Analysis And Assessment Methods Of Track Irregularities For Heavy-Hual/High-Speed Railways

Track regularity is important for operating safety of trains.However,due to the long-term service and environmental factors,the track will inevitably be aged,worn and even damaged,resulting in track irregularity.This threatens operating safety of trains.Traditional methods use time-domain amplitude of track irregularity to judge track state,such as amplitude overlimit standard,track quality index and so on.With the rapid development of heavy-haul and high-speed railways in China,higher requirements are put forward for operating safety.It is urgent to improve methods for track irregularity analysis and evaluation to guide track maintenance more scientifically.Therefore,this paper investigates the analysis and evaluation methods of track irregularity of heavy-haul and high-speed railways.The main works and contributions are as follows:(1)Characterization of track irregularity spectrum for heavy-haul railwayTrack spectrum is an effective method to evaluate track irregularity of the whole line.In recent 10 years,China has established a general track standard spectrum for ordinary and high-speed railways.However,there is still a lack of systematic research on heavy-haul track spectrum,and the track standard spectrum has not been promulgated.Based on the measured data of Datong-Qinhuangdao and Shuozhou-Huanghua railways,this paper calculates the track spectra and compares them with the typical track spectra at home and abroad.Then,this paper proposes the track standard spectrum expression and its parameters of heavy-haul railways.Finally,in view of the problem that the standard spectrum can only simulate the random track irregularity,but cannot simulate the periodic irregularity,this paper proposes the expression and simulation of the periodic irregularity.The research shows that the proposed standard spectral expression is better than the typical track spectral expressions at home and abroad to characterize the track spectral characteristics of Chinese heavy-haul railways;the combined simulation of randomness and periodicity proposed in this paper makes the simulation of track irregularity more reasonable.(2)Prediction of vehicle-body vibration accelerations from track irregularities based on CNN-GRU modelThe vehicle-body vibration acceleration is an important index to indirectly evaluate track regularity state.Based on track inspection data of high-speed railways,this paper uses Gated Recurrent Unit(GRU)to model sequential data,and combines Convolution Neural Network(CNN)to extract waveform features.Finally,the CNN-GRU model is proposed to predict vehicle-body vibration accelerations,and the best performance and applicability of the model are studied.The results show that: The performance of the model is the best with the combined input of 8 track irregularities and vehicle speed;the accuracy of CNN-GRU is better than BP neural network,GRU and long short term memory(LSTM),slightly better than CNN-LSTM,but its parameter amount is 1.33 times less than CNN-LSTM;CNN-GRU is suitable for situations such as vehicle speed changes,different lines and different sub-rail foundations.This work and results provide support for the establishment of a track state evaluation method based on multi-vehicle CNN-GRU models.(3)Track state evaluation method based on multi-vehicle CNN-GRU modelsDue to the different dynamic transfer characteristics of different vehicle types,the evaluation results of vehicle-body vibration accelerations of different vehicle types on the same track state are different.In order to improve the reliability of track state evaluation,this paper proposes a track state evaluation method based on multi-vehicle CNN-GRU models on the basis of the previous content.The experimental results show that the problem of inaccurate evaluation of single vehicle can be effectively solved by using multi-vehicle models to comprehensively evaluate the track state.(4)Estimation of track longitudinal irregularities from vehicle-body accelerations based on AM-CNN-Bi GRU modelAt present,countries all over the world use comprehensive inspection vehicles to inspect track irregularities of high-speed railways,but the comprehensive inspection vehicles have a long inspection cycle and occupy operating resources.In order to use operating vehicles to monitor track status frequently,this paper combines CNN and Bi-directional GRU(Bi GRU)on the basis of aforementioned deep learning model(CNN-GRU),and introduces Attention Mechanism(AM),and finally proposes AM-CNN-Bi GRU model to estimate track longitudinal irregularities from vehicle-body accelerations.The applicability of the model in practice is also evaluated.The research shows that the model can effectively estimate track longitudinal irregularities,and is suitable for vehicle speed changes,different lines,different sub-rail foundations and different vehicle types.