Intelligent monitoring of vibration and structural-borne noise induced by rail transit

Abstract

Abstract The monitoring of urban rail transit vibration and structural-borne noise can well solve the problem of small amount of data and large discreteness in traditional evaluation tests. In this study, a monitoring system is utilized to collect and preprocess vibration and noise signals. By employing cellular network and cloud-based services, real-time acquisition and analysis of vibration and noise signals are achieved. In this article, it is proposed to normalize the vibration data obtained after edge computing. After treatment, the gray correlation analysis method was used to determine the correlation between each vibration data component and vibration data classification. Combining frequency domain analysis of vibration data, the data components with high correlation are used as inputs to an improved K-nearest neighbours (KNN) model. Additionally, the correlation of each data component is introduced into the distance calculation formula. The improved KNN model shows improvements in recall rate, precision rate, F-measure, and accuracy compared to the original KNN model, with increases of 0.76 per cent, 2.76 per cent, 1.81 per cent, and 1.61 per cent respectively. Through practical measurements, it is found that different vehicles cause significant variations in vibration, with differences of up to 11 dB in tunnel wall vibration. The differences in tunnel wall vibration caused by the same vehicle at different passenger loads do not exceed 5 dB. Combining practical application cases, the rail transit environmental noise monitoring system established in this study demonstrates its applicability in monitoring vibration and noise-sensitive areas, as well as analysing the effectiveness of vibration reduction and noise control measures.