Study On Methods For Continuous Measurement And Identification Of The Vertical Interaction Between Wheel/Rail

Railway vehicle running status monitoring, diagnosis, and early warning is important means to ensure vehicle safety. Monitoring wheel-rail forces can find out various abnormalities of the train and its devices timely, then diagnose and predict, propose the necessary corrective measures. So as to realize the locomotive vehicle "condition base maintenance" which improve the using quality and efficiency of vehicles and their equipment. At present there are a lot of different business system of monitoring wheel/rail force, but majority just test tools for peak load, or needed great changes to existing rail structure so can be realized measurement and identify the wheel/rail force. Based on this situation, this paper mainly research and development a method of continuous measurement and recognition of the wheel rail vertical load without altering the track structure.This thesis starting from the railway analysis model, introduced railway analysis model response under train wheel load, detailed analysis and study rail stresses and deformations under vertical loading, which providing theoretical basis for effectively measuring wheel/rail vertical force. Subsequently introduce mechanics principle and implementation of various methods for measuring wheel-rail vertical force. The dependence and sensitivity of different methods for rail support status were analyzed, and disadvantages of and applicability of various methods were pointed out. By using finite element analysis software the finite element model of rail was established. Through loading and solving the model we got the responses of rail model nodes when subjected to a moving vertical load. Through analyzing the simulation results, we found the optimal node for arrangement of shear force sensors, and got the sensitivity and linearity of sensors’output. In order to verify the results of the finite element simulation analysis, an indoor laboratory loading experiment was implemented. Indoor test decorate three different schemes in the same rail span. From compare of the results, simulation and laboratory experiment show similar waveforms, but there are differences in amplitude. Then, the possible factors for these differences are analyzed.Wheel-rail vertical force measurement data may be interfered by baseline drift, and the baseline drift should be eliminated before amplitude analysis. this thesis studied the effects of curve fitting, mobile median filtering, mathematical morphology filter and wavelet threshold denoising method in removing the baseline drift, analyzed the disadvantages of different methods and put forward the improvement Suggestions and solutions. According to the characteristics of the wheel/rail vertical force signal, a new empirical method was proposed to remove baseline drift, which through clustering median values and variance of median values in data segments, and picking up the median values reflected baseline drift trends, then curve fitting the Selected points as the baseline. Compared to the simple curve fitting method, the number of basis points increase as much as possible, more detailed of the true baseline is expression. Compared to moving median filtering and morphological filtering, although the nature of these methods are based on order filter, but the proposed method Inhibit non-baseline noise interference to the extracted baseline with a sifting process, making it relative better performance to the others. Moreover, segmenting the data, making the calculation of whole process is greatly reduced. Then, a scheme through combining with wavelet filtering to suppress other interfering component contains in the extracted baseline was introduced, which obtained better eliminate effect.In last part of this thesis, a continuous measurement and identification method for wheel-rail vertical force was proposed. The method is based on the principle beam frame. The test period is divided into a limited number of rail beam element. Then we arrange multiple sets of observation point on the rail. Assuming that when the loading point locate anywhere in the beam position, the observation values of all observations points unchanged. While the load is applied on a different beam element, all observation values response vector must be different. So, by theoretical calculation method or field testing, standard response vector library of all observation points’response for constant loads moving along all of beam element can be obtained. The request of well-posedness of response vector library was validated by convert multiple observation points for single load location point to single observation point for multiple load location points. After obtained the standard response vector library, similarity measure was employed to observer the relationship between any observation vector and each vector in response vector library. In this thesis, a comparative study of three kind of multidimensional vector space similarity measure Euclidean distance, Angle cosine and Pearson correlation coefficients for the noise sensitivity, in order to select the appropriate measure for the membership function. Then uses the maximum membership degree recognition method finds the most similar vector in the standard library to current vector, and firmly believes that both vector responses from a load located at the same beam element. After getting load point location information, the ratio of vector-mode can be used to determine the amplitude of load. Then the method was validated through simulation data, when applied to field test data from line experiment, it achieved a very good recognition. Finally, a detailed analysis of possible factors caused the recognition method error was present.