Method for detecting and eliminating data time series outlier in high-speed process data sensors

Abstract

The objective is to identify and eliminate measurement errors in data time series of data sensors of high-speed processes. Fractal analysis is used to identify and eliminate outliers in data time series of data sensors of high-speed processes. The proposed method provides an accurate detection of an outlier and an acceptable error in a series characteristic modification when the outlier is replaced with a predicted value. As a result, the time series becomes persistent again. The originality of the method consists in using such a hidden regularity of time series as a correlation of values to detect measurement errors. The importance of the study is determined by the immediacy of the problem of detecting and eliminating outliers in data time series of data sensors with stringent requirements imposed upon to reduce decision-making time in the event of critical situations