CLASSIFICATION OF THE DEGREE OF PARAMETER CHANGE IN REAL TIME BASED ON TIME SERIES POINT CLOUD ANALYSIS

Abstract

The task of building a model for assessing the performance of a technical object has many applications
in the field of controlling various hazardous situations. The need for advanced monitoring of the
technical object state to prevent and control the course of abnormal situations in order to eliminate them
with minimal consequences makes the statement and fulfillment of this task relevant and timely. To perform
the assessment of the state of the technical object it is advisable to use simple models that allow to
obtain the result in real time without significant load on the microcontroller control system. The paper
considers the construction of a model for classifying the dynamics of change in the parameter of a technical
object, which will allow you to predict the change in its state in the process of assessing the degree
of serviceability of the object. The data reflecting the change of parameters in real time and presented in
the form of time series of parameter values are used. The change of the object parameter in time is fixed
with the help of a time window, which moves along the time series, cutting out of the set of initial data a
subset with an unchanged number of time samples. To classify the dynamics of parameter variation, it is proposed to use a representation of the time window points in the form of a Poincaré plot, which is actually
a special type of repetition plot or a type of scatter plot. The ellipse compression factor (ellipticity) is
used as a criterion, which encompasses the point cloud formed during the construction of the scatter diagram,
for the time series of the technical parameter. A methodology for training and using the model,
including the formation of classes of states of the dynamics of the object parameter and the calculation of
criteria, is developed. The model has been tested. The model provides the realization of procedures for
real-time detection of the possibility of an abnormal situation at an early stage of its development with the
help of a microprocessor module located at the lower level of the object monitoring system.