Intelligent identification of railway roadbed diseases based on unsupervised learning

Konferenz: EEI 2022 - 4th International Conference on Electronic Engineering and Informatics
24.06.2022 - 26.06.2022 in Guiyang, China

Tagungsband: EEI 2022

Seiten: 4Sprache: EnglischTyp: PDF

Autoren:
Zhang, Kun (China Academy of Railway Sciences Corporation Limited Railway Engineering Research Institute, Beijing, China & Tsinghua University, School of Environment, Beijing, China)
Du, Cui (China Academy of Railway Sciences Corporation Limited Railway Engineering Research Institute, Beijing, China)

Inhalt:
Ground penetrating radar is the main method for detecting the disease of railway subgrade. It is difficult to establish a GPR image sample set, and it is difficult to carry out image recognition based on supervised learning. This paper proposes an intelligent identification method of railway subgrade diseases based on unsupervised learning, and uses k-means clustering algorithm and LandTrendr algorithm to realize automatic classification of railway subgrade GPR signals. The test results show that this method can distinguish three types of signals: normal, muddy and water-containing, and the classification accuracy rate reaches more than 95%.