Channel wave signal classification based on Riemann MDS algorithm
Hongyu Sun, Xue Liu, Lina Sun, Haiyan Yu, Yan Dong, Fuyu Chu · 2023
China's coal fields are tectonically complex and difficult to mine. Traditional means of physical exploration such as three-dimensional (3D) seismic exploration, radio pit detection and geological radar are greatly affected by geology, and the detect results are not precise enough to meet the requirements of precise exploration. In recent years, channel wave exploration technology has gradually become a new means of geological exploration in coal mines by virtue of its high detection accuracy, large detection distance, strong resistance to electrical interference and easy identification of waveform characteristics, and it can be used to explore the hidden disaster-causing factors, and changes of coal seam thickness in coal mines. The conventional Euclidean space cannot effectively describe the nonlinearity of the channel wave. The Riemann MDS algorithm combined with fisher feature selection method is proposed for classification of small geological structure in coal mines. A three-dimensional (3D) medium geometry model is developed for a complicated coal seam with goaf, collapse column, wash zone, and small fault based on COMSOL Multiphysics. Then the MDS -based Riemann algorithm is employed for feature extraction. Finally, the PSO-SVM classifier is used for classification of small geological structure. In experiment, the extracted Riemannian channel wave dataset is divided into training and testing sets in a 7:3 ratio, and sent to the PSO-SVM optimization model for training. The experimental results show that the classification accuracy based on Riemann MDS and PSOSVM is as high as 98.67%, which demonstrates that the proposed algorithm improves the accuracy of classification.