CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
ANALYSIS OF INFLUENCE OF INTERVAL RESET STATISTICS SIMULATION EXPERIMENT FOR EVALUATING THE ACCURACY OF AVERAGE INTEGRAL
Article retracted 05.02.2018
 |
Read | Solntsev Alexey A., Prikhodko Mikhail V., Kudryavtsev Andrey Yu., Kotov Andrey A. ANALYSIS OF INFLUENCE OF INTERVAL RESET STATISTICS SIMULATION EXPERIMENT FOR EVALUATING THE ACCURACY OF AVERAGE INTEGRAL // Caspian journal : management and high technologies. — 2011. — №3. — pp. 53-62. |
Solntsev Alexey A. - Ph.D., Moscow Automobile and Road State Technical University (MADI), 64 Leningradsky av., Moscow, 125319, Russia, kafedra@asu.madi.ru.
Prikhodko Mikhail V. - Ph.D., Deputy Director, "LonMADI", 64 Leningradsky av., Moscow, 125319, Russia, kafedra@asu.madi.ru.
Kudryavtsev Andrey Yu. - Ph.D., Engineer, company NPVF "WELDING", 9 Cabelnyi passage, Cheboksary, 428003, Chuvash Republic, Russia, kafedra@asu.madi.ru.
Kotov Andrey A. - Ph.D., Moscow Automobile and Road State Technical University (MADI), 64 Leningradsky av., Moscow, 125319, Russia, kafedra@asu.madi.ru.
Most modern methods developed to obtain estimates on the results of simulations suggest stationarity of the output process and, therefore, the rejection of the transition period. This article has evaluated the accuracy of the estimate, taking into account reset statistics accumulated during the initial period of simulation. It is assumed that the covariance function conditionally non-stationary Gaussian process defined on the basis of the theorem on normal correlation for a given stationary Gaussian process with autocorrelation function in the form of a mixture of two exponentials. The analysis of the expectation of mean integral evaluation of conditionally non-stationary process provided clear statistics accumulated during the initial interval of the simulation. The analytical expressions for the expectation and variance of mean integral evaluation depending on the discharge statistics are given. It is shown that the determining factor is the extent of the correlation process, and the duration of the interval reset statistics for the variance significantly reflects on dispersy of integral evaluation. Reset of the initial statistical variance of the estimate increases, and thus reduces its accuracy. The displacement of the expectation of evaluation also depends on the duration of the interval reset. Duration of reset interval statistics reduces the systematic error estimate of the mean value. Thus, the task of choosing the initial reset interval statistics by the contradictory criteria requires the construction of a convolution of the original criteria. As such convolutions are encouraged to use estimates of the probability to a given interval. For small values of the covariance reset interval is optimum in the neighborhood of zero, ie non-record of any values of the transition leads to a decrease in the probability of belonging to the interval. Performed study provides practical recommendations for handling the output of simulation process in order to improve accuracy and reduce the time of modeling.
Key words: imitation,modeling,Gaussian processes,the mean integral evaluation,variance,expectation,the trend in non-stationary processes,autocorrelation,convergence,optimization.,
Причина ретракции: материал статьи в основном заимствован из докторской диссертации В.М. Черненького «Процессно-ориентированная концепция системного моделирования АСУ» (защищена в 2010 г. в г. Москве).
Решение о ретракции принято редакционной коллегией журнала «Прикаспийский журнал: управление и высокие технологии» 05.02.2018 г.