A Study Of Recognition Of Subway Noise Signals Based On Machine Learning

With the world's population to focus on the city,the subway network has become the most effective large-scale transport system to solve urban congestion.However,the vibrations generated by the subway during travel will be transmitted to the surrounding buildings,as well as radiation noise through the building structure,affecting the environment and potentially creating safety problems.In view of the above,the technical and scientific community has made considerable efforts to try to have a better understanding of the problem in order to take control measures.In order to study the impact of subway vibration on the ground,you can use the sensor to the field for field measurement.After the measurement is completed,how to pick up and analyze the subway vibration signal in the whole signal is a more critical problem.The existing methods of manual processing,according to experience to measure,determine and deal with,but in the case of large amounts of data efficiency is very low.In this paper,a new method of subway vibration signal recognition based on machine learning is proposed for the effective part of the meteorological vibration signal measured in the field sensor.The main research contents include the following aspects:(1)The subway vibration signal recognition into a machine learning classification problem.Aiming at the problem of data imbalance in classification,the common under sampling method is used to solve the problem.It is proposed to use the random compression sampling method to ensure that each sample has the same number of features.The performance of different machine learning models for this problem is analyzed.Finally,a random forest algorithm is used to establish the classification model.(2)A sliding window algorithm is proposed to identify the vibration signal of the subway.The algorithm is run for a long time,and the sliding window algorithm is studied in parallel.It is proposed to implement the parallelization using Spark framework.Experiments show that the algorithm can improve the running speed after parallelization.(3)Developed a subway vibration signal identification system.Through the measurement data into the system,the system automatically completes the identification of the subway vibration signal,the operation process can observe the operation of the state,but also the results of the operation of the visual display.And the measured data of a subway line verify the effectiveness of the system.