Evaluation of subway vibration reduction effect based on ultra-weak FBG array using deep forest

Abstract

Monitoring the train vibration is an important issue in subway safety management and maintenance. Aiming at the problems of traditional technology for detection and acquisition of subway vibration, such as unable to achieve dynamic detection or evaluation, and unable to give early warning to the changes of subway tunnel structure, this paper proposes a method to obtain and predict the vibration reduction effect based on grating array. The method uses short-time power spectral density (PSD) to extract train vibration signal, and uses Z-vibration level (VLz) to obtain the vibration reduction effect of subway track. A model based on deep forest (DF) is improved to predict the variation trend of vibration reduction effect. The experimental results on the actual train data illustrate that the proposed method can accurately extract the train vibration signal, and the model can effectively forecast the vibration reduction effect, which has a lower error precision and shows improved performance over other prediction models.