Boriak B. Nonius adaptive filter-predictor – control systems technological process delay compensator

The dissertation is aimed at solving the actual scientific and practical problem of increasing the efficiency of automatic control systems (ACS) with time delays which signals are distorted with non-stationary stochastic noises. The solution of this problem was accomplished by developing the filtering and prediction algorithm of ACS with time delays state-space variables and methods of filtering and prediction algorithm adaptation in the conditions of uncertainty and incompleteness of a priori information about the object, signals and noises. Theuse of theexponential smoothing method as the basic structure of the filter-predictor in technological processcontrol systemshas been scientifically substantiated. A filtering and forecasting model have been developed based on theR. Brown’s double exponential smoothing principle. The use of developed model gave an opportunity to compensates delay error for inertial filters. The method of smoothing factor adaptation that uses the least mean squares method (LMS) was developed for approximation on the interval of filtered signalvalues evaluation. This method allowed to adapt the smoothing factor in the absence of a priori information about changes in the parameters of the noise and the monitored useful signal. This method provides high performance during implementation and the required quality of filtration in the conditions of noise amplitude growth. For the first time the smoothing factor differential method adaptation, which involves the use of two or three nonius filters with smoothing factor different values, is developed. This method allowed to evaluate the quality of the filtration without determining the reference signal on the investigated time domain. Adjustment of the smoothing factor is based on the forecast error difference of two and three filtration loops that function with different values of the smoothing factors. According to the results of the statistical analysis of the relationships between filter parameters and the quality of filtration and forecast, the ranges of parameter values were determined. Setting these ranges of parameters allows to minimize filtering and forecasting errors. Practical recommendations for setting up filters parametersin technological process control systemshave been developed. The possibility of the adaptive filter-predictorintegration into the technological processes of applying insulation to the current-carrying conductor of the high-voltage cable and the production of quartz tubes measuring channels to compensate time delays by obtaining predicted values of filtered signals was investigated. According to the results of the simulation, the measurement error of the predicted coordinate decreased from 5–10 % to 2–3 %.The integration of the adaptive filter-predictor into the industrial robot environmental determination system is implemented. According to the results of the experiment, theservomotor turning time was decreased from 13 % to 33 %, while the error of tracking the distance to the surface was decreased by 21 %, which improved the maneuverability of the IR. Implementation of the developed modifications of the adaptive nonius filter-predictor allows to significantly improve the ACS with time delays working quality in the conditions of a priori uncertainty and the current non-stationarity stochastic characteristics of measuring signals and noise.