Analysis of the Behavior of a Technical Parameter in a Limited Area of Its Time Series for Forecasting Tasks

Abstract

To assess the performance of a technical object, it is necessary to determine the possible values of the controlled parameters at the predicted time interval. Evaluation is carried out during the operation of the object. Calculations are usually performed in the background. Previously measured parameter values are used for evaluation. They represent a time series. The possibility of forecasting depends on the nature of the time series segment on which the calculations are made. If the plot of the series is characterized as persistent, the forecast is possible. Otherwise, the prediction may result in invalid results. It is shown that the fractal index can be used for local estimation of the character of the series segment, on the basis of which the forecasting is carried out. Unlike the Hurst index, which requires a large amount of data, the fractality index works with small samples. This allows you to determine the nature of the plot of the series, which contains several dozen values.