Noise Prediction And Sound Quality Evaluation Of Rail Transit Box-girder Structure

As the speed of high-speed trains continues to increase,and people’s movement becomes more convenient.However,with the development of rail transit,some new problems have been brought.When the high-speed train runs on the bridge,the bridge structure noise will be generated.This noise is a low-frequency noise,which has many effects on the body’s physical health,reflection time and language recognition capabilities.Sound quality is an important standard of acoustic environment.This article focuses on the sound quality of railway traffic box girder structures.The physical properties of noise are studied through the simulation of the sound field,and subjective experiments are used to record and analyze people’s subjective feelings,so as to better analyze and evaluate the sound quality of structural noise from various aspects,analyze the psychoacoustic properties of noise,and obtain subjective evaluation.As a result,on this basis,the sound quality evaluation model,the BP neural network evaluation model,which is the same as the subjective evaluation results,can be constructed to better help people predict their subjective perception of structural noise.(1)Predict the structural noise of box girder.The box girder model was established according to the finite element-statistical energy method,and the main excitation of the box girder during train operation was determined to be wheel-rail force.The wheel-rail force was loaded on the box girder model to calculate the sound pressure level of the box girder structure noise;In the experiment,the structure noise of the box girder when the train is running is collected,the physical properties of the noise are analyzed,and the noise is compared with the predicted noise.The comparison results show that the predicted value is very close to the experimental value,which verifies the validity of the sound field simulation results,thus proving the sound field Simulation predicts the feasibility of this measure;through experiments,it can also be seen that the sound pressure level of box girder structure noise reaches the maximum in the 50Hz～100Hz frequency band,and as the train speed increases,the sound pressure level gradually increases.(2)Subjective and objective analysis of sound quality.In the objective aspect,various parameters of the psychoacoustic attributes are calculated in detail,including loudness,sharpness,roughness and volatility.These parameters are used to describe the objective attributes of sound quality;in the subjective aspect,the evaluation is carried out by subjective evaluation test,and the level is adopted.The scoring method is used to evaluate the test personnel and obtain the result called the subjective evaluation value.Through multiple linear regression analysis,determine the relationship between the four psychoacoustic parameters and the sound pressure level and the subjective evaluation results.(3)Establish sound quality BP neural network evaluation model.First,the four psychoacoustic parameters in the sample are used as input values,and the subjective evaluation values are used as output values to establish a neural network model.Then through a series of randomly selected training data,continuous training and then optimize and determine the network structure and related parameters,so as to predict the results of subjective evaluation;calculate the psychoacoustic parameters of different samples,and bring them into the established BP neural network In the evaluation model,the sound quality prediction value is obtained and compared with the corresponding subjective evaluation value.It is found that its predictive ability is strong and can be accurately predicted;by changing the box girder section size and damping loss factor,the corresponding subjective evaluation value is obtained Value:Using the BP neural network model can skip the process of collecting experiments,predict the results of psychoacoustic properties through sound field simulation,and then directly predict subjective feelings through the connection of the neural network,making the sound quality prediction more efficient.