A Time-Frequency Extraction Model of Structural Vibration Combining VMD and HHT

  Objectives  Time and frequencies are important for investigating dynamic responses of offshore oil platforms. Time-frequency feature extraction is a key issue in structural health monitoring of large civil engineering structures. Current popular methods for signal analysis and processing cannot extract the characteristics of dynamic responses accurately and completely. Therefore, a new approach combining variational mode decomposition (VMD) and Hilbert-Huang transform (HHT) is proposed to extract multi-dimensional vibration characteristics of time, frequency and energy.

  Methods  Firstly, VMD is used to extract the vibration frequency component of events and the noise component could thus be removed. Then, HHT is applied to extract multi-dimensional vibration characteristics of time, frequency and energy using the cleaned accelerometer data. Finally, the frequency domain integration approach is introduced to compute the vibration displacements based on cleaned data processed by VMD. A series of simulation tests and ship impact test on offshore oil platform are performed using accelerometers in the MEMS IMU.

  Results  Experimental results show that the extraction method based on VMD-HHT can obtain the multi-dimensional time-frequency-energy characteristics accurately. The frequency domain integration approach can produce highly accurate and reliable vibration displacements using accelerometer data.

  Conclusions  VMD-HHT feature extraction model can not only extract the vibration frequency more completely and accurately, but also clearly see the time of vibration events, the corresponding frequency range and vibration intensity.VMD-HHT aided frequency domain integration approach can effectively remove the low-frequency noise and determine cut-off frequency quickly.