On Some Key Subjects In The Detaction Of High Speed Railway Overhead Contact System

High-speed railway OCS (Overhead contact system) on-line vehicular detection is the basis used to evaluate the safety performance and current collection quality, and the basis of so-called state repair used to find faults or hidden troubles for maintenance department, and the measure to scout device quality and analyze fault reasons for production scheduling department. Since 1980s, vehicular on-line OCS detection has made considerable progresses from scientific research to practical application. However, with the rapid increase of train velocity nowadays, OCS detection encounters new problems: interference and noise restrict further advance of the measurement accuracy of high-speed railway OCS dynamic parameters; the measurement of pantograph-catenary kinetic parameters needs that the additive mass at the pantograph slide to be minimized; uniform maintenance standard of OCS parameters is required to be set up. Meanwhile, the advancement of other related technology promotes OCS detection to adopt new measures and methods to meet the needs of production development.From the view points of system and interconnection, this thesis studies topics of following aspects: adaptive anti-disturbance technology used in high-speed OCS detection, comprehensive analysis of dynamic parameters, speed value transform treatment, and analytical calculation algorithms of pantograph dynamic response etc.This thesis introduces reference image signal that has undergone the same process as the useful weak signal, and through Lagrange phase shift interpolation and statistic steepest descent algorithm, the system can adaptively filter the low frequency noises whose frequency ranges are hard to distinguish with that of useful signal and stochastic noises. An improved adaptive filter is constructed in this thesis. By the introduction of two cannel of observation samples, the observation samples of useful signal contaminated by noises as expected signal, and the observation samples of approximated noise signal as the input of the filter, the optimized solution of weights of adaptive filter then can be obtained by the minimization of least squares performance. Pull-off value computed from contact pressure may create errors due to sudden change of the pressure value. Through power spectral density analysis, adopting low band filter of 75Hz has certain good effects on this phenomenon. From the systematic view, considering the zigzag feature of pull-off value and the manual measured value of registration arm as teachers, and adopting the self-learning character of neutral network, the neutral network models and the fuzzy logic inference machine model to calculate the pull-offvalue are obtained.Aider centering, de-dimension and standardization processing of the massive measured data, hierarchical clustering and k-means algorithms are used to classify the data through their own inherent characteristics of knowledge, and multivariable regression analysis is then implemented. Through residual analysis and hypothesis testing, computation models of each class are obtained to compute the standard limit value of the hard spot, and the relationship between the dynamic parameters and the velocity at each class.Pantograph slide can be equivalent to engineering spring beam. When pantograph-catenary system is under operation, slide beam will appear lateral vibration and present various modes. Through analysis, we know there are vibration sensitive spots at the bottom of the pantograph slide. Based on superposition principle, vibration offset formula can be obtained, and the response relation matrix between the force and displacement can be established. The transfer function matrix can be obtained by the inversing of response relation matrix. In practical applications, the parameters of pantograph-catenary contact pressure, impact of hard spot, height of contact wire, pull-off value, and time of off-line can be calculated totally from the measured value of sensitive spots of pantograph slide. The method presented in this paper can overcome the inherent defect of both the traditional contact-type and the novel non-contact type of detection effectively, and meet the needs of high-speed railway OCS detection.