Fault-tolerant multi-sensor data fusion system for underground mine gas hazard prediction using Dempster Shafer Evidence Theory

Gas hazards in the coal mine are essential concerns related to the safety of mine workers and mining operations. Hence, a reliable and accurate gas hazard prediction and monitoring system is required. Most available schemes are independent node-based systems that are vulnerable when a sensor gets at fault, resulting in wrong decisions about any state, ambiguous alarm signals, and many associated issues. Therefore, a double-layer, fault-tolerant data fusion (FTDF) model is proposed in this study. In the lower level of FTDF, the gas hazard classification is done individually for each node using an efficient classification algorithm. After that, a refined decision is generated at the higher level of FTDF using Dempster Shafer evidence theory (DSET). The gas hazard prediction accuracy at the lower level is 97 %, and the hazard prediction accuracy at the higher level after implementing the FTDF is 99.99%. The proposed model has been investigated for healthy and faulty node cases, and 99.99% accuracy is observed in both scenarios. The performance of the FTDF model is compared with SVM and ANN models, where SVM and ANN suffer in faulty node cases. On the other hand, FTDF is immune to faulty node scenarios.