Prediction And Analysis Of In-Car Noise In Metro Area With Small Radius Curve Based On BP Neural Network

In recent years,domestic urban rail transit has entered the fast lane of development,but during its development,the problems of vibration and noise have become more and more serious.When a train passes a small radius curve section,it will produce relatively large wheel/rail noise,which will cause a large noise in the car,affect the passenger experience,and even cause people’s complaints in serious cases.This paper applies the principle of BP neural network to the prediction of in-car noise in curved sections for the first time,which has a guiding significance for the research on the prediction of in-car noise in urban rail transit.The analysis of the frequency spectrum characteristics of the noise in the car and the wheel/rail noise to determine the cause of the excessive noise in the car is also of reference value in the research of the wheel/rail noise.This paper establishes a wheel-rail noise calculation model for small-radius curve sections,calculates it for different working conditions,and uses measured data to verify the accuracy of the wheel-rail noise simulation model for small-radius curve sections.Later,the BP neural network was used to establish a noise prediction model in the car,and the correlation coefficient analysis method was used to study the correlation between the noise in the car and the wheel/rail noise,the radius of the curve,and the speed of the train,and to judge the degree of influence of various factors in the curve section on the noise in the car.Size and so on.The research content of this article is as follows:(1)Theoretical research on noise in subway cars and wheel-rail noise in curved sections.Through collecting and consulting relevant literature,we can understand the basic theories of noise in subway cars and wheel-rail noise,such as causes,propagation methods,and influencing factors.,Control and prediction methods,etc.(2)For a city’s Metro Line 7 in-car noise and wheel-rail noise synchronization test,refer to the subway noise evaluation specification,and formulate the early-stage in-car noise test plan for the small radius curve section of the subway,and the simultaneous test plan for the in-car noise and wheel-rail noise.,And carry out on-site measurements to analyze the overall noise level and frequency spectrum characteristics of the small radius curve car noise and wheel/rail noise,and determine the reason for the excessive noise in the car.(3)Establish a wheel-rail noise calculation model for small-radius curve sections,calculate for different working conditions,and use measured data to verify the calculation accuracy of the wheel-rail noise simulation model for small-radius curve sections.The simulation data can be used for subsequent noise prediction in the car,which can greatly reduce the cost of noise prediction in the car.(4)Process the data of in-vehicle noise and wheel-rail noise,and use the improved spectrum subtraction method based on multi-window spectrum estimation to process wheel-rail noise,thereby eliminating the influence of tunnel reverberation,which is the next subway in the small radius curve section.The research on the mapping relationship between noise and wheel-rail noise provides a large amount of accurate and complete wheel-rail noise data.(5)The BP neural network is used to establish a noise prediction model in the car.After the neural network is trained,the test set data is used to verify its calculation accuracy and prediction efficiency.In the following,combined with the simulation data calculated by using the wheel-rail noise model,the noise level in the vehicle can be predicted in the corresponding section.(6)Through data correlation analysis,the relationship between noise in the car and wheel/rail noise,train speed and curve radius can be determined.Combined with linear regression analysis,it is possible to determine the degree of influence of various factors on the noise in the car with a small radius curve.