Machine learning based prediction model and interpretable analysis of airflow temperature in water dropping shaft

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Accurate prediction of the air-flow temperature in the mine water dropping shaft is crucial for scientifically guiding the prevention and control of thermal hazards in mines. To improve the precision, stability, and interpretability of the prediction model for the air-flow temperature in mine drainage shafts, combined with relevant literature research and the actual characteristics of mine air-flow temperature, Pearson correlation analysis was introduced to screen the characteristic variables. The parameters of the LSBoost model are optimized using the WOS algorithm, and a mine drainage shaft air-flow temperature prediction model based on WOS-LSBoost is established. Under the same sample set conditions, compare the prediction precision of the four models with the established DT model, RF model, LSBoost model, and SVM model, and conduct interpretability analysis on the prediction models based on the SHAP analysis method. Research has demonstrated that the WOS-LSBoost model has the best predictive performance, stronger generalization ability, and can effectively improve the precision of air-flow temperature prediction; the temperature of the wellhead air-flow has the largest effect on the prediction model, while the surface pressure has the smallest effect on the prediction model. This study provides scientific reference for the prevention and control of thermal hazards in underground mines.

This publication has 10 references indexed in Scilit: