Processing math: 100%
Application of Neural Networks for Predicting UTC Local Time-Scale With Clock Ensemble | IEEE Journals & Magazine | IEEE Xplore

Application of Neural Networks for Predicting UTC Local Time-Scale With Clock Ensemble


Abstract:

The local time-scale UTC( k ) provides precise time for each country and serves as the basis for the coordinated universal time (UTC). The use of clock ensemble to gene...Show More

Abstract:

The local time-scale UTC( k ) provides precise time for each country and serves as the basis for the coordinated universal time (UTC). The use of clock ensemble to generate local time scales has been used or is advancing in laboratories. To improve the UTC( k ) performance in the case of clock ensembles, a method is proposed to predict the local time-scale based on an ensemble time-scale perceptron (ETSP). The UTC( k ) and clock data are obtained from circular T , frequency drift files, and clock data files. We classify those data according to [UTC–UTC( k )] and use high-performance time scales directly or simply weighted as network training references. Meanwhile, we design an ETSP based on multilayer perceptron (MLP) composed of weighted nonlinear neurons and perform prediction estimation and implicit calculation of weights for time-scale generation. In this article, the official data provided by the International Bureau of Weights and Measures (BIPM) is used for experimental verification. The accuracy is verified by mean—UTC–UTC( k )— and max—UTC–UTC( k )—, and the stability is verified by Allan deviation (ADEV) and maximum time interval error (MTIE). The results show an order-of-magnitude performance improvement in both accuracy and stability compared to the original prediction.
Article Sequence Number: 1000809
Date of Publication: 28 November 2022

ISSN Information:

Funding Agency:


References

References is not available for this document.