M.I. Dli - Dr.Sc. (Eng.), Professor, Head of the Department of Management and Information Technologies in Economy, Branch of National Research University «Moscow Power Engineering Institute» in Smolensk. E-mail: midli@mail.ru A.Y. Puchkov - Ph.D. (Eng.), Associate Professor, Department of Management and Information Technolo-gies in Economy, Branch of National Research University «Moscow Power Engineering Institute» in Smolensk. E-mail: putchkov63@mail.ru

Modern steam boiler management systems include a big set of hardware that realize the required control laws. The high level of responsibility for an object of managing uninterrupted operation demands the existence of diagnostic systems which control data units, metering devices and actuators serviceability. Existing diagnostic systems allow to find in the main such malfunctions as breaking of circuits, signal derating, a current status of object state relay module quantized output non-conformity. Simultaneously managing processes trends identification algorithms in diagnostic systems stay weakly implemented, which can lead to an unwanted managing object state and instrumental inverse tasks. Artificial neural networks application with prefatory fuzzy Kalman filtration of incoming signals is suggested to eliminate algorithm omission. A numeric experiment was carried out in the programming environment Simulink to establish a serviceability of the suggested development of Plant Control Systems architecture. The second stage of the superheater TP-81 was chosen as the managing object. The situation, when the output and input metering is noisy, is created in the model. The drift of an entering measuring instrument from a real object input parameter`s definitions was simulated by sinusoidal signal injection. Artificial neural network solves an inverse task basically on output metering which has passed the prefatory filtration, and forms an approximate solution on its output, that is a signal which really appears at entry to the managing object. Processing of the obtained results of simulating has shown that the ANNs copes well with the task of errors recognition of input counter till the relation signal/noise at the output counter is more about 35 db.

 

1. Tikhonov A.N., Arsenin V.JA.Metody reshenija nekorrektnykh zadach. M.: Nauka. 1979. 284 s.
2. Hansen P.C. Discrete Inverse Problems: Insight and Algorithms. Philadelphia: SIAM2010. 213 p.
3. Sizikov V.S. O sposobakh nevjazki pri reshenii nekorrektnykh zadach // ZHurnal vychislitelnojj matematiki i matematicheskojj fiziki. 2003. T. 43. № 9. S. 1294-1312.
4. Kruglov V.V., Dli M.I., Golunov R.JU. Nechetkaja logika i iskusstvennye nejjronnye seti: Ucheb. posobie. M.: Izd-vo Fiziko-matematicheskojj literatury. 2001. 224 s.
5. Filtracija i stokhasticheskoe upravlenie v dinamicheskikh sistemakh / pod red. K.T. Leondesa. M.: Mir. 1980. 408 s.
6. Izerman R. Cifrovye sistemy upravlenija: Per. s angl. M.: Mir. 1984. 541 s.
7. Puchkov A.JU., Pavlov D.A.Varianty postroenija algoritma poiska reshenija obratnykh zadach s primeneniem nejjronnykh setejj // Programmnye produkty i sistemy. 2012. №2 (98). S. 149-153.
8. Puchkov A.JU., Pavlov D.A.Algoritmy poiska reshenija obratnykh zadach pri nepreryvnom i diskretnom vremeni // Nauchnoe obozrenie. 2013. №1. S. 174-176.
9. Kulakov G.T. Analiz i sintez sistem avtomaticheskogo regulirovanija. Minsk: Tekhnoprint. 2003. 135 s.
10. Puchkov A.Y. Influence of the signal-noise ratio on the accuracy of the approximating metod of the solution of inverse problems // European Science and Technology: Materials of the VII Int. Research and Practice Conf. Munich, Germany. 2014. V. 2. P. 543-546.