WSN-based Railway Track Research On Vibration Monitoring And Feature Recognition

It is an important means to maintain the railway line by using track vibration to assess the operating state of rails. The development of wireless sensor network technology (WSN) has shown significant opportunities for identifying the operating state of rails. The method, which uses WSN to monitor the track state, offers a lot of advantages such as long-term service, no human error, real-time and full-line assessment. In spite of many advantages above, the research efforts are relatively limited compared to the current track static inspection and dynamic detection. In this thesis, in order to establish and improve the monitoring platform, WSN-based track monitoring platform is systematically studied in terms of realization of monitoring platform, denoising of vibration data and identification of vibration characteristic. The main contents are as follows:(1) A load adaptive proactive server is developed using the queuing network model. The concurrent data processing method for the information from multi sensor nodes is proposed. A network server software is developed. The software handles multiple connections and large data requests, and implements design-time performance evaluation and the run-time optimal adjustment of configuration through the asynchronous mechanisms under operation system. It also solves the problem of data packets losing during the transmission, realizes many functions including receiving and saving real-time data, sending and uploading control parameters, etc.(2) Both of the control & communication layer and data storage & analysis layer are designed and implemented, and a performance test is carried out on the monitoring system platform. A gateway is achieved to coordinate heterogeneous network for data forwarding; a data storage and management model is constructed based on relational databases and data analysis machine; the data flow of interactive manipulation for client software is designed.13 gateways and 104 nodes are installed in the test site. Performance testing shows that the loss rates of all nodes are less than 5% during the 30 days’ test; the loss rates are equal to zero with maximum throughput of 0.52 MB/s during an hour; the delay is less than 1 min from data collection to be displayed on the client interface when the delay of packet loss is less than 4 min. Besides, the system has been installed on the Beijing-Tianjin high-speed railway, withstood various tests and got rich track information for two years.(3) The Bayes wavelet packet denoising method is studied to reduce the noise of track vibration signal. After comparing treatment effects of different filtering algorithms, the Bayes wavelet packet denoising method is the optimal denoising algorithm. The decomposition level of wavelet coefficients is determined by their autocorrelation function. The noises of wavelet coefficients are filtered out based on the each coefficient characteristic, while keeping the local non-stationary and jumping feature of vibration signal (the SNRs of denoising results of three track structures lie within 4.50-4.70 range). This is very useful for the final vibration analysis.(4) The vibration characteristics analysis theory is studied and a track vibration experiment is carried out. Vertical wheel-rail force analysis model is established, and two relationships between running speed and wheel-rail interaction force (P1andP2), wheel-rail displacement are studied. A positive correlation between running speed and wheel-rail interaction force is confirmed. The vertical displacement is closely related to forceP2. A wavelet method of track vibration data, which chooses the complex Morlet wavelet as mother wavelet, is established. The acceleration data from the WSN monitoring platform are analyzed to obtain the vibration characteristics with different track structures and running speeds. The results reveal the varying laws of frequency and power with the test positions and motivations changing in the low, medium and high frequency band. A shifted frequency, which has a linear relationship with the travel speed, is found in the medium frequency band. The fastener state can be accurately determined by the time-varying characteristics of frequency and energy.Through theoretical verification and actual experiments, the dissertation completes a series of researches including gateways, network server, data storage and management, client software, denoising and vibration feature extraction, and firstly develops a large-scale WSN track vibration monitoring system to analyze track vibration in the world.