Tesis Doctorales de la Universidad de Alcalá
Más información
Autor/aYuan , Lei
Director/aHernández Alonso, Álvaro
Codirector/aYuan , Yang
Fecha de defensa31/10/2018
CalificaciónSobresaliente Cum Laude
ProgramaElectrónica: Sistemas Electrónicos Avanzados. Sistemas Inteligentes (RD 99/2011)
Mención internacionalNo
ResumenAs one of the lifeblood of national economic development, rail transportation carries heavy passenger and cargo tasks. Rail line is an important part of railway transportation. The accident loss caused by track breakage cannot be estimated. Its integrity affects the safety of people's lives and property. Real-time, accurate and stable fault detection is of great significance. The breakage detection is realized by analyzing and processing the signals transmitted in the track. Therefore, it is of great significance to study the denoising of signals in the breakage detection system and improve the signal recognition algorithm. Firstly, in order to overcome shortcomings of complex wavelet selection, mode mixing, et al. in the current ultrasonic guided wave (UGW) signal denoising algorithm, this work proposes the use of variational mode decomposition (VMD) algorithm to suppress the interference of UGW signals. Based on the denoised UGW signal, a feature extraction method based on the UGW signal under different track status is proposed. The difference between the extracted feature (the root mean square voltage value Vrms and the frequency component Fp corresponding to the highest amplitude point in the amplitude-frequency characteristics) in three different track status and their respective temporal and spatial dependencies are proposed. The idea of real-time identification and classification of track status by the recurrent neural network (RNN) is also proposed. Through the simulation and experimental results of denoising algorithm, the UGW signal can be reconstructed correctly, the extracted features are obviously different in three track conditions, and through the accuracy of the classification results and the display on the two-dimensional plane, it can be known that the RNN network can detect the track status in real time and discover the track breakage situation in time. Secondly, aiming at the situation that the UGW signal cannot propagate through the joint maintenance of broken repaired structure, the previous UGW-based track breakage detection system cannot work anymore, the improved real-time track breakage detection system combined with the detection principle of the track circuit is proposed The improved system is compatible with the transmission and processing of UGW and electrical signals, and is suitable for more complex train operating environments. At the same time, for the part of the improved track breakage detection system based on the principle of track circuit detection, two simulation models of railway line are proposed. These two models provide a suitable simulation platform for analyzing the transmission and detection of electrical signals.