Linear Vibration Identification Using Machine Learning Approaches

As Machine Learning (ML) techniques have been modified and more mainstream, the appli-cations of these techniques are increasing in many scientific and industrial fields. Although vibration-based structure identification techniques have been improved upon to diagnose fail-ure symptoms or identify system parameters, new methods based on machine learning have been studied to increase their performances. This research aims to present an innovative appli-cation of machine learning in structure identification. An aluminum cantilever beam that is randomly excited by using an electrodynamic shaker was selected as the case study example to prove the methodology experimentally. By using a combination of ML-based regression and classification techniques, the vibration responses are measured at different points to identify the beam natural frequencies. The estimated results are validated using the Fast Fourier Trans-form (FFT) and Frequency Domain Decomposition (FDD) methods. The results show that us-ing the proposed ML-based technique can present a new output-only identification method to identify the system accurately.

کلیدواژه ها:

System Identification ، Machine Learning ، Time domain Response.