ANALYSIS AND DATA PROCESSING SYSTEMS

The accuracy of modern information-measuring systems is caused by both measuring schemes and data processing algorithms. In this paper the issues of reducing an algorithmic error of group standards of time and frequency which can be treated as complex information-measuring systems are considered.

The estimation error of the state vector of standard can be reduced by 15-20 percent due to the use of more advanced algorithms for processing measurement information, in particular, through the use of algorithms based on the use of dynamic stochastic models. However, empirical time series containing, at least, no less than a hundred elements are necessary for creating such models. When processing data obtained on daily intervals, this leads to delays of approximately one quarter between the moment of including a new standard in the structure of the group standard and the beginning of using the measurement results obtained with its help. A natural way of reducing this temporary delay consists in creating predictive models for a shortened time series with their subsequent adaptation in the process of obtaining new measurement results.

The article proposes to use the method of a stochastic quasi-gradient which is designed to create a sequence of points in the parameter space of dynamic stochastic models for the adaptation of time series models describing processes of changing relative frequency deviations of hydrogen generators which form the basis of Russian standards of time and frequency. The adaptation algorithm realizing the proposed method is considered. The results of the computational experiment are presented confirming the efficiency of the method when adapting a one-dimensional time series model. The results obtained are generalized for multidimensional models and also for the adaptation of predictive time series models containing deterministic trends along with a stochastic component.

The approach proposed by the authors allows cutting almost by half the temporary delay caused by the accumulation of initial data necessary for creating predictive models used for the estimation of a group standard state.

1. Panfilo G., Harmegnies A., Tisserand L. A new prediction algorithm for the generation of International Atomic Time. Metrologia, 2012, vol. 49, pp. 49–56.

2. Percival D.B., Senior K.L. A wavelet-based multiscale ensemble time-scale algorithm. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency, 2012, vol. 59, N 3, pp. 510–522.

3. Podogova S.D., Mishagin K.G., Medvedev S.Yu., Blinov I.Yu. An algorithm for a group time scale using a moving average over multiple time. Measurements Techniques, 2015, vol. 58, no. 5, pp. 532–538.

4. Levine J. The statistical modeling of atomic clocks and the design of timescales. Review of Scientific Instruments, 2012, vol. 83, art. 021101, pp. 1–28.

5. Khrustalev Yu.P. Statistical and dynamic processing of data obtained when handling time and frequency standards. Measurement Techniques, 2004, vol. 47, no. 6, pp. 555–561.

6. Bezmenov I.V., Blinov I.Yu. Teoreticheskie osnovy postroeniya modelei dlya opisaniya sovremennykh shkal vremeni i standartov chastoty [Theoretical bases of building of models for the description of modern time scales and frequency standards]. Mendeleevo, VNIIFTRI Publ., 2015. 529 p.

7. Greenhall C.A. Reduced Kalman filters for clock ensembles. 2011 Joint Conference of the IEEE International Frequency Control Symposium & European Frequency and Time Forum: Proceedings, San Francisco, CA, 2011, pp. 1–5.

8. Suess M., Greenhall C.A. Combined covariance reductions for Kalman filter composite clocks. Metrologia, 2012, vol. 49, pp. 588–596.

9. Khrustalev Yu.P., Akulov V.M., Ippolitov A.A., Kurysheva L.N. Obrabotka dannykh, poluchennykh po rezul'tatam vzaimnykh izmerenii vtorichnogo etalona vremeni i chastoty [Processing data obtained as a result of reciprocal measuring of secondary standard of time and frequency]. Vestnik Irkutskogo gosudarstvennogo tekhnicheskogo universiteta – Bulletin of Irkutsk State Technical University, 2012, no. 7 (66), pp. 22–28.

10. Khrustalev Yu.P., Serysheva I.A. Avtomatizatsiya protsessa postroeniya dinamicheskikh stokhasticheskikh modelei [Automation of dynamic stochastic model creation process]. Vestnik Irkutskogo gosudarstvennogo tekhnicheskogo universiteta – Bulletin of Irkutsk State Technical University, 2017, vol. 21, no. 9, pp. 95–103.

