A VIBRATION RECOGNITION METHOD BASED ON DEEP LEARNING AND SIGNAL PROCESSING

作     者：CHENG Zhi-gang LIAO Wen-jie CHEN Xing-yu LU Xin-zheng CHENG Zhi-gang;LIAO Wen-jie;CHEN Xing-yu;LU Xin-zheng

作者机构：Civil Engineering DepartmentTsinghua UniversityBeijing 100084China Beijing Engineering Research Center of Steel and Concrete Composite StructuresTsinghua UniversityBeijing 100084China Key Laboratory of Civil Engineering Safety and Durability of China Education MinistryTsinghua UniversityBeijing 100084China 

出 版 物：《工程力学》 (Engineering Mechanics)

年 卷 期：2021年第38卷第4期

页      面：230-246页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：山东省高校土木结构防灾减灾协同创新中心基金资助项目 清华大学大学生研究训练计划项目(2011T0017) 

主　　题：vibration recognition signal processing time-frequency-domain characteristics convolutional neural network(CNN) long short-term memory(LSTM)network 

摘      要：Effective vibration recognition can improve the performance of vibration control and structural damage detection and is in high demand for signal processing and advanced ***-processing methods can extract the potent time-frequency-domain characteristics of signals;however,the performance of conventional characteristics-based classification needs to be *** used deep learning algorithms(e.g.,convolutional neural networks(CNNs))can conduct classification by extracting high-dimensional data features,with outstanding ***,combining the advantages of signal processing and deep-learning algorithms can significantly enhance vibration recognition performance.A novel vibration recognition method based on signal processing and deep neural networks is proposed ***,environmental vibration signals are collected;then,signal processing is conducted to obtain the coefficient matrices of the time-frequency-domain characteristics using three typical algorithms:the wavelet transform,Hilbert-Huang transform,and Mel frequency cepstral coefficient extraction ***,CNNs,long short-term memory(LSTM)networks,and combined deep CNN-LSTM networks are trained for vibration recognition,according to the time-frequencydomain ***,the performance of the trained deep neural networks is evaluated and *** results confirm the effectiveness of the proposed vibration recognition method combining signal preprocessing and deep learning.