interface for the practical implementation of the proposed models in the coal mines.