We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
基于 Piper-PCA-Fisher 的矿井突出水源 判别模型构建及应用.
- Authors
孙延辉; 胡文博; 马红林; 王历民; 杜 华; 李世涛
- Abstract
The efficient and accurate identification of water inrush sources is a critical step to prevent mines from water inrush accidents. In order to improve the reliability of identification of mine water inrush source, a mine water inrush source identification model based on Piper trilinear diagram, Principal Component Analysis (PCA), and Fisher identification method was presented with the data of water sample taken from Linhuan mining area as a study case. Using the concentration tested for six categories of chemical ions as the identification index for water inrush sources, sample classification and identification were carried out by determining the distance between the evaluation set and the center of the four categories of water sources. The research results indicate that the Piper trilinear diagram can be utilized to exclude abnormal samples, and ultimately 37 sets of typical water samples were chosen to be training samples. Principal component analysis was adopted to reduce the dimensionality of identification index. It can be found that the former three principal components presented 91.636% coverage of the original index information, which made the hydro-chemical property of the sample be characterized efficiently. The Piper-PCA-Fisher model was applied to back identify 37 sets of training samples and predict 15 sets of samples to be tested, and the accuracy of both back identification and prediction was 100%. The Piper-PCA-Fisher identification model helps to weaken the correlation between the indexes. The sound identification effects in case of limited availability of water sample data can be acquired.
- Subjects
PRINCIPAL components analysis; WATER supply; WATER sampling; CLASSIFICATION; MINE accidents
- Publication
Industrial Minerals & Processing / Huagong Kuangwu yu Jiagong, 2023, Vol 52, Issue 9, p52
- ISSN
1008-7524
- Publication type
Article
- DOI
10.16283/j.cnki.hgkwyjg.2023.09.009