11. Serysheva I.A., Chekan M.A., Barkhatova L.V., Krupenev E.A. Metod sopryazhennykh gradientov v sisteme avtomaticheskogo postroeniya dinamicheskikh stokhasticheskikh modelei [Conjugate gradient method in the system of predicting model automatic creation]. Vestnik Irkutskogo gosudarstvennogo tekhnicheskogo universiteta – Bulletin of Irkutsk State Technical University, 2018, vol. 22, no. 8, pp. 72–82.

12. Khrustalev Yu.P., Serysheva I.A. Increasing of robustness of estimators of the state of time and frequency standards. Measurement Techniques, 2014, vol. 57, no. 5, pp. 519–525.

13. Box G.E.P., Jenkins G.M. Time series analysis: forecasting and control. San Francisco, Holden-Day Publ., 1970 (Russ. ed.: Boks Dzh., Dzhenkins G. Analiz vremennykh ryadov: prognoz i upravlenie. Vyp. 1. Moscow, Mir Publ., 1974. 406 р.).

14. Borovikov V.P. Populyarnoe vvedenie v sovremennyi analiz dannykh v sisteme STATISTICA [Popular introduction to modern data analysis in the STATISTICA system]. Moscow, Goryachaya liniya-Telekom Publ., 2015. 288 р.

15. Kashyap R.L., Rao A.R. Dynamic stochastic models from empirical data. New York, Academic Press, 1976 (Russ. ed.: Kash'yap R.L., Rao A.R. Postroenie dinamicheskikh stokhasticheskikh modelei po eksperimental'nym dannym. Moscow, Nauka Publ., 1983. 384 p.).

16. Ovcharenko V.N. Adaptive parameter identification in dynamical and static systems. Automation and Remote Control, 2011, vol. 72, no. 3, pp. 570–579.

17. Druzhinina M.V., Nikiforov V.O., Fradkov A.L. Methods of nonlinear adaptive control with respect to the output. Automation and Remote Control, 1996, vol. 57, no. 2, pp. 153–176.

18. Tyukin I.Yu., Terekhov V.A. Adaptatsiya v nelineinykh dinamicheskikh sistemakh [Adaptation in nonlinear dynamic systems]. Moscow, LKI Publ., 2008. 384 р.

19. Ermol'ev Yu.M. Metody stokhasticheskogo programmirovaniya [Methods of stochastic programming]. Moscow, Nauka Publ., 1976. 239 p.

20. Serysheva I.A., Khrustalev Yu.P. Metod stokhasticheskogo kvazigradienta v zadache adaptacii prognoziruyushchikh modelei [Method of stochastic quasigradient in the problem of predictive model adaptation]. Vestnik Irkutskogo gosudarstvennogo tekhnicheskogo universiteta – Bulletin of Irkutsk State Technical University, 2013, no. 12 (83), pp. 25–30.

21. Widrow B., Mccool J.M., Larimore M.G., Johnson C.R. Stationary and nonstationary learning characteristics of the LMS adaptive filter. Trudy Instituta inzhenerov po elektronike i radioelektronike – Proceedings of the IEEE, 1976, vol. 64, no. 8, pp. 37–51. (In Russian).

22. Nevel'son M.B., Khasminskii R.Z. Stokhasticheskaya approksimatsiya i rekurrentnoe otsenivanie [Stochastic approximation and recurrent estimation]. Moscow, Nauka Publ., 1972. 304 p.

23. Serysheva I.A., Khrustalev Yu.P. Adaptatsiya mnogomernykh dinamicheskikh stokhasticheskikh modelei [Method of stochastic quasigradient in the task of adapting of predictive models]. Sovremennye tekhnologii. Sistemnyi analiz. Modelirovanie – Modern Technologies. System analysis. Modeling, 2014, no. 2 (42), pp. 78–84.