Detection of outliers in the data is practically homogeneous technical systems

Abstract

The article presents a method for detecting outliers in data on a particular parameter of practically homogeneous technical systems (hereinafter referred to as PHTS) in the form of time series by their comparative analysis, reflecting cases of impossibility of its implementation. The method is very relevant due to the widespread use of time series in various modern technical systems. To develop the method, a non-exhaustive review of various methods for detecting defects in data was carried out. The main attention in the review was focused on the detection of outliers, since such a defect as omissions in the data is present in them almost explicitly. Some advantages and disadvantages of outlier detection methods are considered, taking into account which the method was developed. The method has broad prospects for its further application, as it can help in identifying outliers in the data of the systems under study, and in case of data similarity, it can give additional confidence to the researcher that the systems whose data are being studied were at the designated time in good condition and almost equal conditions. In turn, the difference may indicate the presence of outliers in the data, the causes of which may be: malfunctions of one or another of the systems or in the systems for collecting and storing data about them, the influence of an unaccounted factor on one or another system. Despite some subjectivity, the method has a significant plus in the form of flexibility, in addition, it does not require the construction of complex models that determine the reference behavior of the investigated parameter of the PHTS for a comparative analysis of the model data and the systems under study. This fact indicates the promise of its application for data analysis even of complex PHTS.