A method for remote automatic monitoring of railway track deformation using a pre-trained generative neural network

Abstract

A method for remote, automated current and predictive monitoring of railway track deformation based on the generation, emission, reception of signals reflected from discrete sections of the railway track, and conversion into a beat frequency F б =F оп -F от difference frequencies, reference frequency F оп and reflected F от, analog-to-digital conversion, inter-period accumulation by radar sensors placed on the monitored section of the railway track, transmission of information via communication channels to a remote automated operator workstation, characterized by the fact that in each range resolution element the amplitudes and phases of the reflected signals are determined, compared element by element with previously recorded reference signals, the results of the comparison are weighted taking into account measurement errors, the coordinates and area of ​​deformation, the levels of subsidence and uplift of the railway track in the range resolution elements are estimated, their results are displayed to the operator by a pre-trained artificial intelligence system, which calculates the tendency of the deformation of the railway track using previous measurements, the periods of reference measurements are determined depending on weather conditions and possible changes in the state of the railway track,In this case, radar sensors emit and receive signals in a narrow beam, within the width of the railway track.