Open AccessA Study of an Engine Anomaly Detection Model IForest-ADOA
Author(s) -
Liu Liu,
Min Xiao
Publication year - 2022
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2171/1/012075
Subject(s) - anomaly detection , anomaly (physics) , cluster analysis , computer science , stability (learning theory) , data mining , artificial intelligence , machine learning , pattern recognition (psychology) , physics , condensed matter physics
In the environment of the Industrial Internet, automatic and effective anomaly detection methods are of great importance to ensure the safety and stability of engines. However, in practical engine applications, traditional machine learning anomaly detection methods often have limitations due to a large number of unmarked samples and high latitude data. In this paper, an engine anomaly detection model based on the ADOA model is proposed, combining the unsupervised learning iForest model with the semi-supervised learning ADOA model. Through the anomalous samples detected by iForest, the ADOA model is used to learn their features and patterns and perform clustering, to achieve the anomaly detection of the samples and improve its accuracy at the same time. Experiments based on a real engine dataset show that this model has good results in terms of anomaly recognition rate and anomaly misclassification rate for anomaly detection of high-dimensional data generated by real environments.