Research On The Evaluation Of Railway Track Quality State Based On Dynamic Detection Data

With the continuous development of railway technology,the detection and analysis of railway track smoothness have gradually attracted people’s attention.A good railway track quality can ensure the safety of the railway and provide passengers with a comfortable riding experience.Based on the track inspection vehicle,it can carry out dynamic safety inspection of track quality,which can not only provide data support and decision-making plan for the safety management of track quality,but also can dig out the uneven development law of track by referring to a large amount of inspection data.Therefore,how to accurately obtain track morphology data and use the detection data to effectively guide on-site maintenance is becoming a key topic for domestic and foreign researchers.In the railway track inspection environment,the inspection equipment is affected by sunlight interference and equipment failure factors to produce some abnormal values,which will not only affect the accurate evaluation of the quality status of the railway track,but also affect the follow-up track quality status prediction research.At the same time,in the prediction research of the orbit quality index,the final prediction accuracy of the model is not ideal due to factors such as the weight ratio of the detection data and the background value,and the random adaptability is low.In order to solve the above problems,this article has carried out the following researches.(1)Research on the correction processing of detected data abnormal value.The research on the correction of abnormal value of track detection data can be divided into the correction of abnormal value of sampling point and the correction of abnormal mileage of sampling point.Aiming at the research on the correction of the numerical anomaly of the sampling point,this paper adopts the numerical correction model of the sampling point of the combined method.By combining the Rheinda criterion in the gross error elimination theory and the absolute mean method,the effective identification and correction of the numerical value of the sampling point is realized.The experimental results show that: in terms of root mean square error,the outlier correction model based on the combination method reduces11.7% and 9.35% on average compared with the wavelet change method and the change rate correction method;For the research on mileage abnormal value of track detection data,this paper constructs a combined correction model of mileage abnormality based on Lagrangian piecewise interpolation method,through preliminary mileage correction-data segmentation processing-segment data interpolation-improved DTW algorithm mileage abnormal accuracy A series of steps of correction processing,the accuracy of the mileage abnormality correction can reach 0.25 m,and the correlation coefficient ρ＿ij after the mileage abnormality correction can reach 0.98 on average.This effectively solves the problem of abnormal mileage at the track inspection data sampling point,and facilitates subsequent inspection data quality evaluation.Forecast research.(2)Construction of TQI prediction model based on optimized non-equal interval gray model.In order to solve the problem of unsatisfactory prediction results in the TQI prediction model due to factors such as the weight ratio of the detection data,the background value and other factors,this paper proposes a gray model that optimizes the non-equal time interval combined with the TQI prediction model of the BP neural network.The time interval gray model is optimized from three perspectives: background value construction,initial value selection,and detection data weight,which improves the prediction accuracy of the trend component of the track TQI prediction.Aiming at the random fluctuation component in the orbit TQI prediction,this paper adopts the BP neural network model.With its good fusion and learning characteristics,it effectively corrects the residual sequence in the detection data,improves the prediction accuracy of the final model,and realizes Good track quality life prediction.The experiment proves that the relative error of the orbit TQI prediction model constructed in this paper is reduced by 18.2% and 15.7% respectively compared with the improved gray and PSVM model and GM＿GA＿Elman model.The validity and robustness of the model is verified,and the preventive maintenance decision plan of railway tracks can be realized based on the model.(3)Design and development of a visual track inspection data management platform.Based on the research on the evaluation of railway track quality status based on dynamic inspection data,the design of a visual track inspection data management platform is realized.According to the application scenarios of different lines,real-time track dynamic data collection,historical data processing and analysis,and current track geometric irregularity data prediction can be realized,and the original detection data,data processing results and track geometric irregularity prediction can be realized through the cloud data management platform The storage management of the results is convenient for users to query and analyze at any time.