CLASSIFICATION OF SEISMOACOUSTIC EMISSION SOURCES IN FIBER-OPTIC SYSTEMS FOR MONITORING EXTENDED OBJECTS
A. V. Timofeev, D. I. Groznov
LLC В«Flagman-Geo», Saint-Petersburg, Russia
Keywords: классификация, XGBoost, SVM, DAS, П†-OTDR, classification
Abstract
A new method is proposed for automatic classification of seismoacoustic emission sources in fiber-optic monitoring systems based on the principles of optical reflectometry in a time domain using the time reconstruction of the interference signal phase. The novelty of this approach lies in the original principles of forming a space of classification features and the use of ensemble classifiers. This method works provided that a relatively small training database is used. High practical effectiveness of the proposed approach is shown when working on real data. A classification reliability value previously inaccessible to alternative classification methods implemented in serial fiber-optic monitoring systems is evenly achieved for different classes of goals
|