Study On The Methods Of Data Preprocessing And Performance Analysis For Atomic Clocks

Act as the key equipment establishing and maintaining the high-precision time-frequency standard, atomic clock, its performance affects the precision of navigation, positioning and time keeping directly. For accurately analyzing the performance of atomic clocks, it is very necessary to study the presenting methods of atomic clock's property specifications, and remove or weaken the influence of the gross error data, noise and some other abnormal datas in raw atomic clock data on analysis of application. Therefore, based on deep research of the fundmental theory of atomic clock signal, this thesis focuses on the systematic research and study of the methods of data preprocessing and performance analysis for atomic clocks. The main contents are as follows:1. For the measured atomic clock data, the median gross error detection method is analyzed. Because of the limitations of its application range, then the improved median gross error detection method is put forward. In addition, combination of the characteristics of wavelet analysis, the wavelet analysis method for dealing with the gross error data is studied. The comparative analysis results indicate that the two methods are not only simple and effective, but also of great applicability.2. Based on the basic principles of wavelet threshold denoising, it is studied that the parameters factors, such as the wavelet functions, threshold selection rules and regulating methods and so on, which influence the denoising effects of commonly used threshold denoising algorithm. Grounded on the parameters selection results, it adopts the improved algorithm for denoising, that is, wavelet packet algorithm, the second wavelet transformation and improved threshold algorithm and otherwise. Then, the best combination of parameters and improved algorithms which fit for the atomic clock data denoising are got by numerical examples.3. The various time-domain variances and periodic components suffered by atomic clocks are analyzed by the measured atomic clock data. Then, the general calculation formula of a variety of time-domain variances is studied and discussed. Some of the calculation formulas and the power law models of time-domain variances are derived. And then, the characteristics of the analysis methods of time-domain variances are summed up.4. According to the principles of definition and calculation of the time-domain variances, the concept of wavelet variance series is put forward, and the calculation methods of the wavelet variances estimation and confidence interval are made clear. By calculation and analysis, the relationships between wavelet variances and time-domain variances are proved and supplementally explained. Then, the feasibility and superiority of wavelet variance methods for stability analysis are pointed out by synthesizing. Thereby, we perfect the scale-domain wavelet variance methods which is used to analyze the frequency stability of atomic clocks systematically and thoroughly.5. Based on the scale pollution model of atomic clock inaccurate data, the robust M estimation algorithm of wavelet variance is studied. Through the examples, the robust estimation effects are analyzed, the gross error detection steps of this model is discussed as well. The mixed robust estimation algorithm and scale-based gross error detection method of wavelet variance fit for that model are achieved finally.