Тип публикации: статья из журнала
Год издания: 2025
Ключевые слова: industrial sensor data, anomaly detection, clustering, k-means, dbscan, equipment monitoring, промышленные сенсорные данные, обнаружение аномалий, кластеризация, k-средних, мониторинг оборудования
Аннотация: The paper investigates the possibility of automated anomaly detection in industrial mining sensor data based on clustering algorithms. Time series of technological parameters obtained during continuous equipment operation over several dozen days were used as input data. After data cleaning, normalization, and preliminary processingПоказать полностью, k-means and DBSCAN algorithms were applied to identify stable operating modes and deviations from them. Experimental results demonstrated the formation of typical clusters corresponding to the main equipment operating regimes, as well as a limited proportion of anomalous observations associated with transitional and unstable states. A comparative analysis of the algorithms revealed differences in their sensitivity to data structure and types of anomalies. The results confirm that clustering methods can be effectively used for primary condition monitoring of industrial equipment and anomaly detection without the need for labeled datasets or complex predictive models.
Журнал: Известия Тульского государственного университета. Науки о Земле
Выпуск журнала: № 4
Номера страниц: 382-388
ISSN журнала: 22185194
Место издания: Тула
Издатель: Тульский государственный университет