Open AccessTHE PROCESS SIGNAL ANOMALY DETECTION USING CLASSIFIER ENSEMBLE AND WAVELET TRANSFORMS
Author(s) -
Damir A. Murzagulov,
А В Замятин
Publication year - 2021
Publication title -
avtomatizaciâ processov upravleniâ
Language(s) - English
Resource type - Journals
ISSN - 1991-2927
DOI - 10.35752/1991-2927-2021-1-63-20-26
Subject(s) - anomaly detection , wavelet , preprocessor , computer science , artificial intelligence , pattern recognition (psychology) , process (computing) , data mining , classifier (uml) , wavelet transform , signal (programming language) , ensemble learning , data pre processing , machine learning , programming language , operating system
The IT-infrastructure development level of manufacturing plants allows to collect and storage technological information, thereby creating the possibilities to adapt data-mining systems. The article deals with the problem of anomaly detection in process signals with a view to improving the quality of control object monitoring. The ensemble of base classifiers based on algorithms of machine learning and wavelet transforms is proposed to detect anomalies. Authors examine the process signal characteristics and wavelet analysis advantages for signal preprocessing. An approach to the anomaly detection was developed based on a model ensemble. This approach was previously tested on actual process signals.