DEVELOPMENT OF MODELS OF CONDITION FORECASTING FOR TECHNICALLY COMPLICATED OBJECT’S ELEMENTS 

Tyurin Mikhail Vladimirovich, candidate of technical sciences, senior stuff scientist, Scientific Research Institute of Physical Measurement (440026, 8 Volodarsky street, Penza, Russia), preobrazovatel@niifi.ru
Fomin Sergey Anatol'evich, chief specialist, department of industrial safety and environmental supervision, Center for exploitation of ground-based space infrastructure (107996, 42 Shchepkina street, Moscow, Russia)
Prokof'ev Oleg Vladimirovich, candidate of technical sciences, associate professor, sub-department of computer science, mathematics and humanities sciences, Financial University under the Government of the Russian Federation (Penza branch) (440052, 33b Kalinina street, Penza, Russia)

629.78.05

Background. Primary signal transducers, that convey information about the state of technically complex object (TCO), differ in variety of set of output quantities, of change speeds and of dynamic value ranges. There is the need for formulation of the adaptive selection procedure of the best model of time series to predict the state of the system element. The selection process should take into account the possibility of software and hardware implementation in terms of the information environment monitoring and control system. The assessment of the future element state is important for the prevention of accidents and emergency situation of technically complex object.
Matherials and methods. The paired-dominant choice of the time series model was applied. UML-modeling methodology was used, which provides the transition from the analysis and structuring of the primary measurement data till the generation procedure of database structures and of program code. The procedure of definitive choice of the "best" model was formulated as the multicriterial task of discrete optimization.
Results. The complex description of information environment of monitoring and control system of TCO was built based on the diagram MMD of physical objects. The algorithm of the paired-dominant choice of the mathematical model of the time series has been proposed on the basis of decision tree. Additional limitations were designed for the case of occurrence of ambiguous decision about the choice. Recommendations for choosing model in the conditions of small samples of observations of the time series were made.
Conclusions. Choosing the best time series model to predict the state of the TCO is possible on the basis of a decision tree implementing the algorithm of the paireddominant choice. When short samples of the time series observations are necessary, the models set is corrected.

predictive modeling of the state, technically complex object, monitoring and control system, prevention of accidents and emergency situation.