| With the continuous development of railway,the traditional railway quality detection methods no longer meet the needs of high-speed and accurate development,and the detection technology based on line laser displacement is gradually becoming an important monitoring means of rail corrugation measurement.However,due to the influence of complex railway line measurement environment and random free vibration of car body,the accurate detection of rail corrugation has been greatly challenged.Based on the linear laser displacement sensor and its measurement platform,this paper studies the dynamic detection method of rail corrugation,focusing on the key problems and technical difficulties in realizing the dynamic detection of rail corrugation based on linear laser displacement.The main research contents and innovations are as follows:(1)A dynamic detection method of rail corrugation based on line laser displacement is proposed,and its basic process and principle are introduced.A large number of measured experiments on rail corrugation in indoor environment and outdoor real railway line are carried out by using the measurement platform built by line laser displacement sensor.Through the analysis and summary of the test results of the measured data,five main problems in the conventional detection process of rail corrugation based on linear laser displacement are found,including the loss of sampling data,the inclination of corrugation measurement data under nodding vibration,noise interference in sampling data registration,accurate restoration of long-distance corrugation data and extraction of wide-band corrugation value.The impact of these problems on the accurate detection of corrugation is emphatically analyzed,It lays a solid foundation for the research content carried out in the subsequent chapters.(2)In order to solve the problem of data loss caused by the reflection of the bright area on the rail surface and the interference of the external environment,a neural network based method for prediction and compensation of rail corrugation loss data is proposed in this paper.This method mainly includes three steps:data preprocessing such as missing data location and outlier data detection,training the prediction network based on LSTM and data filling.The performance test results show that this method has higher accuracy than the traditional filling method and the existing neural network prediction method,and can effectively solve the impact of missing data on the accuracy of rail corrugation measurement.At the same time,this filling method has good robustness under different loss rates.(3)Aiming at the problem of corrugation data tilt caused by vehicle body vibration when the laser plane of line laser displacement sensor is no longer perpendicular to the longitudinal direction of track,a blind identification and correction method of sensor tilt based on scanning line projection deformation is proposed in this paper.Firstly,without the asistant of external monitoring equipment,the scanning line state of the sensor in each sampling is identified by the blind identification method based on sampling data to judge whether the sensor has nodding vibration.Then,a mapping method is proposed to correct the skewed data set.Finally,the derivative dynamic time warping algorithm is used to match and calibrate the point sets with inconsistent size in the overlapping area,find the best matching point pair,and finally realize the accurate registration of adjacent sampling data sets.The experimental results show that the proposed recognition and correction method can accurately correct the data under nod vibration,and has good stability for different degrees of nodding vibration intensity.(4)Aiming at the interference of noise on repeated measurement interval data registration in complex measurement environment,a data registration method based on sub-trend is proposed in this paper.Firstly,the sampled rail corrugation data are decomposed and classified into long and short trends;then,starting from the actual data,the short trend and long trend data are de-noised by the adaptive filtering method based on fuzzy threshold.This filtering method can filter the noise data in the sampling data,so as to improve the similarity of repeated measurement interval data;finally,the long and short trend registration is carried out for the repeated measurement interval data in the adjacent sampling data.This method can reduce the registration interference of different trend data on the repeated measurement area data,and complete the overlap of adjacent sampling data through trend synthesis.The experimental results show that the proposed method can effectively solve the problem of low data registration accuracy in noise environment,greatly improve the accuracy of subsequent data splicing,and has high reliability for noise interference in different measurement environments.(5)In order to accurately restore all scattered sampling irregularity data to the complete rail corrugation curve representing the measurement section,a full surface rail corrugation restoration method based on layered splicing is proposed in this paper.The method is mainly divided into two parts.First,aiming at the data splicing problem between two adjacent sampling data,it is mainly realized through the data splicing process from coarse to fine.The coarse splicing is mainly used to locate the repeated measurement interval and ensure that the two data sets can overlap in this area.Then,the overlapping degree between the data sets is improved through the fine splicing method based on multiple population genetic algorithm,and the two data sets are combined;Secondly,a simple and feasible layered splicing strategy is proposed to reduce the cumulative error caused by splicing between a large number of segmented data sets,and realize the high-precision measurement of rail corrugation based on line laser displacement technology.The indoor and outdoor measurement experiments show that the proposed data splicing and corrugation restoration method can realize the high-precision and rapid measurement of rail corrugation and its detection performance has good engineering practicability.(6)Aiming at a series of problems faced by rail corrugation in wide-band measurement and wide-band corrugation value extraction,a wide-band rail corrugation value extraction method based on virtual scale is proposed in this paper.The extraction method of corrugation value proposed in this paper mainly includes three aspects:first,according to the existing corrugation evaluation system and the classification of rail corrugation in China,the wide band of rail corrugation based on line laser displacement as measurement technology is divided;Secondly,the virtual chord measurement model is proposed to restore the corrugation data.The advantage of this method is that it can extract the corrugation data of specific wavelength;Thirdly,the virtual scale method is proposed to calculate the corrugation value in a specific band range,which shows an excellent search ability of the maximum wave depth,so as to realize the task of extracting the corrugation value of each band.The experimental results show that the proposed corrugation value extraction method can achieve accurate classification of wave bands for corrugation measurement based on line laser displacement technology,and accurately extract the maximum wave depth,so as to provide accurate data support for subsequent corrugation evaluation and rail maintenance.To sum up,facing the real railway measurement scene,this paper studies the practical problems existing in the rail corrugation detection process based on line laser displacement,and puts forward the corresponding solutions.The proposed method can effectively improve the detection accuracy,reliability and wide-band detection ability of rail corrugation,and has great theoretical significance and practical value for promoting the efficient detection and intelligent level of rail corrugation. |
PDF Full Text Request