Abstract
Purpose of research. The article is devoted to the development of production predicting models and their stability conditions.Methods: Predicting models are actively used in modern control systems, in information support intellectual systems of decision-making. They have a huge role in any activity connected with signals' processing including anomalies detection of various technological processes and assessment of risk potential of critical information infrastructure objects. They can also be used in monitoring systems of security threats. Special class among predicting models is represented by the models based on experiences of proceeding processes (for example, regularities taken from the data which are saved up as a result of an object work).Results. Virtual "instant" model of an object belonging to this class is described in the article. It is presented taking into account multiple and large-scale decomposition of entrance influences vectors and the forecast of an object output. The described model gives the forecast without possible future conditions of an expected background. The approach based on the wavelet-analysis which is characterized by a unique opportunity of detailed frequency analysis in time is developed for stability study of virtual "instant" model. Stability conditions of the predicting model are received on the basis of this approach. This model has allocation conditions for approximating and detailing components for four types of ratios between memory depth on input and output.Conclusion: Predicting model of oil processing in which memory depth on an input is more than memory depth on output is described in the article. It is shown that the accuracy of virtual "instant" model forecast is higher than linear predicting model has at rare data of laboratory analysis. One of stability conditions depending on decomposition depth is shown for the constructed model. On the basis of received results analysis it is possible to draw a conclusion on applicability of received stability conditions for risk potential assessment of process development forecast implementation in monitoring systems of security threats.
Highlights
Predicting model of oil processing in which memory depth on an input is more than memory depth on output is described in the article
It is shown that the accuracy of virtual "instant" model forecast is higher than linear predicting model has at rare data of laboratory analysis
One of stability conditions depending on decomposition depth is shown for the constructed model
Summary
Predicting models are actively used in modern control systems, in information support intellectual systems of decision-making. They have a huge role in any activity connected with signals' processing including anomalies detection of various technological processes and assessment of risk potential of critical information infrastructure objects. They can be used in monitoring systems of security threats. Special class among predicting models is represented by the models based on experiences of proceeding processes (for example, regularities taken from the data which are saved up as a result of an object work)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have