Формування простору діагностичних ознак на основі перетинів ядер Вольтерра

• Дата: 2018
• Автор: Фомін, Олександр Олексійович
• Формат: Стаття
• Мова: Ukrainian
• Опубліковано: Kamianets-Podilskyi National Ivan Ohiienko University 2018
• Онлайн доступ: http://mcm-tech.kpnu.edu.ua/article/view/140049
• Назва журналу: Mathematical and computer modelling. Series: Technical sciences

The class of problems of indirect control and diagnostics of complex continuous nonlinear dynamic objects of various physical nature is considered. These problems belong to the class of inductive modeling problems, the essence of which lies in the transition from empirical information to the mathematical model with the aim of obtaining new knowledge and decision-making under conditions of substantial incompleteness and a priori uncertainty of information.The purpose of the work is to increase the reliability of the diagnostic procedure on the basis of the formation of informative feature spaces for the creation of effective tools for diagnosing objects of different nature.The method of model diagnostics of non-linear dynamic objects is considered, based on the description of objects in the form of Volterra integro-power series whose multidimensional kernels are used in constructing the space of diagnostic features.A method is proposed for the formation of a space of diagnostic features based on Volterra kernels by a directional search of arbitrary cross sections of nuclei, which, unlike the existing method, which uses for the formation of a space of diagnostic features of diagonal nuclear bonds, can significantly increase the reliability of diagnosis.A step-by-step algorithm for the formation of a space of diagnostic features is presented. This algorithm consists in the sequential implementation of identification operations of the diagnostic object, the formation of a family of diagnostic models of the object, the construction of a state classifier, and the selection of the resulting space of diagnostic features.It is established that the second-order kernel provides the most complete information for diagnosing the states of an object of research. Analysis of the diagnostic value is formed on the basis of the diagonal cross section of the Volterra kernels of the second order of feature spaces showed that the initial region of intersection corresponding to the first three samples has the highest informative value.