Study On Fatigue Damage Prediction Method Of Metro Frame Based On Working Condition Recognition

In recent years,with the rapid rise of China’s economic level and the accelerating process of urbanization,a large number of subway lines have been laid in every big city to relieve the traffic pressure on the ground.With the continuous increase of the mileage of subway trains,a large number of tracking tests have been carried out at home and abroad,and the damage evolution law of key parts of the frame has been preliminarily proved.However,the influence of different line conditions and train running states on the damage of key parts of the frame still needs to be further studied.In this paper,a B-type subway bogie frame is taken as the research object,and the dynamic stress tracking test of the frame is carried out in Beijing subway.Relying on a large number of measured data,an analysis software platform integrating data management and data processing is built.On this basis,the identification methods of subway train working conditions are studied,and the transfer relationship between the combined working condition parameters and the damage of key parts of the frame is calculated according to the identification method.Finally,the prediction method of the frame damage is established.Main work contents include:(1)The types of data generated in the process of dynamic stress testing are studied.Combined with the characteristics of each type of data,the test data is managed according to the experiment and test levels.Meanwhile,corresponding storage templates are proposed for different types of data.At the same time,the process flow of measured data is analyzed,the algorithm of the main links in the data pre-processing and post-processing is deeply analyzed,and new improved algorithms are proposed in view of the disadvantages of the traditional algorithms.(2)Based on the study of data management method and processing optimization algorithm,C#.NET is used to develop the dynamic stress test data analysis platform.After running and testing,the goal of rapid extraction and efficient processing from massive measured data is realized,and the effective connection between data management and data processing is completed.(3)The working conditions of subway trains are divided,and the train running acceleration is extracted by using the measured speed signal and the least squre linear fitting method based on LU decomposition,which is used as the characteristic identification parameter of running conditions.The identification interval of traction,braking,coasting conditions is studied.In this paper,two types of models,data-driven model and program discriminant model,are used to identify straight line and curve conditions in the main line of subway,and the identification results are integrated by voting method.Finally,the effect of running condition and line condition recognition is evaluated by combining with the measured data.(4)The equivalent stress distribution of the frame measuring points under the same working condition parameters is studied,and the minimum sample size of the same equivalent stress for damage analysis is determined.Based on this,the combined working condition-damage sample set is established.Based on BP neural network,the transmission model between the combined working condition parameters and frame damage is trained,and the genetic algorithm is used to optimize the model training process.Finally,the specified line section is identified and divided,multiple sets of identification parameters are input into the model to obtain the damage prediction value,compared with the measured damage value of the section,and the prediction effect is relatively ideal.The research content of this paper lays a foundation for the establishment of high precision load spectrum according to different working conditions,and provides a new idea for the damage prediction reasearch of the bogie frame of the same platform in a new line.80 figures,26 tables,83 references.