COMPARATIVE ANALYSIS OF FILTRATION METHODS IN EXPERIMENTAL STUDY OF DYNAMIC PROPERTIES OF OBJECTS

The condition for successful application at the local level of control of industrial controllers and programmable logic controllers is the presence of integrated software for automatic tuning of the parameters of the control algorithm. The dynamic properties of an object are most often determined by the results of an active experiment directly on the object through an identification procedure based on pre-processed data. The main technologies of time series filtering are analyzed and the system of indicators for their comparison is selected. A series of simulation experiments to obtain overclocking characteristics with different degrees of addition of additive barriers to entry and exit are planned and conducted. Simple engineering algorithms of exponential, median, moving average simple and weighted and other window filters are investigated for the efficiency of smoothing in the mode of off-line and on-line processing of the obtained data on the indicators of relative root mean square and integral mean deviation from the baseline. methods. Based on the results of the research, conclusions are drawn about the real effectiveness of on-line filtering, recommendations for their practical use on the hardware platforms of local controls and the low level of smoothing of off-line methods in a limited experimental sample. The directions of further researches in the direction of application of more difficult filtering algorithms for signals with a high level of interference are established.