A.V. Eltishev - Student, Department of Automation and Telemechanics, Perm National Research Polytechnic University. E-mail: eltysheval322@gmail.com L.M. Oniskiva - Ph.D. (Eng.), Associate Professor, Department of Applied Mathematic, Perm National Research Polytechnic University. E-mail: oniskivf@gmail.com A.I. Posyagin - Assistant, Department of Automation and Telemechanics, Perm National Research Polytechnic University. E-mail: posyagin.anton@gmail.com A.A. Yuzhakov - Dr.Sc. (Eng.), Professor, Head of Department of Automation and Telemechanics, Perm National Research Polytechnic University

The given article concerns the comparative analysis of the developed simulation and analytic models of the neural network self-routing analog-to-digital convertor (ADC) which were created in order to calculate the required number of serving neurons depending on the set parameters of ADC and the probability of failure. At this stage the simplified structure of the neural network (NN) was taken as the basic model; its topology presents a circle of sequentially connected basic neurons without regard to the additional bypass connections between them. The simulation model was developed with the help of the AnyLogic tool using the agent modelling method. This model has the logic of posing and serving the requests in the NN with the regard to the spontaneous failure of the neurons either available or occupied by serving a request at the given moment. The probability of the failure at the serving of the incoming request in the given model is calculated by the relation of the total time during which the system was being in the failure status (couldn-t pose the incoming request in the NN) to the whole modelling time. Then, the analytical model of the neural network self-routing analog-to-digital convertor was considered. It is described with the help of the queuing theory (QT). Based on input and output routes properties the kind of the queue was determined, then the vector graph of the given system status was built which helps to define the failure probability of serving - this is the sum of status probabilities when a failure takes place. However, together with the increase of parameters describing the NN the vector graph of the queuing system becomes more complicated which is why it was decided to develop the model allowing to accelerate the required calculations. This model was developed using the AnyLogic tool and it also has the vector graph structure of the queuing system of ADC. The probability of the failure is calculated by the relation of the total time during which the system was being in the failure status to the whole modelling time. Usage of the developed model accelerates the calculations significantly but doesn-t solve the problem of the automatic scaling of the status graph. In order to solve this problem, the improved analytical ADC model was designed. The peculiarity of this model is in the automatic production of status agents corresponding to the status vector graph of queuing system which is produced by the top-level agent presenting the network of queuing serving. The calculation of the failure probability is made the same way. According to the results of the modelling, the comparative analysis was provided, the conclusion of simulation and ana-lytical models adequacy was made, and the further problems of the research were formulated.

 

1. Posjagin A.I., JUzhakov A.A. Razrabotka dvukhslojjnojj nejjronnojj seti dlja samomarshrutizirujushhegosja analogo-cifrovogo preobrazovatelja na osnove nejjronnojj seti // EHlektrotekhnika. 2013. № 11. S. 10−13.
2. Dadenkov S.A., Kon E.L. Analiz modelejj i metodov agentnogo i diskretno-sobytijjnogo imitacionnogo modelirovanija // Izv. SPbGEHTU «LEHTI». 2015. № 5. S. 35-41.
3. JUzhakov A.A. Voprosy analiza analogo-cifrovykh preobrazovatelejj na osnove sistem massovogo obsluzhivanija, funkcionirujushhikh v sluchajjnojj srede. Perm: Perm. gos. tekhn. un-t. 1997. 52 s.
4. Matushkin N.N., Nazarov A.A., JUzhakov A.A. Multiplikativnost raspredelenija sostojanijj zamknutojj SMO pri neodnorodnom vkhodjashhem potoke // Informacionnye upravljajushhie sistemy. Sb. nauchn. trudov. Perm: PermGTU. 1996. S. 34-38.