Source Discrimination of Mine Water Inrush Based on Spectral Data and EGA–PNN Model: A Case Study of Huangyuchuan mine; [基于频谱数据和EGA-PNN模型的矿井突水水源识别-以黄玉川矿为例]; [Fallstudie zur Eingrenzung von Wassereinbrüchen in Bergwerken mittels Spektraldaten und EGA–PNN Modell]; [Discriminación de la fuente de irrupción de agua en la mina basada en datos espectrales y en un modelo EGA-PNN: un estudio de caso]

Methods currently used to identify water sources and predict mine water inrush points are mostly based on water temperatures, water levels, or hydrochemical data. However, these methods are relatively complex. Therefore, to rapidly and effectively identify the source of mine water inrush, an elite strategy was used to optimize a genetic algorithm. The optimized elitist genetic algorithm (EGA) was applied to parameter selection of a probabilistic neural network (PNN) using spectral data to construct an EGA–PNN discriminant model. Water samples were collected from four different sources in the Huangyuchuan Mine in north China, including Permian sandstone fissure water, Carboniferous coal seam roof water, Ordovician limestone water, and Permian mined-out water. Full spectral data of the water samples were obtained, and 80% of the sample data were randomly selected as the training set for the EGA–PNN model. Theoretical analysis and experimental results showed that after 100 iterations, the average recognition rate was ≈ 95.69% with a mean square error average of only ≈ 8.21 × 10−3. This study provides a novel methodology that can be used to rapidly identify mine water inrush sources in similar geological coalfields elsewhere in northern China. © 2022, The Author(s) under exclusive licence to International Mine Water Association.