An upgrade of predictorfunctions based on the analysis of time series for mashing beer wort
Микола Володимирович Чернецький, Василь Дмитрович Кишенько, Анатолій Петрович Ладанюк · Eastern-European Journal of Enterprise Technologies · 2015
The paper presents research findings concerning the complex system of the mashing and brewing section of the brewery by methods of nonlinear dynamics. We have analysed the time series of the variable technological processesof making beer wort. The prior information and analysis on the dimensions of the reconstituted phase trajectories of the object determined our selection of the structures and functions of the basic models of the dynamics in the process of mashing wort; the structures and functions had the form of nonlinear algebraic polynomials and fractional rational functions of the 3-6 order, respectively. As a result, the calculated and established coefficients of the dynamic models have provided adequacy of the experimental data model that is sufficient for practical purposes: the prediction error does not exceed 6 %. Simulation in the VectorODE package made it possible to upgrade predictor functions for predicting the progress of mashing beer wort in chaotic regimes of the process. We have determined the optimal structure and parameters of differential equations as predictors for the criteria of time and prediction accuracy. Besides, we have proved effectiveness of the obtained models for predicting the behaviour of the object in a wide range, including chaotic regimes. The derived models were checked through time test series that showed the effectiveness of these predictor functions: the depth of the prediction was 8-17 minutes with an accuracy of 3 % to 7 %. Using predictor functions ensures the implementation of effective management strategies in the technological complex of beer production in conditions of intermittency.