Nadezhda Glebovna Yarushkina, Ulyanovsk State Technical University, Doctor of Engineering, Professor; graduated from the Faculty of Radioengineering at Ulyanovsk State Technical University; First ViceRector  ViceRector for Scientific Affairs of Ulyanovsk State Technical University; an author of more than 250 papers in the field of soft computing, fuzzy logic, and hybrid systems. [email:
This email address is being protected from spambots. You need JavaScript enabled to view it.
]N. Yarushkina, Evgenii Nilolaevich Egov, Ulyanovsk State Technical University, Postgraduate Student and Assistant at the Department of Information Systems of Ulyanovsk State Technical University; graduated from the Faculty of Information Systems and Technologies of Ulyanovsk State Technical University; an author of articles in the field of intellectual analysis of time series. [email:
This email address is being protected from spambots. You need JavaScript enabled to view it.
]E. Egov


The Algorithm for Identifying New Anomalies in Technical Time Series Diagnosis
The article discusses the ways to diagnose the time series in order to detect anomalies in them. The authors propose to determine the number of each point of the values of the two parameters. Also a set of situations related to changes in the values of these parameters between the points should be prepared. While analyzing series, the frequency of each situation occurrence should be determined. If the probability of situations occurrence is less than 0.01, then such situations may be attributed to an abnormal ones. On the basis of the previous situation choice, a template that allows identifying these anomalies in the future is created. As one of the pairs for situations identification, the entropy measures values obtained from fuzzy time series are proposed to use. The first measure of entropy is calculated by the value of the membership function point compared to the fuzzy label. The second measure of entropy is calculated on the basis of the deviation of the actual value trends from the forecasting one. Series analysis is performed on the basis of the second pair. This pair represent “fuzzy label  fuzzy trend” one. This pair was introduced to identify longterm stay in the areas of certain states, which can be attributed to the abnormal ones. It also describes the algorithm to identify previously unknown anomalies and search of anomalies patterns. The experiment was carried out in order to check the efficiency of the algorithm. Time series of physical quantities characterizing work of important units of helicopter engines in which it was necessary to reveal the presence of defects were investigated. The main interest of this paper is the anomaly detection algorithm based on the measure of the uncertainty of the time series. The article is intended for professionals diagnosing technical systems. Entropy measure, diagnosis, time series, anomalies.

