Research On ZPW-2000 Frequency-shift Signal Detection Methods Based On Nonlinear Least Squares Criterion

The ZPW-2000 track circuit is extensively applied to existing railways and passenger dedicated lines in China.In the track section equipped with ZPW-2000 track circuits,the ZPW-2000 frequency-shift signal is transmitted through the rail and is continuously sensed by the on-board receiving coil.The on-board computer resolves the carrier frequency information and the low frequency information from the samples of the sensed voltage.According to the carrier frequency information and the low frequency information,on-board devices and train drivers can grasp the states of the front sections which are either idle or occupied.In order to ensure the reliability of this track circuit based ground-train communication approach,the accuracy,the real-time detection capability and the interference immunity of the ZPW-2000 frequency-shifted signal detection method must be concerned simultaneously.At present,the periodogram method is dominantly applied in on-board devices,and by which the accuracy,the interference immunity,and the real-time detection capability can be ensured simultaneously.However,under the scenario that the sampling duration is limited by the real-time detection requirement,the detection accuracy and the in-band interference immunity of the periodogram method are susceptible to some problems such as spectral leakages.In order to further enhance the ZPW-2000 signal detection performance,according to the existing literature,this dissertation attempts to study detection methods based on nonlinear least squares criterion.So as to clearly describe the typical characteristics of the ZPW-2000 frequency-shift signal,the model of the on-board-device-received signal is established at first.This model is composed of the ideal ZPW-2000 signal model,the additive interference model,the multiplicative interference model,and white Gaussian noise.Secondly,to simplify the problem,the approximate model of the received signal is established.According to the approximate model,the maximum likelihood estimation criterion,namely the nonlinear least squares criterion,for the carrier frequency and the low frequency is derived.By deriving and analyzing the maximum periodogram criterion,the problems encountered by the periodogram method are summarised.Then,for on-board detection of the ZPW-2000 signal,a detection algorithm based on nonlinear least squares method is designed.For obtaining the spectrum under fault conditions,the simplified model of the received signal is established by analysis,and a detection algorithm based on rapid RELAX(RELAXation)method,which also follows the nonlinear least squares criterion,is designed.Finally,the real-time performance of the designed algorithms are tested by time complexity analysis and simulations,the Cramér–Rao bound for the carrier frequency estimates and the low frequency estimates under the simplified model of the received signal and the approximate model of the received signal are all derived,and with the simulation examples generated according to the model of the on-board-device-received signal,the accuracies and interference immunities of the designed algorithms are tested by simulations.The simulation results and the analysis results can be concluded as follows: the designed algorithms are able to meet the real-time detection requirements in the simulation environment.When the sampling duration is short,the estimation accuracies of the carrier frequency and low frequency of the designed algorithms are significantly better than those of the periodogram method,and thus the sampling duration required by detection methods for correct decoding can be further reduced.Under white Gaussian noise assumption,the estimation biases and variances of the carrier frequencies and the low frequencies of the designed algorithms are all lower than those of the periodogram method.The detection algorithm based on nonlinear least squares method has favorable immunities to the in-band traction harmonic interferences and the interferences of the same type track circuit signals,but its immunity to the transient impulse interferences still needs to be improved.Comparing with the periodogram method,the detection algorithm based on rapid RELAX method possesses a better performance in frequency resolution,but its interference immunity still needs improvement.