Research On Prediction Model Of Train-induced Vibration Database Based On Machine Learning Algorithm

Traditional methods for predicting environmental train-induced vibration are difficult to balance the needs of fast prediction speed,wide range and high accuracy.The application of machine learning and database technology to the field of environmental vibration of rail transit provides a new direction for the research of train-induced environmental vibration prediction methods.In this paper,a framework of metro traffic environmental vibration database system is constructed,and nested machine learning algorithms are used to achieve fast and accurate prediction of metro traffic environmental vibration.The major research contents and results are as follows.(1)Database system design.The main influencing factors of the prediction of traininduced environmental vibration are investigated through the literature.And the database target storage information should include working condition information,vibration response information,model information and test information.The design of the database structure was completed by using the Entity Relationship Diagram.(2)Database system development.By analyzing the functional requirements,a database application framework for train-induced environmental vibration was designed,and a database system was built using Django+My SQL+Vue+u WSGI to realize the scientific management of train-induced environmental vibration data.(3)The data samples of machine learning algorithm are obtained by using the sliced finite element-infinite element coupling model calculation.Firstly,the numerical calculation method of the sliced finite element-infinite element coupling model was verified based on the actual measurement cases,and then 38 sets of working conditions were designed based on the orthogonal test design method and random sampling method.And 228 sets of vibration data were calculated and used as the training set and test set of the machine learning algorithm model.(4)The prediction model of train-induced environment vibration neural network algorithm is constructed to achieve the prediction of the VLzmax.And the effects of optimizer,learning rate and the number of hidden layer neurons on the learning speed and prediction accuracy of the algorithm prediction model are investigated.(5)Establishing a database prediction model based on the machine learning algorithm.The machine learning algorithm prediction model is embedded in the database system,using JSON strings as the data format for front and back-end interaction,and using ORM and AJAX for data transfer and format conversion.Using these methods,the interaction between the machine learning algorithm prediction model and the visual operation of the database system were implemented.(6)Verification by actual measurement.The prediction results of the database prediction model under the corresponding conditions are compared and analyzed by selecting the actual measurement cases to further verify the prediction accuracy of the database prediction model on the VLzmax of the surface points.With a large amount of data support,the machine learning algorithm is used for traininduced environment vibration prediction,which can significantly reduce the modeling time and calculation time.The train-induced environment vibration database system can realize the scientific management of vibration data,as well as provide a visualization software ecology for machine learning algorithm prediction model and provide a platform support for the accumulation of datasets.The research in this paper shows that it is feasible to combine the database with machine learning algorithms for fast and accurate prediction of train-induced environmental vibration.