Identification of stochastic disturbance sources of air doors in mine ventilation systems

Abstract

Identifying stochastic disturbance sources at air doors in mine ventilation systems is essential for ensuring both safety and operational efficiency. To address this issue, a time-dependent disturbance model, based on the Navier-Stokes equations, was developed to facilitate the rapid and accurate localization of disturbances. Particle Image Velocimetry (PIV) experiments were performed to investigate the characteristics of air door disturbances and their dynamic effects on airflow. A novel approach for identifying disturbance sources, based on time-series monitoring signals, was proposed, wherein air volume and air pressure serve as key features for classifying disturbance locations. The PIV experiments revealed both logarithmic and linear relationships between tunnel wind speed and contraction wind speed, as well as between the air door opening ratio and the contraction coefficient. By leveraging unsteady disturbance theory, training samples were generated for the stochastic disturbance source identification task. Experimental results demonstrated that Long Short-Term Memory (LSTM) networks achieved a high accuracy of 90.28% in identifying disturbances based solely on air volume. In both T-type and complex ventilation systems, the inclusion of air pressure features significantly improved identification accuracy by 9.66% and 9.16%, respectively. However, the combination of air volume and air pressure features resulted in minimal additional improvements (0.03% and 0.12%). The evaluation of model performance confirmed that the LSTM-based framework, supported by unsteady disturbance theory, is highly effective in identifying stochastic disturbance sources within mine ventilation systems.